Virtualization - Running RTOS and UNIX on one machine - Linux

This is a discussion on Virtualization - Running RTOS and UNIX on one machine - Linux ; Hi, It seems a very large marketplace for the virtualization in embedded field. Many works which need to be done by 2 or more processors can be done by only one processor if adopt a technique that allow a supervisor ...

+ Reply to Thread
Results 1 to 12 of 12

Thread: Virtualization - Running RTOS and UNIX on one machine

  1. Virtualization - Running RTOS and UNIX on one machine

    Hi,

    It seems a very large marketplace for the virtualization in embedded
    field. Many works which need to be done by 2 or more processors can be
    done by only one processor if adopt a technique that allow a
    supervisor software to control both a UNIX OS(typically Linux) and a
    RTOS(or more?).

    I am very interested in this technology and I know there has been a
    lot of startup in this fields, such as http://www.virtuallogix.com/.
    And there are many open source virtualization project, such as XEN and
    OpenVZ, but I can't find a project for running a general purpose OS
    plus RTOS.

    Because RTOS has some constraints on timing, so I think there should
    be many differences between existing open source general purpose
    virtualization techniques and the commercial RTOS vitualization
    techniques.

    I really don't have enough experience on RTOS, so can anybody who are
    good at this can comment for is it very hard to make this done?

    Thanks for your comments!

    Bin


  2. Re: Virtualization - Running RTOS and UNIX on one machine

    On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    wrote in
    <1183302769.974254.151050@i38g2000prf.googlegroups. com>:

    >Hi,
    >
    >It seems a very large marketplace for the virtualization in embedded
    >field. Many works which need to be done by 2 or more processors can be
    >done by only one processor if adopt a technique that allow a
    >supervisor software to control both a UNIX OS(typically Linux) and a
    >RTOS(or more?).
    >
    >I am very interested in this technology and I know there has been a
    >lot of startup in this fields, such as http://www.virtuallogix.com/.
    >And there are many open source virtualization project, such as XEN and
    >OpenVZ, but I can't find a project for running a general purpose OS
    >plus RTOS.
    >
    >Because RTOS has some constraints on timing, so I think there should
    >be many differences between existing open source general purpose
    >virtualization techniques and the commercial RTOS vitualization
    >techniques.
    >
    >I really don't have enough experience on RTOS, so can anybody who are
    >good at this can comment for is it very hard to make this done?
    >
    >Thanks for your comments!
    >
    >Bin


    I dived into the virtualization some year ago, and was disappointed by the performance.
    I can see use of it to sell virtual servers for customers...
    But for embedded where real time is important, and resources are very limited,
    I see no advantages in this over say an FPGA with 2 or more processors.
    Doing the parts that need to be fast in FPGA as gate operations
    or state machines, and having one or more processors for sequential code,
    seems to me to satisfy both real time and complex code requirements.
    So I am a non-believer in virtualization for small -possibly embedded- systems.
    If you go your way, there is real time Linux, but real time Linux under a virtualizer
    seems a joke to me.
    If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    Never tried that.

    I will try to explain some aspects of this.
    If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    have as requirement that when input pin of the par port goes high, then
    within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    then you have a huge problem.
    We must then go back to basic understanding of the hardware, sequential processing
    or use simple logic.
    Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    to carry a specific signal from one pin to an other in usually about 1 clock tick.
    This happens as a separate circuit, an completely different 'machine' withing the same
    gate array, while the other part of the thousands of gates available can for example
    form a sequential system, a processor, running Linux if must be.
    Never ever can you beat this with virtualizing.
    never.
    Those who come from a hardware background will know what I am talking about.
    Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    of heat, and those solutions will be slower and less reliable.

    So use what is best for the requirement you have.
    It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make the real time requirement
    if the same can be done with a few milliwatt logic gates.
    Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    description languages), I most fortunately have a hardware background, played with gates long
    before there were _any_ computers for sale.
    In 1979 I digitised my first video signal ... designed the digitiser.
    The max processor speed available at that time was just a few MHz (less then 10).
    Think how that can be.
    I did not use a processor :-)
    LOL
    Even then some Unix people were big ego boasters.
    You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    Or at least basic electronics.

    El Pante

    Copyright El Pante 2007-always
    Reading and understanding is required.
    All rights reserved, non OSI compliant.
    You can quote El Pante on this.


  3. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 2, 12:37 am, Jan Panteltje wrote:
    > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > wrote in
    > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:
    >
    >
    >
    > >Hi,

    >
    > >It seems a very large marketplace for the virtualization in embedded
    > >field. Many works which need to be done by 2 or more processors can be
    > >done by only one processor if adopt a technique that allow a
    > >supervisor software to control both a UNIX OS(typically Linux) and a
    > >RTOS(or more?).

    >
    > >I am very interested in this technology and I know there has been a
    > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > >And there are many open source virtualization project, such as XEN and
    > >OpenVZ, but I can't find a project for running a general purpose OS
    > >plus RTOS.

    >
    > >Because RTOS has some constraints on timing, so I think there should
    > >be many differences between existing open source general purpose
    > >virtualization techniques and the commercial RTOS vitualization
    > >techniques.

    >
    > >I really don't have enough experience on RTOS, so can anybody who are
    > >good at this can comment for is it very hard to make this done?

    >
    > >Thanks for your comments!

    >
    > >Bin

    >
    > I dived into the virtualization some year ago, and was disappointed by the performance.
    > I can see use of it to sell virtual servers for customers...
    > But for embedded where real time is important, and resources are very limited,
    > I see no advantages in this over say an FPGA with 2 or more processors.
    > Doing the parts that need to be fast in FPGA as gate operations
    > or state machines, and having one or more processors for sequential code,
    > seems to me to satisfy both real time and complex code requirements.
    > So I am a non-believer in virtualization for small -possibly embedded- systems.
    > If you go your way, there is real time Linux, but real time Linux under a virtualizer
    > seems a joke to me.
    > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > Never tried that.
    >
    > I will try to explain some aspects of this.
    > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > have as requirement that when input pin of the par port goes high, then
    > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > then you have a huge problem.
    > We must then go back to basic understanding of the hardware, sequential processing
    > or use simple logic.
    > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > This happens as a separate circuit, an completely different 'machine' withing the same
    > gate array, while the other part of the thousands of gates available can for example
    > form a sequential system, a processor, running Linux if must be.
    > Never ever can you beat this with virtualizing.
    > never.
    > Those who come from a hardware background will know what I am talking about.
    > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > of heat, and those solutions will be slower and less reliable.
    >
    > So use what is best for the requirement you have.
    > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make the real time requirement
    > if the same can be done with a few milliwatt logic gates.
    > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > description languages), I most fortunately have a hardware background, played with gates long
    > before there were _any_ computers for sale.
    > In 1979 I digitised my first video signal ... designed the digitiser.
    > The max processor speed available at that time was just a few MHz (less then 10).
    > Think how that can be.
    > I did not use a processor :-)
    > LOL
    > Even then some Unix people were big ego boasters.
    > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > Or at least basic electronics.
    >

    Thanks for your comment, but AFAIK, your case is only a type of real
    time, many other real time system doesn't rely on hardware like this,
    but is constraint by protocol timing. Such as if a protocol says you
    must respond a ACK within 5 ms, otherwise the other end treat this
    failure.

