[THIN] Re: Alternative fileservers (EMC Celerra)

Hi Matt,

I have an active pain avoidance mechanism so once I get burned with a
product I tend to avoid it, and advise others to do the same. Sometimes
that's unfair because a product has actually improved in the interim but you
genarally only get one chance with me.

The deal with the Celerra was that despite EMC protestations that it ws
optimally tuned, whatever that means in a heterogeneous environment, we were
seeing system pauses that just had to be network i/o related with
consistently high redirector current commands on the TS systems.

If you've got a windows based file server, the default values (set at the
file server end)  are way too low for TS systems and will often result in TS
pauses and even hangs. But you can tweak maxmpxct and maxworkitems on the
file server to increase the resilience of the network i/o request mechanism
and get rid of the hangs.

Many NAS CIFS implementations use SAMBA or a derivative and with SAMBA you
can tune the Max Mux value (equivalent to maxmpxct) to tailor it for a TS
environment. But if the NAS implementation is a closed box then you've got a
problem. Approaches to EMC about this issue didn't produce any results at
the time so it was a choice of living with the pause issues or using
something else.

This may no longer be the case with EMC. Provided the Max Mux/(could also be
MaxConnections) or MaxMPxCT equivalent may now be tunable so that everything
will be fine. But if they aren't, then the Celerra remains as a possible bad
choice for a busy TS environment, particularly if you're doing folder
redirection and all that other fun stuff that increases network i/o
requests.

regards,

Rick

-- 
Ulrich Mack
Quest Software
Provision Networks Division


On 3/26/08, M <mathras@xxxxxxxxxxxxxxxx> wrote:
>
>  Hello Rick,
>
> Can you elaborate on your interesting comment.
> "The Celerra style NAS approach can have throughput problems in a TS
> environment. It's fine anywhere else"
>
> I am interested because we use this (it also cost a bomb so im told) and
> have had real issues with pausing Citrix sessions. Our Citrix
> team eventually pinpointed this to the way the disks were configured and the
> way departmental data was distributed.
>
> Matt
>
>
>
>
>

Other related posts: