Thread: Choosing between Intel 320, Intel 510 or OCZ Vertex 3 SSD for db server

A few quick thoughts:

1. 320 would be the only SSD I'd trust from your short-list. It's the
only one with proper protection from unexpected power loss.
2. Multiple RAID'ed SSDs sounds like (vast) overkill for your workload.
A single SSD should be sufficient (will get you several thousand TPS on
pgbench for your DB size).
3. Consider not using the magnetic disks at all (saves on space, power
and the cost of the RAID controller for them).
4. Consider using Intel 710 series rather than 320 (pay for them with
the money saved from #3 above). Those devices have much, much higher
specified endurance than the 320s and since your DB is quite small you
only need to buy one of them.

On 10/24/2011 8:09 AM, Amitabh Kant wrote:
> Hello
>
> I need to choose between Intel 320 , Intel 510 and OCZ Vertex 3 SSD's
> for my database server. From recent reading in the list and other
> places, I have come to understand that OCZ Vertex 3 should not be
> used, Intel 510 uses a Marvel controller while Intel 320 had a nasty
> bug which has been rectified. So the list narrows down to only 510 and
> 320, unless I have understood the OCZ Vertex reviews incorrectly.
>
> The server would itself be built along these lines: Dual CPU Xeon
> 5620, 32 or 48 GB RAM, 2 SAS 10K disk in RAID 1 for OS, 2 SAS 10K disk
> in RAID 1 for pg_xlog and 4 SSD in RAID 10 for data directory
> (overkill??). OS would be FreeBSD 8.2 (I would be tuning the sysctl
> variables). PG version would be 9.1 with replication set to another
> machine (Dual CPU Xeon 54xx, 32 GB RAM, 6 15K SAS 146 GB: 4 in RAID 10
> for data and 2 in RAID 1 for OS + pg_xlog). The second machine hosts
> my current db , and there is not much of an issue with the
> performance. We need better redundancy now(current was to take a
> dump/backup every 12 hours), so the new machine.
>
> My database itself is not very big, approx 40 GB as of now, and would
> not grow beyond 80 GB in the next year or two. There are some tables
> where insert & updates are fairly frequent. From what I could gather,
> we are not doing more than 300-400 tps at the moment, and the growth
> should not be very high in the short term.
>
> Hope someone can give some pointers to which SSD I should go for at
> the moment.
>
>
> Amitabh
>

On Mon, Oct 24, 2011 at 11:53 AM, David Boreham <david_list@boreham.org> wrote:
>
> A few quick thoughts:
>
> 1. 320 would be the only SSD I'd trust from your short-list. It's the only
> one with proper protection from unexpected power loss.

yeah.

> 2. Multiple RAID'ed SSDs sounds like (vast) overkill for your workload. A
> single SSD should be sufficient (will get you several thousand TPS on
> pgbench for your DB size).

Also, raid controllers interfere with TRIM.

> 3. Consider not using the magnetic disks at all (saves on space, power and
> the cost of the RAID controller for them).

Agree.  If one SSD did not deliver the tps, I'd consider buying more
and optimizing with jbod/tablespaces -- really doubt that's necessary
however. Maybe a single large slow magnetic disk is a good idea for
retaining backups though.

> 4. Consider using Intel 710 series rather than 320 (pay for them with the
> money saved from #3 above). Those devices have much, much higher specified
> endurance than the 320s and since your DB is quite small you only need to
> buy one of them.

710 is good idea if and only if you are worried about write durability
(in which case it's a great idea).

merlin

On Mon, Oct 24, 2011 at 6:31 PM, Merlin Moncure <mmoncure@gmail.com> wrote:
>> 2. Multiple RAID'ed SSDs sounds like (vast) overkill for your workload. A
>> single SSD should be sufficient (will get you several thousand TPS on
>> pgbench for your DB size).
>
> Also, raid controllers interfere with TRIM.

What about redundancy?

How do you swap an about-to-die SSD?

Software RAID-1?

On 10/24/2011 3:31 PM, Merlin Moncure wrote:
>> 4. Consider using Intel 710 series rather than 320 (pay for them with the
>> >  money saved from #3 above). Those devices have much, much higher specified
>> >  endurance than the 320s and since your DB is quite small you only need to
>> >  buy one of them.
> 710 is good idea if and only if you are worried about write durability
> (in which case it's a great idea).

I disagree with this (that it is the only reason to select 710 series).
The write endurance (specified at least) is orders of magnitude higher.
Doing 100's of TPS constantly it is possible to burn through the 320's
endurance
lifetime in a year or two.

On 10/24/2011 4:47 PM, Claudio Freire wrote:
> What about redundancy?
>
> How do you swap an about-to-die SSD?
>
> Software RAID-1?

The approach we take is that we use 710 series devices which have
predicted reliability similar to all the other components in the
machine, therefore the unit of replacement is the entire machine. We
don't use trays for example (which saves quite a bit on data center
space). If I were running short endurance devices such as 320 series I
would be interested in replacing the drives before the machine itself is
likely to fail, but I'd do so by migrating the data and load to another
machine for the replacement to be done offline. Note that there are
other operations procedures that need to be done and can not be done
without downtime (e.g. OS upgrade), so some kind of plan to deliver
service while a single machine is down for a while will be needed
regardless of the storage device situation.

Am 24.10.2011 16:09, schrieb Amitabh Kant:

> while Intel 320 had a nasty bug which has been rectified

Be careful with that Intel SSD.
This one is still very buggy.
Take a look at the Intel forums
http://communities.intel.com/community/tech/solidstate?view=discussions#/ about
users who are complaining that they´ve lost all their data.
Even the firmware upgrade didn´t completely resolved the issues.

Regards
Thilo

On Mon, Oct 24, 2011 at 11:37 PM, David Boreham <david_list@boreham.org> wrote:
>> What about redundancy?
>>
>> How do you swap an about-to-die SSD?
>>
>> Software RAID-1?
>
> The approach we take is that we use 710 series devices which have predicted
> reliability similar to all the other components in the machine, therefore
> the unit of replacement is the entire machine. We don't use trays for
> example (which saves quite a bit on data center space). If I were running
> short endurance devices such as 320 series I would be interested in
> replacing the drives before the machine itself is likely to fail, but I'd do
> so by migrating the data and load to another machine for the replacement to
> be done offline. Note that there are other operations procedures that need
> to be done and can not be done without downtime (e.g. OS upgrade), so some
> kind of plan to deliver service while a single machine is down for a while
> will be needed regardless of the storage device situation.

Interesting.

But what about unexpected failures. Faulty electronics, stuff like that?

I really don't think a production server can work without at least raid-1.

On 10/25/2011 8:55 AM, Claudio Freire wrote:
> But what about unexpected failures. Faulty electronics, stuff like
> that? I really don't think a production server can work without at
> least raid-1.

Same approach : a server either works or it does not. The transition
between working and not working may be expected or not expected. The
solution is the same : use another machine to perform the work the now
not working machine was doing. The big benefit of this approach is that
you now do not need to worry about specific kinds of failure or
individual components, including storage.

If it helps, think of this architecture as raid-1, but with the whole
machine being the "drive" rather than individual storage devices.

On 10/28/2011 12:40 AM, Amitabh Kant wrote:
>
>
> Sadly, 710 is not that easily available around here at the moment.
>

All three sizes are in stock at newegg.com, if you have a way to export
from the US to your location.