|
Home > Archive > Data Storage > July 2005 > Any experiences with EMC NS500 series?
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 |
Any experiences with EMC NS500 series?
|
|
| cameron.kim@gmail.com 2005-06-17, 5:47 pm |
| I am looking for any experiences/opinions with the EMC NS501 or NS502
series. In particular in using the NAS gateway for CIFS/NFS and ISCSI
connected to a CX300 with a mixed shelf (1 shelf FC, 1 shelf ATA).
I also heard that the EMC takes a overall performance hit when trying
to rebuild a broken volume (due to disk failure). Is that right?
Thanks.
| |
|
| On 17 Jun 2005 07:06:47 -0700, cameron.kim@gmail.com wrote:
>I am looking for any experiences/opinions with the EMC NS501 or NS502
>series. In particular in using the NAS gateway for CIFS/NFS and ISCSI
>connected to a CX300 with a mixed shelf (1 shelf FC, 1 shelf ATA).
Personally, I like the NS products. They are stable, quick and
flexible with a decent feature set. Although the GUI has improved
immeasurably over the past couple of years, the CLI remains the most
complete management tool.
If you're minded to go with EMC for disk, you'll like the NS.
>I also heard that the EMC takes a overall performance hit when trying
>to rebuild a broken volume (due to disk failure). Is that right?
Technically the one word answer is: Yes. But so does pretty much
every disk array that uses RAID, especially RAID 4 or 5 (it has to
calculate the original data from the parity bits).
However, it depends on how you have configured it. The CX allows you
to set a priority for the rebuild. That is, you can tell it that
application i/o is far more (or less) important than the disk rebuild.
This allows you to choose how much of an impact the rebuild operation
has on performance of your application - many other arrays don't give
you this choice.
For the record, you have four choices on the CX: ASAP, High, Med, Low.
These pertain to the priority of the rebuild. ASAP is the quickest
rebuild, but will have the greatest potential impact on application
performance. Low will take the longest to rebuild the data but the
least potential to impact app performance.
Depending on your application, if it's not hammering the array, you
might not notice any performance degradation, even if it's set to
'ASAP'.
You have a choice. Look at what your app is doing. You need to
balance your application i/o prioriy against the likelihood of
multiple disk failures in a single RAID group.
HVB.
| |
| cameron.kim@gmail.com 2005-06-17, 5:47 pm |
| Thanks for the reply, HVB.
In my second question, i was particularly interested in the performance
impact... to FC shelf volumes, when there is a disk failure on an ata
shelf. With the NS500 trying to rebuild the ata shelf, i am wondering
if it impacts the performance on the FC shelves.
Also, how is the performance with snapshots?
I am comparing this unit to a Netapp system (using cifs/nfs/iscsi)
Ironically, your posting is the only positive thing i have heard
regarding this product. Everyone from competitors (thats expected),
previous EMC reps, and so forth, have informed of the either the poor
performance or the instability of this device.
Thanks in advance for your thoughts.
| |
|
| On 17 Jun 2005 08:58:08 -0700, cameron.kim@gmail.com wrote:
>Thanks for the reply, HVB.
No probs!
>In my second question, i was particularly interested in the performance
>impact... to FC shelf volumes, when there is a disk failure on an ata
>shelf. With the NS500 trying to rebuild the ata shelf, i am wondering
>if it impacts the performance on the FC shelves.
Rebuilding data from parity (i.e. RAID 5) will consume CPU cycles on
the Service Processor (controller). The more CPU cycles you use for
various activities (RAID rebuild, mirroring, etc.) the less is
available for regular I/O.
So, to answer your question, if you have a mix of FC and ATA disks,
and you have an ATA disk failure, performane of the OVERALL system
could be affected, depending on how you've set the rebuild priority.
I don't believe this is any different to the majority of modular
arrays available from other vendors (if anyone knows different, please
post!).
It's worth noting that the rebuild priority can be configured per LUN.
You could choose to have 'LOW' priority on your ATA LUNs to minimise
any potential performance hit on production I/O.
>Also, how is the performance with snapshots?
Are you asking if you kill disk performance by using snapshots?
No, not unless you're doing something really funky, in which case EMC
Engineering would be all over you like a rash anyway. I doubt you'd
be able to detect any performance difference with snapshots on or off.
>I am comparing this unit to a Netapp system (using cifs/nfs/iscsi)
Which NetApp box? Just out of interest.
Are you planning on using it just for regular file sharing, or are you
looking at applications also? (for EMC and NetApp)
>Ironically, your posting is the only positive thing i have heard
>regarding this product. Everyone from competitors (thats expected),
>previous EMC reps, and so forth, have informed of the either the poor
>performance or the instability of this device.
Lots of people seem to have a chip on their shoulder when it comes to
EMC and I think that's responsible for a lot of the negative press you
hear.