    Many handset has a architecture of two or more cores. Most smartphone
    or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    to processing applications, another core to processing wireless
    protocol. They connect with each other with some common interface,
    such as SPI, UART etc.

    Man, embedded is a broader concept now, not only a device with 8-bit
    micro processor now, many PDA has a 600Mhz processor but still is
    regarded as "embedded".

    Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    the main application processor, the cost is greatly cut down. And it
    is viable. Take GSM as an example, most time the GSM is idle and when
    GSM is working, we can yield most the CPU power to it. Most GSM
    operation only sustain for several msec, but a default Linux time
    slice is 100ms!! So I think this is absolutely doable.

    Thanks.
    Bin



  4. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 2, 3:41 am, Bin Chen wrote:
    > On Jul 2, 12:37 am, Jan Panteltje wrote:
    >
    >
    >
    > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > wrote in
    > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > >Hi,

    >
    > > >It seems a very large marketplace for thevirtualizationin embedded
    > > >field. Many works which need to be done by 2 or more processors can be
    > > >done by only one processor if adopt a technique that allow a
    > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > >RTOS(or more?).

    >
    > > >I am very interested in this technology and I know there has been a
    > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > >plus RTOS.

    >
    > > >Because RTOS has some constraints on timing, so I think there should
    > > >be many differences between existing open source general purpose
    > > >virtualizationtechniques and the commercial RTOS vitualization
    > > >techniques.

    >
    > > >I really don't have enough experience on RTOS, so can anybody who are
    > > >good at this can comment for is it very hard to make this done?

    >
    > > >Thanks for your comments!

    >
    > > >Bin

    >
    > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > I can see use of it to sell virtual servers for customers...
    > > But for embedded wherereal timeis important, and resources are very limited,
    > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > Doing the parts that need to be fast in FPGA as gate operations
    > > or state machines, and having one or more processors for sequential code,
    > > seems to me to satisfy bothreal timeand complex code requirements.
    > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > seems a joke to me.
    > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > Never tried that.

    >
    > > I will try to explain some aspects of this.
    > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > have as requirement that when input pin of the par port goes high, then
    > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > then you have a huge problem.
    > > We must then go back to basic understanding of the hardware, sequential processing
    > > or use simple logic.
    > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > gate array, while the other part of the thousands of gates available can for example
    > > form a sequential system, a processor, running Linux if must be.
    > > Never ever can you beat this with virtualizing.
    > > never.
    > > Those who come from a hardware background will know what I am talking about.
    > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > of heat, and those solutions will be slower and less reliable.

    >
    > > So use what is best for the requirement you have.
    > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > if the same can be done with a few milliwatt logic gates.
    > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > description languages), I most fortunately have a hardware background, played with gates long
    > > before there were _any_ computers for sale.
    > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > The max processor speed available at that time was just a few MHz (less then 10).
    > > Think how that can be.
    > > I did not use a processor :-)
    > > LOL
    > > Even then some Unix people were big ego boasters.
    > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > Or at least basic electronics.

    >
    > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > but is constraint by protocol timing. Such as if a protocol says you
    > must respond a ACK within 5 ms, otherwise the other end treat this
    > failure.
    >
    > Many handset has a architecture of two or more cores. Most smartphone
    > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > to processing applications, another core to processing wireless
    > protocol. They connect with each other with some common interface,
    > such as SPI, UART etc.
    >
    > Man, embedded is a broader concept now, not only a device with 8-bit
    > micro processor now, many PDA has a 600Mhz processor but still is
    > regarded as "embedded".
    >
    > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > the main application processor, the cost is greatly cut down. And it
    > is viable. Take GSM as an example, most time the GSM is idle and when
    > GSM is working, we can yield most the CPU power to it. Most GSM
    > operation only sustain for several msec, but a default Linux time
    > slice is 100ms!! So I think this is absolutely doable.
    >
    > Thanks.
    > Bin- Hide quoted text -
    >
    > - Show quoted text -


    There are real-time hypervisors from various companies (available or
    at least announced):

    Depending on processor architecture you have at least the following
    companies to choose from:

    ARM etc.
    Trango www.trango-vp.com
    Virtuallogix www.virtuallogix.com mentioned in the first
    post (also PPC and maybe more)

    x86:
    Tenasys www.tenasys.com
    Real-Time Systems www.real-time-systems.com

    If singlecore processors are used you will always have a performance
    hit for context switching between the OSes and you (or the respective
    vendors) might run into patent issues with e.g. IBM and the well known
    RT-Linux patent (now owned by Wind River).

    On multicore architectures you have the advantage of much better
    interrupt latencies and you can reboot individual OSes (cores) anytime
    during operation - but your hardware gets more expensive.

    Regards,

    Gerd






  5. Re: Virtualization - Running RTOS and UNIX on one machine

    On a sunny day (Mon, 02 Jul 2007 01:41:34 -0000) it happened Bin Chen
    wrote in
    <1183340494.626702.322280@x35g2000prf.googlegroups. com>:

    >So I think this is absolutely doable.


    Then do it.
    Why ask?

  6. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:
    > On Jul 2, 3:41 am, Bin Chen wrote:
    >
    >
    >
    > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > wrote in
    > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > >Hi,

    >
    > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > >field. Many works which need to be done by 2 or more processors can be
    > > > >done by only one processor if adopt a technique that allow a
    > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > >RTOS(or more?).

    >
    > > > >I am very interested in this technology and I know there has been a
    > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > >plus RTOS.

    >
    > > > >Because RTOS has some constraints on timing, so I think there should
    > > > >be many differences between existing open source general purpose
    > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > >techniques.

    >
    > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > >good at this can comment for is it very hard to make this done?

    >
    > > > >Thanks for your comments!

    >
    > > > >Bin

    >
    > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > I can see use of it to sell virtual servers for customers...
    > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > or state machines, and having one or more processors for sequential code,
    > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > seems a joke to me.
    > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > Never tried that.

    >
    > > > I will try to explain some aspects of this.
    > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > have as requirement that when input pin of the par port goes high, then
    > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > then you have a huge problem.
    > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > or use simple logic.
    > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > gate array, while the other part of the thousands of gates available can for example
    > > > form a sequential system, a processor, running Linux if must be.
    > > > Never ever can you beat this with virtualizing.
    > > > never.
    > > > Those who come from a hardware background will know what I am talking about.
    > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > So use what is best for the requirement you have.
    > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > if the same can be done with a few milliwatt logic gates.
    > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > before there were _any_ computers for sale.
    > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > Think how that can be.
    > > > I did not use a processor :-)
    > > > LOL
    > > > Even then some Unix people were big ego boasters.
    > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > Or at least basic electronics.

    >
    > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > but is constraint by protocol timing. Such as if a protocol says you
    > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > failure.

    >
    > > Many handset has a architecture of two or more cores. Most smartphone
    > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > to processing applications, another core to processing wireless
    > > protocol. They connect with each other with some common interface,
    > > such as SPI, UART etc.

    >
    > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > micro processor now, many PDA has a 600Mhz processor but still is
    > > regarded as "embedded".

    >
    > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > the main application processor, the cost is greatly cut down. And it
    > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > operation only sustain for several msec, but a default Linux time
    > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > Thanks.
    > > Bin- Hide quoted text -

    >
    > > - Show quoted text -

    >
    > There are real-time hypervisors from various companies (available or
    > at least announced):
    >
    > Depending on processor architecture you have at least the following
    > companies to choose from:
    >
    > ARM etc.
    > Trango www.trango-vp.com
    > Virtuallogix www.virtuallogix.commentioned in the first
    > post (also PPC and maybe more)
    >
    > x86:
    > Tenasys www.tenasys.com
    > Real-Time Systems www.real-time-systems.com
    >
    > If singlecore processors are used you will always have a performance
    > hit for context switching between the OSes and you (or the respective
    > vendors) might run into patent issues with e.g. IBM and the well known
    > RT-Linux patent (now owned by Wind River).
    >
    > On multicore architectures you have the advantage of much better
    > interrupt latencies and you can reboot individual OSes (cores) anytime
    > during operation - but your hardware gets more expensive.
    >

    Thank you very much!
    Do you have more infomation on how they implement this? I want it
    eagerly. RTOS is simple one but the important part is the software
    built on RTOS. I want to know how many ways can I try to achieve this,
    and has any reference already?

    Software such as XEN is quite large, and at the first glance it can't
    fulfill my requirement. May be I can start a new open source project
    if I know exactly how to do it

    Thanks.
    Bin



  7. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 2, 1:52 pm, Bin Chen wrote:
    > On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:
    >
    >
    >
    > > On Jul 2, 3:41 am, Bin Chen wrote:

    >
    > > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > > wrote in
    > > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > > >Hi,

    >
    > > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > > >field. Many works which need to be done by 2 or more processors can be
    > > > > >done by only one processor if adopt a technique that allow a
    > > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > > >RTOS(or more?).

    >
    > > > > >I am very interested in this technology and I know there has been a
    > > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > > >plus RTOS.

    >
    > > > > >Because RTOS has some constraints on timing, so I think there should
    > > > > >be many differences between existing open source general purpose
    > > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > > >techniques.

    >
    > > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > > >good at this can comment for is it very hard to make this done?

    >
    > > > > >Thanks for your comments!

    >
    > > > > >Bin

    >
    > > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > > I can see use of it to sell virtual servers for customers...
    > > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > > or state machines, and having one or more processors for sequential code,
    > > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > > seems a joke to me.
    > > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > > Never tried that.

    >
    > > > > I will try to explain some aspects of this.
    > > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > > have as requirement that when input pin of the par port goes high, then
    > > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > > then you have a huge problem.
    > > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > > or use simple logic.
    > > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > > gate array, while the other part of the thousands of gates available can for example
    > > > > form a sequential system, a processor, running Linux if must be.
    > > > > Never ever can you beat this with virtualizing.
    > > > > never.
    > > > > Those who come from a hardware background will know what I am talking about.
    > > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > > So use what is best for the requirement you have.
    > > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > > if the same can be done with a few milliwatt logic gates.
    > > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > > before there were _any_ computers for sale.
    > > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > > Think how that can be.
    > > > > I did not use a processor :-)
    > > > > LOL
    > > > > Even then some Unix people were big ego boasters.
    > > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > > Or at least basic electronics.

    >
    > > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > > but is constraint by protocol timing. Such as if a protocol says you
    > > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > > failure.

    >
    > > > Many handset has a architecture of two or more cores. Most smartphone
    > > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > > to processing applications, another core to processing wireless
    > > > protocol. They connect with each other with some common interface,
    > > > such as SPI, UART etc.

    >
    > > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > > micro processor now, many PDA has a 600Mhz processor but still is
    > > > regarded as "embedded".

    >
    > > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > > the main application processor, the cost is greatly cut down. And it
    > > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > > operation only sustain for several msec, but a default Linux time
    > > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > > Thanks.
    > > > Bin- Hide quoted text -

    >
    > > > - Show quoted text -

    >
    > > There arereal-timehypervisors from various companies (available or
    > > at least announced):

    >
    > > Depending on processor architecture you have at least the following
    > > companies to choose from:

    >
    > > ARM etc.
    > > Trango www.trango-vp.com
    > > Virtuallogix www.virtuallogix.commentionedin the first
    > > post (also PPC and maybe more)

    >
    > > x86:
    > > Tenasys www.tenasys.com
    > >Real-TimeSystems www.real-time-systems.com

    >
    > > If singlecore processors are used you will always have a performance
    > > hit for context switching between the OSes and you (or the respective
    > > vendors) might run into patent issues with e.g. IBM and the well known
    > > RT-Linux patent (now owned by Wind River).

    >
    > > On multicore architectures you have the advantage of much better
    > > interrupt latencies and you can reboot individual OSes (cores) anytime
    > > during operation - but your hardware gets more expensive.

    >
    > Thank you very much!
    > Do you have more infomation on how they implement this? I want it
    > eagerly. RTOS is simple one but the important part is the software
    > built on RTOS. I want to know how many ways can I try to achieve this,
    > and has any reference already?
    >
    > Software such as XEN is quite large, and at the first glance it can't
    > fulfill my requirement. May be I can start a new open source project
    > if I know exactly how to do it
    >
    > Thanks.
    > Bin- Hide quoted text -
    >
    > - Show quoted text -



    I work for one of the vendors I mentioned therefore I do not want to
    advertise one solution here or discuss competing vendor's products.
    It is best if you approach these companies directly.

