Need Advice on Windows File Cluster - Storage

This is a discussion on Need Advice on Windows File Cluster - Storage ; I have a single active \ passive cluster running 2003 Storage Server as a File Server for 2500 - 3000 people. It is currently at about 75% capacity with ~15TB of data on it. We were looking to implement a ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Need Advice on Windows File Cluster

  1. Need Advice on Windows File Cluster

    I have a single active \ passive cluster running 2003 Storage Server as a
    File Server for 2500 - 3000 people. It is currently at about 75% capacity
    with ~15TB of data on it. We were looking to implement a data archival
    device using CAS technology. The vendor is telling us that the current
    amount of data is too much for the archival policies to process every night.
    We are discussing breaking it out into two seperate file clusters and using
    DFS to keep the existing name space. Does this line of thought make sense or
    are we overly complicating this? Also if I use DFS will the end users
    connection be with the DFS server or with the actual cluster on the backend.
    I need to size the DFS server properly based on this info. Thanks.

  2. Re: Need Advice on Windows File Cluster

    Normally the issue is the amount of data churn, not the size of the data;
    though this would depend on the specific implementation of the CAS solution.
    Splitting the data across multiple nodes may or may not help - if you still
    have a single scanner & it has to scan the entire corpus every night then
    you haven't really changed anything.

    Out of curiosity - have you enabled Single Instance Storage (SIS) on the
    Storage Server? We use it here & it cuts our storage by ~40% (a lot of
    duplicates in source files). Most companies see a cut closer to 20% or 30%.


    Pat



    "Tim" wrote in message
    news:BC8265A6-E7D7-4DA6-992C-C6AA15D2E6A9@microsoft.com...
    >I have a single active \ passive cluster running 2003 Storage Server as a
    > File Server for 2500 - 3000 people. It is currently at about 75% capacity
    > with ~15TB of data on it. We were looking to implement a data archival
    > device using CAS technology. The vendor is telling us that the current
    > amount of data is too much for the archival policies to process every
    > night.
    > We are discussing breaking it out into two seperate file clusters and
    > using
    > DFS to keep the existing name space. Does this line of thought make sense
    > or
    > are we overly complicating this? Also if I use DFS will the end users
    > connection be with the DFS server or with the actual cluster on the
    > backend.
    > I need to size the DFS server properly based on this info. Thanks.



  3. Re: Need Advice on Windows File Cluster

    Pat,
    Thanks for your response. We had not enabled SIS. We are running a pre-R2
    version of Windows 2003 Storage server, is SIS still available to us?

    The CAS solution we are looking to implement is an EMC Centerra using
    DiskXtender which does utilize SIS. For the most part we have 2TB LUNS
    (constrained by the Win2003 clustering limit). They have told us that
    anything more than 7 physical drives will have difficulty completing in the
    overnight window that we have allocated for it to run.

    I don't think we have a massive amount of data churn, but I don't have any
    hard and fast numbers to provide proof either way. Any other suggestions you
    have are definitly appreciated. -Tim


    "Pat [MSFT]" wrote:

    > Normally the issue is the amount of data churn, not the size of the data;
    > though this would depend on the specific implementation of the CAS solution.
    > Splitting the data across multiple nodes may or may not help - if you still
    > have a single scanner & it has to scan the entire corpus every night then
    > you haven't really changed anything.
    >
    > Out of curiosity - have you enabled Single Instance Storage (SIS) on the
    > Storage Server? We use it here & it cuts our storage by ~40% (a lot of
    > duplicates in source files). Most companies see a cut closer to 20% or 30%.
    >
    >
    > Pat
    >
    >
    >
    > "Tim" wrote in message
    > news:BC8265A6-E7D7-4DA6-992C-C6AA15D2E6A9@microsoft.com...
    > >I have a single active \ passive cluster running 2003 Storage Server as a
    > > File Server for 2500 - 3000 people. It is currently at about 75% capacity
    > > with ~15TB of data on it. We were looking to implement a data archival
    > > device using CAS technology. The vendor is telling us that the current
    > > amount of data is too much for the archival policies to process every
    > > night.
    > > We are discussing breaking it out into two seperate file clusters and
    > > using
    > > DFS to keep the existing name space. Does this line of thought make sense
    > > or
    > > are we overly complicating this? Also if I use DFS will the end users
    > > connection be with the DFS server or with the actual cluster on the
    > > backend.
    > > I need to size the DFS server properly based on this info. Thanks.

    >


  4. Re: Need Advice on Windows File Cluster

    SIS is only on R2. You can contact your OEM and see if they have an upgrade
    program for your devices.

    The 2TB limit on clusters was raised with the release of Service Pack 2 when
    GPT disks are used in lieu of MBRs, with some caveats - namely that the rest
    of the storage stack (drivers, HBAs, etc.) has to support >2TB LUNs as well.
    A surprising number do not.

    EMC would know their solution better than I do. It sounds like they are
    doing a full scan every night vs tracking a change log. In which case it is
    the corpus size & not the change rate that is the controlling factor.


    Pat


    "Tim" wrote in message
    news:1C6F7324-5715-4677-BC1A-091D5BCF3AD7@microsoft.com...
    > Pat,
    > Thanks for your response. We had not enabled SIS. We are running a
    > pre-R2
    > version of Windows 2003 Storage server, is SIS still available to us?
    >
    > The CAS solution we are looking to implement is an EMC Centerra using
    > DiskXtender which does utilize SIS. For the most part we have 2TB LUNS
    > (constrained by the Win2003 clustering limit). They have told us that
    > anything more than 7 physical drives will have difficulty completing in
    > the
    > overnight window that we have allocated for it to run.
    >
    > I don't think we have a massive amount of data churn, but I don't have any
    > hard and fast numbers to provide proof either way. Any other suggestions
    > you
    > have are definitly appreciated. -Tim
    >
    >
    > "Pat [MSFT]" wrote:
    >
    >> Normally the issue is the amount of data churn, not the size of the data;
    >> though this would depend on the specific implementation of the CAS
    >> solution.
    >> Splitting the data across multiple nodes may or may not help - if you
    >> still
    >> have a single scanner & it has to scan the entire corpus every night then
    >> you haven't really changed anything.
    >>
    >> Out of curiosity - have you enabled Single Instance Storage (SIS) on the
    >> Storage Server? We use it here & it cuts our storage by ~40% (a lot of
    >> duplicates in source files). Most companies see a cut closer to 20% or
    >> 30%.
    >>
    >>
    >> Pat
    >>
    >>
    >>
    >> "Tim" wrote in message
    >> news:BC8265A6-E7D7-4DA6-992C-C6AA15D2E6A9@microsoft.com...
    >> >I have a single active \ passive cluster running 2003 Storage Server as
    >> >a
    >> > File Server for 2500 - 3000 people. It is currently at about 75%
    >> > capacity
    >> > with ~15TB of data on it. We were looking to implement a data archival
    >> > device using CAS technology. The vendor is telling us that the current
    >> > amount of data is too much for the archival policies to process every
    >> > night.
    >> > We are discussing breaking it out into two seperate file clusters and
    >> > using
    >> > DFS to keep the existing name space. Does this line of thought make
    >> > sense
    >> > or
    >> > are we overly complicating this? Also if I use DFS will the end users
    >> > connection be with the DFS server or with the actual cluster on the
    >> > backend.
    >> > I need to size the DFS server properly based on this info. Thanks.

    >>



+ Reply to Thread