to increase the efficiency of SAN . I am thinking of a design which
provides an in band virtualization which recieves data requst n form
of ISCSI packets. splits th requst into two . and simultanously maks
connection with two physical disk remotely placed in different
locations in sync with each other. is this thing possible !! if not
why not ?

varun wrote:
> to increase the efficiency of SAN . I am thinking of a design which
> provides an in band virtualization which recieves data requst n form
> of ISCSI packets. splits th requst into two . and simultanously maks
> connection with two physical disk remotely placed in different
> locations in sync with each other. is this thing possible !! if not
> why not ?


Not only possible, but quite widely implemented in a variety of storage
virtualization implementations. The only real gotcha with doing this
synchronously is distance of the two locations which will impact
latency, potentially up to the point that the host will think the I/O
timed out and force a retry, or worse.

--
Nik Simpson

On Aug 25, 7:19*am, varun wrote:
> to increase the efficiency of *SAN . I am thinking of a design which
> provides an in band virtualization which recieves data requst n form
> of ISCSI packets. splits th requst into two . and simultanously maks
> connection with two physical disk remotely placed in different
> locations in sync with each other. is this thing possible !! if not
> why not ?

This only makes the SAN more efficient if you're only doing reads. If
you're doing a lot of writes, this makes the SAN very inefficient
since you're doing synchronous replication between 2 locations. If
they're under 100km apart and you have really low latency, you can
certainly do this. If you're trying to do this between NYC and Kansas
City, your performance is going to suck. In any case, you will
*always* slow down writes in a synchronous replication solution. You
can not write any faster than your network.

You need to determine if you're trying to build a DR solution or a
high-performance solution - they have different trade-offs.

.../Ed

Ed Wilts, RHCE, BCFP, BCSD, SCSP, SCSE
mailto:ewilts@ewilts.org

On Aug 25, 8:13*pm, Ed Wilts wrote:
> On Aug 25, 7:19*am, varun wrote:
>
> > to increase the efficiency of *SAN . I am thinking of a design which
> > provides an in band virtualization which recieves data requst n form
> > of ISCSI packets. splits th requst into two . and simultanously maks
> > connection with two physical disk remotely placed in different
> > locations in sync with each other. is this thing possible !! if not
> > why not ?
>
> This only makes the SAN more efficient if you're only doing reads. *If
> you're doing a lot of writes, this makes the SAN very inefficient
> since you're doing synchronous replication between 2 locations. *If
> they're under 100km apart and you have really low latency, you can
> certainly do this. *If you're trying to do this between NYC and Kansas
> City, your performance is going to suck. *In any case, you will
> *always* slow down writes in a synchronous replication solution. *You
> can not write any faster than your network.
>
> You need to determine if you're trying to build a DR solution or a
> high-performance solution - they have different trade-offs.
>
> * *.../Ed
>
> Ed Wilts, RHCE, BCFP, BCSD, SCSP, SCSE
> mailto:ewi...@ewilts.org

ok , beside keeping two remote location in sync ! what are the
advantages of striping of data blocks as implemented in RAID
controllers in remote locations ? what are the pros and cons of this
solution??