    Thanks for your understanding.


    Gerd


  8. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 2, 9:33 pm, gerd.lamm...@real-time-systems.com wrote:
    > On Jul 2, 1:52 pm, Bin Chen wrote:
    >
    >
    >
    > > On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > On Jul 2, 3:41 am, Bin Chen wrote:

    >
    > > > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > > > wrote in
    > > > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > > > >Hi,

    >
    > > > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > > > >field. Many works which need to be done by 2 or more processors can be
    > > > > > >done by only one processor if adopt a technique that allow a
    > > > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > > > >RTOS(or more?).

    >
    > > > > > >I am very interested in this technology and I know there has been a
    > > > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > > > >plus RTOS.

    >
    > > > > > >Because RTOS has some constraints on timing, so I think there should
    > > > > > >be many differences between existing open source general purpose
    > > > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > > > >techniques.

    >
    > > > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > > > >good at this can comment for is it very hard to make this done?

    >
    > > > > > >Thanks for your comments!

    >
    > > > > > >Bin

    >
    > > > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > > > I can see use of it to sell virtual servers for customers...
    > > > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > > > or state machines, and having one or more processors for sequential code,
    > > > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > > > seems a joke to me.
    > > > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > > > Never tried that.

    >
    > > > > > I will try to explain some aspects of this.
    > > > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > > > have as requirement that when input pin of the par port goes high, then
    > > > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > > > then you have a huge problem.
    > > > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > > > or use simple logic.
    > > > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > > > gate array, while the other part of the thousands of gates available can for example
    > > > > > form a sequential system, a processor, running Linux if must be.
    > > > > > Never ever can you beat this with virtualizing.
    > > > > > never.
    > > > > > Those who come from a hardware background will know what I am talking about.
    > > > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > > > So use what is best for the requirement you have.
    > > > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > > > if the same can be done with a few milliwatt logic gates.
    > > > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > > > before there were _any_ computers for sale.
    > > > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > > > Think how that can be.
    > > > > > I did not use a processor :-)
    > > > > > LOL
    > > > > > Even then some Unix people were big ego boasters.
    > > > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > > > Or at least basic electronics.

    >
    > > > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > > > but is constraint by protocol timing. Such as if a protocol says you
    > > > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > > > failure.

    >
    > > > > Many handset has a architecture of two or more cores. Most smartphone
    > > > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > > > to processing applications, another core to processing wireless
    > > > > protocol. They connect with each other with some common interface,
    > > > > such as SPI, UART etc.

    >
    > > > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > > > micro processor now, many PDA has a 600Mhz processor but still is
    > > > > regarded as "embedded".

    >
    > > > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > > > the main application processor, the cost is greatly cut down. And it
    > > > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > > > operation only sustain for several msec, but a default Linux time
    > > > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > > > Thanks.
    > > > > Bin- Hide quoted text -

    >
    > > > > - Show quoted text -

    >
    > > > There arereal-timehypervisors from various companies (available or
    > > > at least announced):

    >
    > > > Depending on processor architecture you have at least the following
    > > > companies to choose from:

    >
    > > > ARM etc.
    > > > Trango www.trango-vp.com
    > > > Virtuallogix www.virtuallogix.commentionedinthe first
    > > > post (also PPC and maybe more)

    >
    > > > x86:
    > > > Tenasys www.tenasys.com
    > > >Real-TimeSystems www.real-time-systems.com

    >
    > > > If singlecore processors are used you will always have a performance
    > > > hit for context switching between the OSes and you (or the respective
    > > > vendors) might run into patent issues with e.g. IBM and the well known
    > > > RT-Linux patent (now owned by Wind River).

    >
    > > > On multicore architectures you have the advantage of much better
    > > > interrupt latencies and you can reboot individual OSes (cores) anytime
    > > > during operation - but your hardware gets more expensive.

    >
    > > Thank you very much!
    > > Do you have more infomation on how they implement this? I want it
    > > eagerly. RTOS is simple one but the important part is the software
    > > built on RTOS. I want to know how many ways can I try to achieve this,
    > > and has any reference already?

    >
    > > Software such as XEN is quite large, and at the first glance it can't
    > > fulfill my requirement. May be I can start a new open source project
    > > if I know exactly how to do it

    >
    > > Thanks.
    > > Bin- Hide quoted text -

    >
    > > - Show quoted text -

    >
    > I work for one of the vendors I mentioned therefore I do not want to
    > advertise one solution here or discuss competing vendor's products.
    > It is best if you approach these companies directly.
    >
    > Thanks for your understanding.
    >

    Thank you. I totally understand.
    Hope others here can give more detailed info.

    Bin


  9. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 2, 10:32 pm, Bin Chen wrote:
    > On Jul 2, 9:33 pm, gerd.lamm...@real-time-systems.com wrote:
    >
    > > On Jul 2, 1:52 pm, Bin Chen wrote:

    >
    > > > On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > > On Jul 2, 3:41 am, Bin Chen wrote:

    >
    > > > > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > > > > wrote in
    > > > > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > > > > >Hi,

    >
    > > > > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > > > > >field. Many works which need to be done by 2 or more processors can be
    > > > > > > >done by only one processor if adopt a technique that allow a
    > > > > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > > > > >RTOS(or more?).

    >
    > > > > > > >I am very interested in this technology and I know there has been a
    > > > > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > > > > >plus RTOS.

    >
    > > > > > > >Because RTOS has some constraints on timing, so I think there should
    > > > > > > >be many differences between existing open source general purpose
    > > > > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > > > > >techniques.

    >
    > > > > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > > > > >good at this can comment for is it very hard to make this done?

    >
    > > > > > > >Thanks for your comments!

    >
    > > > > > > >Bin

    >
    > > > > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > > > > I can see use of it to sell virtual servers for customers...
    > > > > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > > > > or state machines, and having one or more processors for sequential code,
    > > > > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > > > > seems a joke to me.
    > > > > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > > > > Never tried that.

    >
    > > > > > > I will try to explain some aspects of this.
    > > > > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > > > > have as requirement that when input pin of the par port goes high, then
    > > > > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > > > > then you have a huge problem.
    > > > > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > > > > or use simple logic.
    > > > > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > > > > gate array, while the other part of the thousands of gates available can for example
    > > > > > > form a sequential system, a processor, running Linux if must be.
    > > > > > > Never ever can you beat this with virtualizing.
    > > > > > > never.
    > > > > > > Those who come from a hardware background will know what I am talking about.
    > > > > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > > > > So use what is best for the requirement you have.
    > > > > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > > > > if the same can be done with a few milliwatt logic gates.
    > > > > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > > > > before there were _any_ computers for sale.
    > > > > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > > > > Think how that can be.
    > > > > > > I did not use a processor :-)
    > > > > > > LOL
    > > > > > > Even then some Unix people were big ego boasters.
    > > > > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > > > > Or at least basic electronics.

    >
    > > > > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > > > > but is constraint by protocol timing. Such as if a protocol says you
    > > > > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > > > > failure.

    >
    > > > > > Many handset has a architecture of two or more cores. Most smartphone
    > > > > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > > > > to processing applications, another core to processing wireless
    > > > > > protocol. They connect with each other with some common interface,
    > > > > > such as SPI, UART etc.

    >
    > > > > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > > > > micro processor now, many PDA has a 600Mhz processor but still is
    > > > > > regarded as "embedded".

    >
    > > > > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > > > > the main application processor, the cost is greatly cut down. And it
    > > > > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > > > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > > > > operation only sustain for several msec, but a default Linux time
    > > > > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > > > > Thanks.
    > > > > > Bin- Hide quoted text -

    >
    > > > > > - Show quoted text -

    >
    > > > > There arereal-timehypervisors from various companies (available or
    > > > > at least announced):

    >
    > > > > Depending on processor architecture you have at least the following
    > > > > companies to choose from:

    >
    > > > > ARM etc.
    > > > > Trango www.trango-vp.com
    > > > > Virtuallogix www.virtuallogix.commentionedinthefirst
    > > > > post (also PPC and maybe more)

    >
    > > > > x86:
    > > > > Tenasys www.tenasys.com
    > > > >Real-TimeSystems www.real-time-systems.com

    >
    > > > > If singlecore processors are used you will always have a performance
    > > > > hit for context switching between the OSes and you (or the respective
    > > > > vendors) might run into patent issues with e.g. IBM and the well known
    > > > > RT-Linux patent (now owned by Wind River).

    >
    > > > > On multicore architectures you have the advantage of much better
    > > > > interrupt latencies and you can reboot individual OSes (cores) anytime
    > > > > during operation - but your hardware gets more expensive.

    >
    > > > Thank you very much!
    > > > Do you have more infomation on how they implement this? I want it
    > > > eagerly. RTOS is simple one but the important part is the software
    > > > built on RTOS. I want to know how many ways can I try to achieve this,
    > > > and has any reference already?

    >
    > > > Software such as XEN is quite large, and at the first glance it can't
    > > > fulfill my requirement. May be I can start a new open source project
    > > > if I know exactly how to do it

    >
    > > > Thanks.
    > > > Bin- Hide quoted text -

    >
    > > > - Show quoted text -

    >
    > > I work for one of the vendors I mentioned therefore I do not want to
    > > advertise one solution here or discuss competing vendor's products.
    > > It is best if you approach these companies directly.

    >
    > > Thanks for your understanding.

    >
    > Thank you. I totally understand.
    > Hope others here can give more detailed info.


    A friend recommend a project, seems like what i want:

    http://home.gna.org/adeos/



  10. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 3, 4:37 pm, Bin Chen wrote:
    > On Jul 2, 10:32 pm, Bin Chen wrote:
    >
    >
    >
    >
    >
    > > On Jul 2, 9:33 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > On Jul 2, 1:52 pm, Bin Chen wrote:

    >
    > > > > On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > > > On Jul 2, 3:41 am, Bin Chen wrote:

    >
    > > > > > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > > > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > > > > > wrote in
    > > > > > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > > > > > >Hi,

    >
    > > > > > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > > > > > >field. Many works which need to be done by 2 or more processors can be
    > > > > > > > >done by only one processor if adopt a technique that allow a
    > > > > > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > > > > > >RTOS(or more?).

    >
    > > > > > > > >I am very interested in this technology and I know there has been a
    > > > > > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > > > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > > > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > > > > > >plus RTOS.

    >
    > > > > > > > >Because RTOS has some constraints on timing, so I think there should
    > > > > > > > >be many differences between existing open source general purpose
    > > > > > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > > > > > >techniques.

    >
    > > > > > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > > > > > >good at this can comment for is it very hard to make this done?

    >
    > > > > > > > >Thanks for your comments!

    >
    > > > > > > > >Bin

    >
    > > > > > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > > > > > I can see use of it to sell virtual servers for customers...
    > > > > > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > > > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > > > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > > > > > or state machines, and having one or more processors for sequential code,
    > > > > > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > > > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > > > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > > > > > seems a joke to me.
    > > > > > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > > > > > Never tried that.

    >
    > > > > > > > I will try to explain some aspects of this.
    > > > > > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > > > > > have as requirement that when input pin of the par port goes high, then
    > > > > > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > > > > > then you have a huge problem.
    > > > > > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > > > > > or use simple logic.
    > > > > > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > > > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > > > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > > > > > gate array, while the other part of the thousands of gates available can for example
    > > > > > > > form a sequential system, a processor, running Linux if must be.
    > > > > > > > Never ever can you beat this with virtualizing.
    > > > > > > > never.
    > > > > > > > Those who come from a hardware background will know what I am talking about.
    > > > > > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > > > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > > > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > > > > > So use what is best for the requirement you have.
    > > > > > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > > > > > if the same can be done with a few milliwatt logic gates.
    > > > > > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > > > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > > > > > before there were _any_ computers for sale.
    > > > > > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > > > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > > > > > Think how that can be.
    > > > > > > > I did not use a processor :-)
    > > > > > > > LOL
    > > > > > > > Even then some Unix people were big ego boasters.
    > > > > > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > > > > > Or at least basic electronics.

    >
    > > > > > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > > > > > but is constraint by protocol timing. Such as if a protocol says you
    > > > > > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > > > > > failure.

    >
    > > > > > > Many handset has a architecture of two or more cores. Most smartphone
    > > > > > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > > > > > to processing applications, another core to processing wireless
    > > > > > > protocol. They connect with each other with some common interface,
    > > > > > > such as SPI, UART etc.

    >
    > > > > > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > > > > > micro processor now, many PDA has a 600Mhz processor but still is
    > > > > > > regarded as "embedded".

    >
    > > > > > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > > > > > the main application processor, the cost is greatly cut down. And it
    > > > > > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > > > > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > > > > > operation only sustain for several msec, but a default Linux time
    > > > > > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > > > > > Thanks.
    > > > > > > Bin- Hide quoted text -

    >
    > > > > > > - Show quoted text -

    >
    > > > > > There arereal-timehypervisors from various companies (available or
    > > > > > at least announced):

    >
    > > > > > Depending on processor architecture you have at least the following
    > > > > > companies to choose from:

    >
    > > > > > ARM etc.
    > > > > > Trango www.trango-vp.com
    > > > > > Virtuallogix www.virtuallogix.commentionedinthefirst
    > > > > > post (also PPC and maybe more)

    >
    > > > > > x86:
    > > > > > Tenasys www.tenasys.com
    > > > > >Real-TimeSystems www.real-time-systems.com

    >
    > > > > > If singlecore processors are used you will always have a performance
    > > > > > hit for context switching between the OSes and you (or the respective
    > > > > > vendors) might run into patent issues with e.g. IBM and the well known
    > > > > > RT-Linux patent (now owned by Wind River).

    >
    > > > > > On multicore architectures you have the advantage of much better
    > > > > > interrupt latencies and you can reboot individual OSes (cores) anytime
    > > > > > during operation - but your hardware gets more expensive.

    >
    > > > > Thank you very much!
    > > > > Do you have more infomation on how they implement this? I want it
    > > > > eagerly. RTOS is simple one but the important part is the software
    > > > > built on RTOS. I want to know how many ways can I try to achieve this,
    > > > > and has any reference already?

    >
    > > > > Software such as XEN is quite large, and at the first glance it can't
    > > > > fulfill my requirement. May be I can start a new open source project
    > > > > if I know exactly how to do it

    >
    > > > > Thanks.
    > > > > Bin- Hide quoted text -

    >
    > > > > - Show quoted text -

    >
    > > > I work for one of the vendors I mentioned therefore I do not want to
    > > > advertise one solution here or discuss competing vendor's products.
    > > > It is best if you approach these companies directly.

    >
    > > > Thanks for your understanding.

    >
    > > Thank you. I totally understand.
    > > Hope others here can give more detailed info.

    >
    > A friend recommend a project, seems like what i want:
    >
    > http://home.gna.org/adeos/- Hide quoted text -
    >
    > - Show quoted text -


    There is also www.xenomai.org, www.rtai.org and www.osadl.org
    Just like ADEOS these are all Real-Time Extensions for Linux or in the
    case of OSADL a "real-time linux" itself.
    In the same world plays a commercial version "rt-linux" from Wind
    River (formerly FSM-Labs) www.windriver.com

    Hope this helps.

    Gerd


  11. Re: Virtualization - Running RTOS and UNIX on one machine

    On Jul 4, 12:01 am, gerd.lamm...@real-time-systems.com wrote:
    > On Jul 3, 4:37 pm, Bin Chen wrote:
    >
    >
    >
    > > On Jul 2, 10:32 pm, Bin Chen wrote:

    >
    > > > On Jul 2, 9:33 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > > On Jul 2, 1:52 pm, Bin Chen wrote:

    >
    > > > > > On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > > > > On Jul 2, 3:41 am, Bin Chen wrote:

    >
    > > > > > > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > > > > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > > > > > > wrote in
    > > > > > > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > > > > > > >Hi,

    >
    > > > > > > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > > > > > > >field. Many works which need to be done by 2 or more processors can be
    > > > > > > > > >done by only one processor if adopt a technique that allow a
    > > > > > > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > > > > > > >RTOS(or more?).

    >
    > > > > > > > > >I am very interested in this technology and I know there has been a
    > > > > > > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > > > > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > > > > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > > > > > > >plus RTOS.

    >
    > > > > > > > > >Because RTOS has some constraints on timing, so I think there should
    > > > > > > > > >be many differences between existing open source general purpose
    > > > > > > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > > > > > > >techniques.

    >
    > > > > > > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > > > > > > >good at this can comment for is it very hard to make this done?

    >
    > > > > > > > > >Thanks for your comments!

    >
    > > > > > > > > >Bin

    >
    > > > > > > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > > > > > > I can see use of it to sell virtual servers for customers...
    > > > > > > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > > > > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > > > > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > > > > > > or state machines, and having one or more processors for sequential code,
    > > > > > > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > > > > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > > > > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > > > > > > seems a joke to me.
    > > > > > > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > > > > > > Never tried that.

    >
    > > > > > > > > I will try to explain some aspects of this.
    > > > > > > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > > > > > > have as requirement that when input pin of the par port goes high, then
    > > > > > > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > > > > > > then you have a huge problem.
    > > > > > > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > > > > > > or use simple logic.
    > > > > > > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > > > > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > > > > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > > > > > > gate array, while the other part of the thousands of gates available can for example
    > > > > > > > > form a sequential system, a processor, running Linux if must be.
    > > > > > > > > Never ever can you beat this with virtualizing.
    > > > > > > > > never.
    > > > > > > > > Those who come from a hardware background will know what I am talking about.
    > > > > > > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > > > > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > > > > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > > > > > > So use what is best for the requirement you have.
    > > > > > > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > > > > > > if the same can be done with a few milliwatt logic gates.
    > > > > > > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > > > > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > > > > > > before there were _any_ computers for sale.
    > > > > > > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > > > > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > > > > > > Think how that can be.
    > > > > > > > > I did not use a processor :-)
    > > > > > > > > LOL
    > > > > > > > > Even then some Unix people were big ego boasters.
    > > > > > > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > > > > > > Or at least basic electronics.

    >
    > > > > > > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > > > > > > but is constraint by protocol timing. Such as if a protocol says you
    > > > > > > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > > > > > > failure.

    >
    > > > > > > > Many handset has a architecture of two or more cores. Most smartphone
    > > > > > > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > > > > > > to processing applications, another core to processing wireless
    > > > > > > > protocol. They connect with each other with some common interface,
    > > > > > > > such as SPI, UART etc.

    >
    > > > > > > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > > > > > > micro processor now, many PDA has a 600Mhz processor but still is
    > > > > > > > regarded as "embedded".

    >
    > > > > > > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > > > > > > the main application processor, the cost is greatly cut down. And it
    > > > > > > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > > > > > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > > > > > > operation only sustain for several msec, but a default Linux time
    > > > > > > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > > > > > > Thanks.
    > > > > > > > Bin- Hide quoted text -

    >
    > > > > > > > - Show quoted text -

    >
    > > > > > > There arereal-timehypervisors from various companies (available or
    > > > > > > at least announced):

    >
    > > > > > > Depending on processor architecture you have at least the following
    > > > > > > companies to choose from:

    >
    > > > > > > ARM etc.
    > > > > > > Trango www.trango-vp.com
    > > > > > > Virtuallogix www.virtuallogix.commentionedinthefirst
    > > > > > > post (also PPC and maybe more)

    >
    > > > > > > x86:
    > > > > > > Tenasys www.tenasys.com
    > > > > > >Real-TimeSystems www.real-time-systems.com

    >
    > > > > > > If singlecore processors are used you will always have a performance
    > > > > > > hit for context switching between the OSes and you (or the respective
    > > > > > > vendors) might run into patent issues with e.g. IBM and the well known
    > > > > > > RT-Linux patent (now owned by Wind River).