Other than Microsoft, I've never known a company to be as disliked by
so many people as EMC. I know EMC have a reputation for being
aggressive, but they do have some great products too. Not everybody
is willing to look past their prejudices! ;-)
There are not *that* many people, even in EMC, with direct experience
of the NAS products. They are too specialised for the regular account
teams and SEs - I'm not knocking the guys, but that's just the way it
is. There is a specialist group that know pretty much everything
there is to know about them, but outside of that, knowledge is patchy.
The big players in the NAS market are NetApp and EMC. EMC rules in
the highest echelons of that space. If you have HUGE (and I mean
50TB+) the scalability and management features of the EMC products
really make them the only game in town.
If you have less than that, and most of us do... well, NetApp own
about 80% of the market and for a great reason: they do what most
people want, most of the time.
Compared to NetApp, EMC's NAS GUI sucks. Compared to NetApp, EMC
don't have the fullest range of software functionality. If you really
want to use the Fibre Channel capabilities, EMC have a far better
story to tell. None of this matters if you don't need all those extra
features.
>Thanks in advance for your thoughts.
You're welcome!
HVB.
| |
| cameron.kim@gmail.com 2005-06-17, 5:47 pm |
| I am comparing the NS502G (clustered) with a CX300 or CX500 backend vs
the new Netapp 3020 with a shelf of fiber and a shelf of ata. (7 TB)
We have previous experience with a netapp, so we are somewhat inclined
to them. They run great and have yet to have them crash.
Primarily we will be using them for ISCSI volumes. It will also server
CIFS and NFS shares, however, we are looking at both products in
consolidating our storage. Around 15 servers, varying LUN sizes of
200GB to 3 TB).
Thanks again.
| |
| Ramesh Pun 2005-06-18, 2:46 am |
|
>
> Lots of people seem to have a chip on their shoulder when it comes to
> EMC and I think that's responsible for a lot of the negative press you
> hear.
>
Really?
> Other than Microsoft, I've never known a company to be as disliked by
> so many people as EMC. I know EMC have a reputation for being
> aggressive, but they do have some great products too. Not everybody
> is willing to look past their prejudices! ;-)
>
Really?Really?Really?
> There are not *that* many people, even in EMC, with direct experience
> of the NAS products. They are too specialised for the regular account
> teams and SEs - I'm not knocking the guys, but that's just the way it
> is. There is a specialist group that know pretty much everything
> there is to know about them, but outside of that, knowledge is patchy.
>
I doubt it. If that is true, then the dislike described above is justified.
| |
|
| On Fri, 17 Jun 2005 23:19:23 -0400, "Ramesh Pun" wrote:
>
>Really?
I think so, yes.
>
>I doubt it. If that is true, then the dislike described above is justified.
What do you doubt, specifically?
- That there are not many people with EMC NAS expertise?
- That the EMC sales and SE teams don't have much expertise in NAS?
- That there is a specialist NAS group in EMC?
I stand by all three of these statements.
HVB.
| |
|
| Hey Cameron & the NG,
On 17 Jun 2005 10:54:43 -0700, cameron.kim@gmail.com wrote:
>I am comparing the NS502G (clustered) with a CX300 or CX500 backend vs
>the new Netapp 3020 with a shelf of fiber and a shelf of ata. (7 TB)
EMC will tell you that you're not making a like-for-like comparison.
Although they can sell a single-headed NAS device, EMC like to use at
least two. This allows for Data Mover failover (or dual active if you
prefer), so it's a resonable assertion, in my view.
They'll tell you that you should be looking at a FAS2030C, because
that provides somewhere to failover (and doubles the price for the
NetApp kit).
However... (see below)
>We have previous experience with a netapp, so we are somewhat inclined
>to them. They run great and have yet to have them crash.
The NetApp kit seems extremely reliable, even in single Filer
environments. I think NetApp quote a real-world 99.6% uptime, or
something like that, measured against all Filers that report their
stats to NetApp (i.e. lots of them).
If you were to have multiple NetApp devices in your company, you
should also look at VFM software (NuView), as it will give you lots of
additional flexibility.
>Primarily we will be using them for ISCSI volumes. It will also server
>CIFS and NFS shares, however, we are looking at both products in
>consolidating our storage. Around 15 servers, varying LUN sizes of
>200GB to 3 TB).
I'd stick with NetApp, unless EMC are doing you a fantastic deal. If
you are used to Snapshot, SnapRestore and any of the other NetApp
software features, I think you'd be disappointed with the EMC way of
doing things. YMMV.
HVB.
| |
| carmelomcc 2005-07-15, 5:47 pm |
| remember if you have a NS it realies on the hotspare from the CX. The
nas does not do anything when it comes to a failed drive. The normal
amount of time for a Hot spare to buil in is 8 to 12 minutes. The I/O
will only take a suttle hit if you are running raid 5. Raid 10 or Raid
1 would not show the perfromance hit.
|
|
|
|
|