SEO

#1 (**permalink**) 01-04-2008, 09:40 PM

The cache on the ESS makes using RAID5 less troublesome than doing so
on SSA does; but I was curious when write acknowledgements occur. Is
a write acknowledged after it is written to cache, or after it is
written to disk?

Thanks,
Chris

#2 (**permalink**) 01-04-2008, 09:40 PM

On 16 Feb 2004 14:59:38 -0800, chrish@po.wr.com (Chris) wrote:

>The cache on the ESS makes using RAID5 less troublesome than doing so
>on SSA does; but I was curious when write acknowledgements occur. Is
>a write acknowledged after it is written to cache, or after it is
>written to disk?
>
>Thanks,
>Chris

Write is acknowledged as soon as the data is written to the
battery-backed/accessable by both clusters write cache. (Or is it nvram
and not battery backed -- can't remember right now). In any case, you need
to loose Both sides of the ESS before you loose the data.

Tom

#3 (**permalink**) 01-04-2008, 09:40 PM

On Mon, 16 Feb 2004 22:05:31 -0500, TomK wrote:

> On 16 Feb 2004 14:59:38 -0800, chrish@po.wr.com (Chris) wrote:
>
>>The cache on the ESS makes using RAID5 less troublesome than doing so
>>on SSA does; but I was curious when write acknowledgements occur. Is
>>a write acknowledged after it is written to cache, or after it is
>>written to disk?
>>
>>Thanks,
>>Chris
>
> Write is acknowledged as soon as the data is written to the
> battery-backed/accessable by both clusters write cache. (Or is it nvram
> and not battery backed -- can't remember right now). In any case, you need
> to loose Both sides of the ESS before you loose the data.
>
> Tom

We have literally had a UPS burn up, crashing 70+ servers simultaneously.
The Shark used it's battery powered time to stage cached writes to disk
before terminating. Other than a few fsck's when the systems all came
back up (and not all were AIX, and all were not even journalled
filesystems), there were no problems. Very, very reliable storage array.

#4 (**permalink**) 01-04-2008, 09:40 PM

"F. Michael Orr" <morr@utility.vccs.edu> wrote in message
news

an.2004.02.17.14.58.48.426453@utility.vccs.e du...
> On Mon, 16 Feb 2004 22:05:31 -0500, TomK wrote:
>
> > On 16 Feb 2004 14:59:38 -0800, chrish@po.wr.com (Chris) wrote:
> >
> >>The cache on the ESS makes using RAID5 less troublesome than doing so
> >>on SSA does; but I was curious when write acknowledgements occur. Is
> >>a write acknowledged after it is written to cache, or after it is
> >>written to disk?
> >>
> >>Thanks,
> >>Chris
> >
> > Write is acknowledged as soon as the data is written to the
> > battery-backed/accessable by both clusters write cache. (Or is it nvram
> > and not battery backed -- can't remember right now). In any case, you
need
> > to loose Both sides of the ESS before you loose the data.
> >
> > Tom
>
> We have literally had a UPS burn up, crashing 70+ servers simultaneously.
> The Shark used it's battery powered time to stage cached writes to disk
> before terminating. Other than a few fsck's when the systems all came
> back up (and not all were AIX, and all were not even journalled
> filesystems), there were no problems. Very, very reliable storage array.
>
Michael,

every write goes to cache of the cluster and it's NVS which sits in the
adjacent cluster. As soon as this is done, then the write is acknowledged.
The write is later destaged to disk.

The NVS is battery backed for 7 days and there is sufficient power destage
if there is a power outage.

The size of the NVS depends on the model of ESS you have.

Regards Paresh

#5 (**permalink**) 01-04-2008, 09:40 PM

"F. Michael Orr" <morr@utility.vccs.edu> wrote:
>[ESS] Very, very reliable storage array.

Though I have yet to meet one where the doors close properly...

Cu
Jan
--
( )
The above .sig is equipped with state-of-the-art stealth technology.

#6 (**permalink**) 01-04-2008, 09:40 PM

On Tue, 17 Feb 2004 23:09:13 +0100, Jan van Beers <tnp@gmx.ch> wrote:

>"F. Michael Orr" <morr@utility.vccs.edu> wrote:
>>[ESS] Very, very reliable storage array.
>
>Though I have yet to meet one where the doors close properly...
>
>Cu
>Jan

Our IBM storage consultant told us that was the main reason the ESS
800 has got a corrugated door, to make it a bit more rigid and not
flap about

Steve

#7 (**permalink**) 01-04-2008, 09:41 PM

Steven Langdale <steven.langdale@toyland.demon.co.uk> wrote:
>On Tue, 17 Feb 2004 23:09:13 +0100, Jan van Beers <tnp@gmx.ch> wrote:
>
>>"F. Michael Orr" <morr@utility.vccs.edu> wrote:
>>>[ESS] Very, very reliable storage array.

Oh, almost forgot:
I copy that, we had only one (of a dozen or so, would have to count)
really fail in the past three years, this had its primary cluster node
dead, but not dead enough for the second one to take over. So it was
just sittung around and creating warm air, but it lost _no_ _data_.

>>Though I have yet to meet one where the doors close properly...
>
>Our IBM storage consultant told us that was the main reason the ESS
>800 has got a corrugated door, to make it a bit more rigid and not
>flap about

ESS 800 are the new ones which bear a Logo which says "Total Storage"?
Then I should go over and have a look at their doors

btw: Is it true that you can´t get those fins for on top of the shark
anymore?

>Steve
Cu
Jan

#8 (**permalink**) 01-04-2008, 09:41 PM

On Wed, 18 Feb 2004 16:59:06 +0100, Jan van Beers <tnp@gmx.ch> wrote:

>btw: Is it true that you can´t get those fins for on top of the shark
>anymore?
>
>Cu
>Jan

You may be able to get them, the same storage consultant also told me
(although he was smiling at the time) that the 800 development name
was the silvertip, and you can get a fin with a silver tip!

I look after a couple of F20's and wanted fins, but they wanted to
charge for them!!!

Steven

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

SEO

write consistency on ESS