    >
    > > > > > > On multicore architectures you have the advantage of much better
    > > > > > > interrupt latencies and you can reboot individual OSes (cores) anytime
    > > > > > > during operation - but your hardware gets more expensive.

    >
    > > > > > Thank you very much!
    > > > > > Do you have more infomation on how they implement this? I want it
    > > > > > eagerly. RTOS is simple one but the important part is the software
    > > > > > built on RTOS. I want to know how many ways can I try to achieve this,
    > > > > > and has any reference already?

    >
    > > > > > Software such as XEN is quite large, and at the first glance it can't
    > > > > > fulfill my requirement. May be I can start a new open source project
    > > > > > if I know exactly how to do it

    >
    > > > > > Thanks.
    > > > > > Bin- Hide quoted text -

    >
    > > > > > - Show quoted text -

    >
    > > > > I work for one of the vendors I mentioned therefore I do not want to
    > > > > advertise one solution here or discuss competing vendor's products.
    > > > > It is best if you approach these companies directly.

    >
    > > > > Thanks for your understanding.

    >
    > > > Thank you. I totally understand.
    > > > Hope others here can give more detailed info.

    >
    > > A friend recommend a project, seems like what i want:

    >
    > >http://home.gna.org/adeos/-Hide quoted text -

    >
    > > - Show quoted text -

    >
    > There is alsowww.xenomai.org,www.rtai.org andwww.osadl.org
    > Just like ADEOS these are all Real-Time Extensions for Linux or in the
    > case of OSADL a "real-time linux" itself.
    > In the same world plays a commercial version "rt-linux" from Wind
    > River (formerly FSM-Labs)www.windriver.com


    Thanks. What I want to achieve is make the RTOS running without
    modification (or only a small modification), so I think RT-Linux can't
    do this

    Bin



  12. Re: Virtualization - Running RTOS and UNIX on one machine

    On 4 Jul, 02:51, Bin Chen wrote:
    > On Jul 4, 12:01 am, gerd.lamm...@real-time-systems.com wrote:
    >
    >
    >
    > > On Jul 3, 4:37 pm, Bin Chen wrote:

    >
    > > > On Jul 2, 10:32 pm, Bin Chen wrote:

    >
    > > > > On Jul 2, 9:33 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > > > On Jul 2, 1:52 pm, Bin Chen wrote:

    >
    > > > > > > On Jul 2, 6:01 pm, gerd.lamm...@real-time-systems.com wrote:

    >
    > > > > > > > On Jul 2, 3:41 am, Bin Chen wrote:

    >
    > > > > > > > > On Jul 2, 12:37 am, Jan Panteltje wrote:

    >
    > > > > > > > > > On a sunny day (Sun, 01 Jul 2007 15:12:49 -0000) it happened Bin Chen
    > > > > > > > > > wrote in
    > > > > > > > > > <1183302769.974254.151...@i38g2000prf.googlegroups. com>:

    >
    > > > > > > > > > >Hi,

    >
    > > > > > > > > > >It seems a very large marketplace for thevirtualizationin embedded
    > > > > > > > > > >field. Many works which need to be done by 2 or more processors can be
    > > > > > > > > > >done by only one processor if adopt a technique that allow a
    > > > > > > > > > >supervisor software to control both a UNIX OS(typically Linux) and a
    > > > > > > > > > >RTOS(or more?).

    >
    > > > > > > > > > >I am very interested in this technology and I know there has been a
    > > > > > > > > > >lot of startup in this fields, such ashttp://www.virtuallogix.com/.
    > > > > > > > > > >And there are many open sourcevirtualizationproject, such as XEN and
    > > > > > > > > > >OpenVZ, but I can't find a project for running a general purpose OS
    > > > > > > > > > >plus RTOS.

    >
    > > > > > > > > > >Because RTOS has some constraints on timing, so I think there should
    > > > > > > > > > >be many differences between existing open source general purpose
    > > > > > > > > > >virtualizationtechniques and the commercial RTOS vitualization
    > > > > > > > > > >techniques.

    >
    > > > > > > > > > >I really don't have enough experience on RTOS, so can anybody who are
    > > > > > > > > > >good at this can comment for is it very hard to make this done?

    >
    > > > > > > > > > >Thanks for your comments!

    >
    > > > > > > > > > >Bin

    >
    > > > > > > > > > I dived into thevirtualizationsome year ago, and was disappointed by the performance.
    > > > > > > > > > I can see use of it to sell virtual servers for customers...
    > > > > > > > > > But for embedded wherereal timeis important, and resources are very limited,
    > > > > > > > > > I see no advantages in this over say an FPGA with 2 or more processors.
    > > > > > > > > > Doing the parts that need to be fast in FPGA as gate operations
    > > > > > > > > > or state machines, and having one or more processors for sequential code,
    > > > > > > > > > seems to me to satisfy bothreal timeand complex code requirements.
    > > > > > > > > > So I am a non-believer invirtualizationfor small -possibly embedded- systems.
    > > > > > > > > > If you go your way, there isreal timeLinux, butreal timeLinux under a virtualizer
    > > > > > > > > > seems a joke to me.
    > > > > > > > > > If the virtualizer itself is a RTOS and gives spare time to a Linux running perhaps....
    > > > > > > > > > Never tried that.

    >
    > > > > > > > > > I will try to explain some aspects of this.
    > > > > > > > > > If you run a virtualizer, with Linux, and say MSDOS, say on a very fast PC, and you
    > > > > > > > > > have as requirement that when input pin of the par port goes high, then
    > > > > > > > > > within 1 us some output on some PCI card must go high, either using the DOS or Linux,
    > > > > > > > > > then you have a huge problem.
    > > > > > > > > > We must then go back to basic understanding of the hardware, sequential processing
    > > > > > > > > > or use simple logic.
    > > > > > > > > > Programmable gate logic (FPGA) allows you to just use simple logic nand / nor /xor etc)
    > > > > > > > > > to carry a specific signal from one pin to an other in usually about 1 clock tick.
    > > > > > > > > > This happens as a separate circuit, an completely different 'machine' withing the same
    > > > > > > > > > gate array, while the other part of the thousands of gates available can for example
    > > > > > > > > > form a sequential system, a processor, running Linux if must be.
    > > > > > > > > > Never ever can you beat this with virtualizing.
    > > > > > > > > > never.
    > > > > > > > > > Those who come from a hardware background will know what I am talking about.
    > > > > > > > > > Those who believe in the Holy Penguin Sequential Solution To Everything Under Heaven are
    > > > > > > > > > doomed to be the late arrivals, using huge amounts of power, producing huge amounts
    > > > > > > > > > of heat, and those solutions will be slower and less reliable.

