|
Home > Archive > Data Storage > July 2005 > SANergy and Polyserve
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
SANergy and Polyserve
|
|
| Faeandar 2005-06-28, 5:48 pm |
| Anyone have comments regarding a comparison of the two? I'm fairly
well versed in Polyserve so anyone with background in SANergy would be
a welcome commentator.
I find little information, even from IBM, on this product; it's
requirements, performance, and architecture. The basics are there but
I need to understand it much better than the marketing slides allow.
Thanks.
~F
| |
| kirmse@netaxs.com 2005-06-29, 5:47 pm |
| Faeandar wrote:
> Anyone have comments regarding a comparison of the two? I'm fairly
> well versed in Polyserve so anyone with background in SANergy would be
> a welcome commentator.
>
> I find little information, even from IBM, on this product; it's
> requirements, performance, and architecture. The basics are there but
> I need to understand it much better than the marketing slides allow.
>
> Thanks.
>
> ~F
I have a fair amount of experience with SANergy. The product is ok,
but IBM's support of the product has seemed lackluster the last year
or so. Whether or not SANergy would work will depend on the particular
application in mind.
In general, you have several machines attached to the storage you are
sharing. One machine acts as a Meta Data Controller (MDC) which manages
the file system locks. All machines but the MDC view the storage as a
network share. File management traffic and small data accesses travel
over the LAN while larger accesses make use of the SAN. Which network
is used can be tuned by file size and file type.
Performance is good with large files but when dealing with lots of
small files and frequent file system changes such as file creation and
deletion performance suffers. In general most SAN file systems seem to
suffer from this so you would need to do some testing.
As long as the hardware you have is reasonable modern there should not
have any problems. We did one sale / installation into a site where
some of the machines had ISA 10BT network cards and the performance
was less than stellar.
| |
|
| If you are surveying this technology area you might want also want to
include Red Hat GFS (formerly Sistina) or IBM's SAN File System (also
known as Storage Tank in some circles).
Faeandar wrote:
> Anyone have comments regarding a comparison of the two? I'm fairly
> well versed in Polyserve so anyone with background in SANergy would be
> a welcome commentator.
>
> I find little information, even from IBM, on this product; it's
> requirements, performance, and architecture. The basics are there but
> I need to understand it much better than the marketing slides allow.
>
> Thanks.
>
> ~F
| |
| Faeandar 2005-06-30, 8:46 pm |
| On 30 Jun 2005 10:25:42 -0700, "mf" <marc@buildingstorage.com> wrote:
[vbcol=seagreen]
>If you are surveying this technology area you might want also want to
>include Red Hat GFS (formerly Sistina) or IBM's SAN File System (also
>known as Storage Tank in some circles).
>
>Faeandar wrote:
Thanks but GFS is a very poor performer. Against Polyserve it was 30%
slower, and that was PS out of the box and GFS configured by an
engineer.
IBM's SANFs has a decent paper presentation but like so many other
products it doesn't do NFS failover.
~F
| |
|
| SGI's CXFS will not only out perform SANergy and Polyserve, it will
also do NFS/Samba failover.
David
| |
| Faeandar 2005-07-01, 5:49 pm |
| On 1 Jul 2005 02:26:34 -0700, "Chatz" <David.Chatterton@gmail.com>
wrote:
>SGI's CXFS will not only out perform SANergy and Polyserve, it will
>also do NFS/Samba failover.
>
>David
Actual failover? With file handles and everything? Where mid-stream
writes will continue after a few retries?
So many products think NFS failover is simply restarting the service
on another host. This is by no means failover. Anything that
requires clients to remount is not failover.
But if CXFS can do this then they're a possibility. Do they run on
Windows as well? That's one of Polyserve's strengths in my case.
Thanks.
~F
| |
|
| The NFS/Samba failover is the usual two machine setup behind a virtual
IP address. The NFS/Samba clients think they are talking to the same
machine even though it has failed over. No remount required.
http://www.sgi.com/products/software/failsafe/iris.html
NFS/Samba scale exception well under IRIX, SGI now claim up to 800MB/s
for streaming reads, 400MB/s streaming writes, on an 8CPU and 8x1Gb
NICs for NFS.
CXFS then gives you the ability for those machines and others (IRIX,
linux, windows etc) to work with a clustered filesystem. ie each
machine is on the SAN accessing the same filesystem at fibre channel
speeds, rather than NFS/network speeds.
http://www.sgi.com/products/storage...le_systems.html
Its then possible to have CXFS clients exporting the shared filesystem
via NFS or Samba to scale even further, but in general SGI recommend
just exporting from the CXFS server.
David
|
|
|
|
|