    >
    > > > > > > > > > So use what is best for the requirement you have.
    > > > > > > > > > It makes no sense to use a 3.5GHz dual Pentium 10 or whatever to make thereal timerequirement
    > > > > > > > > > if the same can be done with a few milliwatt logic gates.
    > > > > > > > > > Some programmers, if not most programmers, have a very hard time learning HDL (hardware
    > > > > > > > > > description languages), I most fortunately have a hardware background, played with gates long
    > > > > > > > > > before there were _any_ computers for sale.
    > > > > > > > > > In 1979 I digitised my first video signal ... designed the digitiser.
    > > > > > > > > > The max processor speed available at that time was just a few MHz (less then 10).
    > > > > > > > > > Think how that can be.
    > > > > > > > > > I did not use a processor :-)
    > > > > > > > > > LOL
    > > > > > > > > > Even then some Unix people were big ego boasters.
    > > > > > > > > > You will all have to learn at least _one_ HDL if you ever want to design anything "real time".
    > > > > > > > > > Or at least basic electronics.

    >
    > > > > > > > > Thanks for your comment, but AFAIK, your case is only a type ofrealtime, many otherreal timesystem doesn't rely on hardware like this,
    > > > > > > > > but is constraint by protocol timing. Such as if a protocol says you
    > > > > > > > > must respond a ACK within 5 ms, otherwise the other end treat this
    > > > > > > > > failure.

    >
    > > > > > > > > Many handset has a architecture of two or more cores. Most smartphone
    > > > > > > > > or PDA which integrates a Bluetooth, GSM or WiFi module has one core
    > > > > > > > > to processing applications, another core to processing wireless
    > > > > > > > > protocol. They connect with each other with some common interface,
    > > > > > > > > such as SPI, UART etc.

    >
    > > > > > > > > Man, embedded is a broader concept now, not only a device with 8-bit
    > > > > > > > > micro processor now, many PDA has a 600Mhz processor but still is
    > > > > > > > > regarded as "embedded".

    >
    > > > > > > > > Accordingly, if we can merge the core for GSM, Bluetooth, or WiFi with
    > > > > > > > > the main application processor, the cost is greatly cut down. And it
    > > > > > > > > is viable. Take GSM as an example, most time the GSM is idle and when
    > > > > > > > > GSM is working, we can yield most the CPU power to it. Most GSM
    > > > > > > > > operation only sustain for several msec, but a default Linux time
    > > > > > > > > slice is 100ms!! So I think this is absolutely doable.

    >
    > > > > > > > > Thanks.
    > > > > > > > > Bin- Hide quoted text -

    >
    > > > > > > > > - Show quoted text -

    >
    > > > > > > > There arereal-timehypervisors from various companies (available or
    > > > > > > > at least announced):

    >
    > > > > > > > Depending on processor architecture you have at least the following
    > > > > > > > companies to choose from:

    >
    > > > > > > > ARM etc.
    > > > > > > > Trango www.trango-vp.com
    > > > > > > > Virtuallogix www.virtuallogix.commentionedinthefirst
    > > > > > > > post (also PPC and maybe more)

    >
    > > > > > > > x86:
    > > > > > > > Tenasys www.tenasys.com
    > > > > > > >Real-TimeSystems www.real-time-systems.com

    >
    > > > > > > > If singlecore processors are used you will always have a performance
    > > > > > > > hit for context switching between the OSes and you (or the respective
    > > > > > > > vendors) might run into patent issues with e.g. IBM and the well known
    > > > > > > > RT-Linux patent (now owned by Wind River).

    >
    > > > > > > > On multicore architectures you have the advantage of much better
    > > > > > > > interrupt latencies and you can reboot individual OSes (cores) anytime
    > > > > > > > during operation - but your hardware gets more expensive.

    >
    > > > > > > Thank you very much!
    > > > > > > Do you have more infomation on how they implement this? I want it
    > > > > > > eagerly. RTOS is simple one but the important part is the software
    > > > > > > built on RTOS. I want to know how many ways can I try to achieve this,
    > > > > > > and has any reference already?

    >
    > > > > > > Software such as XEN is quite large, and at the first glance it can't
    > > > > > > fulfill my requirement. May be I can start a new open source project
    > > > > > > if I know exactly how to do it

    >
    > > > > > > Thanks.
    > > > > > > Bin- Hide quoted text -

    >
    > > > > > > - Show quoted text -

    >
    > > > > > I work for one of the vendors I mentioned therefore I do not want to
    > > > > > advertise one solution here or discuss competing vendor's products.
    > > > > > It is best if you approach these companies directly.

    >
    > > > > > Thanks for your understanding.

    >
    > > > > Thank you. I totally understand.
    > > > > Hope others here can give more detailed info.

    >
    > > > A friend recommend a project, seems like what i want:

    >
    > > >http://home.gna.org/adeos/-Hidequoted text -

    >
    > > > - Show quoted text -

    >
    > > There is alsowww.xenomai.org,www.rtai.organdwww.osadl.org
    > > Just like ADEOS these are all Real-Time Extensions for Linux or in the
    > > case of OSADL a "real-time linux" itself.
    > > In the same world plays a commercial version "rt-linux" from Wind
    > > River (formerly FSM-Labs)www.windriver.com

    >
    > Thanks. What I want to achieve is make the RTOS running without
    > modification (or only a small modification), so I think RT-Linux can't
    > do this
    >
    > Bin


    You could possibly try www.ok-labs.com. They can virtualise eCos and
    Linux but I am not sure if simultaneously. While I personally think
    the L4 microkernel is quite promising for virtualisation, I am
    sceptical about the virtualisation concept in general on embedded
    systems. Companies are pushing time-to-market and cost reduction too
    hard, which causes virtualisation to become desirable. They want real-
    time, they want all linux applications, they want everything, and in a
    very quick and low-cost way. But in the end time will tell that this
    is only a temporary solution that will collapse.

    I understand its use in the enterprise, but on embedded systems I
    believe it is no more than a quick dirty hack. This even applies to
    porting applications from desktop to embedded linux. It's just not a
    very clever idea. The kernel I believe runs reasonably well for most
    embedded uses (e.g, a settop box) but I would replace the userspace
    starting from INIT and the shell.


    Bahadir


+ Reply to Thread