Обсуждение: SSD selection

On 5/15/2012 9:21 AM, Віталій Тимчишин wrote:
>
>
> We've reached to the point when we would like to try SSDs. We've got a
> central DB currently 414 GB in size and increasing. Working set does
> not fit into our 96GB RAM server anymore.
> So, the main question is what to take. Here what we've got:
> 1) Intel 320. Good, but slower then current generation sandforce drives
> 2) Intel 330. Looks like cheap 520 without capacitor
> 3) Intel 520. faster then 320 No capacitor.
> 4) OCZ Vertex 3 Pro - No available. Even on OCZ site
> 5) OCZ Deneva - can't find in my country :)
>

Is the 710 series too costly for your deployment ?
I ask because that would be the obvious choice for a database (much
better write endurance than any of the drives above, and less likely to
suffer from firmware bugs or unpleasant GC behavior).
We've been running them in production for a few months with zero
problems and great performance.
The price on the 710's tends to vary on whether they're in stock :
NewEgg is currently showing $1100 for the 300G drive, but no stock.

On Tue, May 15, 2012 at 12:08 PM, David Boreham <david_list@boreham.org> wrote:
>> We've reached to the point when we would like to try SSDs. We've got a
>> central DB currently 414 GB in size and increasing. Working set does not fit
>> into our 96GB RAM server anymore.
>> So, the main question is what to take. Here what we've got:
>> 1) Intel 320. Good, but slower then current generation sandforce drives
>> 2) Intel 330. Looks like cheap 520 without capacitor
>> 3) Intel 520. faster then 320 No capacitor.
>> 4) OCZ Vertex 3 Pro - No available. Even on OCZ site
>> 5) OCZ Deneva - can't find in my country :)
>>
>
> Is the 710 series too costly for your deployment ?
> I ask because that would be the obvious choice for a database (much better
> write endurance than any of the drives above, and less likely to suffer from
> firmware bugs or unpleasant GC behavior).
> We've been running them in production for a few months with zero problems
> and great performance.
> The price on the 710's tends to vary on whether they're in stock : NewEgg is
> currently showing $1100 for the 300G drive, but no stock.

this.  I think you have two choices today -- intel 320 and intel 710
depending on how much writing you plan to do.  ocz vertex 3 might be a
3rd choice, but it's vaporware atm.

merlin

On Tue, May 15, 2012 at 8:21 AM, Віталій Тимчишин <tivv00@gmail.com> wrote:
> We are using Areca controller with BBU. So as for me, question is: Can 520
> series be set up to handle fsyncs correctly?

No.

The cause for capacitors on SSD logic boards is that fsyncs aren't
flushed to NAND media, and hence persisted, immediately. SSDs are
divided into "pages", called "erase blocks" (usually much larger than
the filesystem-level block size; I don't know offhand what the block
size is on the 710, but on the older X-25 drives, it was 128K).  All
writes are accumulated in the on-board cache into erase block sized
chunks, and *then* flushed to the NAND media. In a power-loss
situation, the contents of that cache won't be preserved unless you
have a capacitor.  In some drives, you can disable the on-board cache,
but that does absolutely atrocious things both to your drive's
performance, and its longevity.

As the other posters in this thread have said, your best bet is
probably the Intel 710 series drives, though I'd still expect some
320-series drives in a RAID configuration to still be pretty
stupendously fast.

rls

--
:wq

On 5/15/2012 12:16 PM, Rosser Schwarz wrote:
> As the other posters in this thread have said, your best bet is
> probably the Intel 710 series drives, though I'd still expect some
> 320-series drives in a RAID configuration to still be pretty
> stupendously fast.
One thing to mention is that the 710 are not faster than 320 series
(unless in your definition of fast you count potential GC pauses of course).
The 710's primary claim to fame is that it has endurance and GC
characteristics designed for server and database use (constant load,
heavy write load).

So 320 drives will be just as fast, if not faster, but they will wear
out much more quickly (possibly not a concern for the OP in his
deployment) and may suffer from unwelcome GC pauses.

On Tue, May 15, 2012 at 3:00 PM, David Boreham <david_list@boreham.org> wrote:
> On 5/15/2012 12:16 PM, Rosser Schwarz wrote:
>>
>> As the other posters in this thread have said, your best bet is
>> probably the Intel 710 series drives, though I'd still expect some
>> 320-series drives in a RAID configuration to still be pretty
>> stupendously fast.
>
> One thing to mention is that the 710 are not faster than 320 series (unless
> in your definition of fast you count potential GC pauses of course).
> The 710's primary claim to fame is that it has endurance and GC
> characteristics designed for server and database use (constant load, heavy
> write load).
>
> So 320 drives will be just as fast, if not faster, but they will wear out
> much more quickly (possibly not a concern for the OP in his deployment) and
> may suffer from unwelcome GC pauses.

Although your assertion 100% supported by intel's marketing numbers,
there are some contradicting numbers out there that show the drives
offering pretty similar performance.  For example, look here:
http://www.anandtech.com/show/4902/intel-ssd-710-200gb-review/4 and
you can see that 4k aligned writes are giving quite similar results
(14k iops) even though the 710 is only rated for 2700 iops while the
320 is rated for 21000 IOPS.  Other benchmarks also show similar
results.

???

I have a theory that Intel rates their drives for IOPS based on the
results of 'Wheel of Fortune'.  This will be confirmed when you start
seeing drives with ratings of 'Bankrupt', 'Trip to Las Vegas', etc.
These must be the same guys that came up with the technical
explanation for the write through caching for the X25-M.

merlin

On 5/16/2012 11:01 AM, Merlin Moncure wrote:
> Although your assertion 100% supported by intel's marketing numbers,
> there are some contradicting numbers out there that show the drives
> offering pretty similar performance.  For example, look here:
> http://www.anandtech.com/show/4902/intel-ssd-710-200gb-review/4  and
> you can see that 4k aligned writes are giving quite similar results
> (14k iops) even though the 710 is only rated for 2700 iops while the
> 320 is rated for 21000 IOPS.  Other benchmarks also show similar
> results.
Actually I said the same thing you're saying : that the two series will
deliver similar performance.

The spec numbers however would be for worst case conditions (in the case
of the 710).
I'm not convinced that those tests were exercising the worst case part
of the envelope.

On Wed, May 16, 2012 at 12:45 PM, David Boreham <david_list@boreham.org> wrote:
> On 5/16/2012 11:01 AM, Merlin Moncure wrote:
>>
>> Although your assertion 100% supported by intel's marketing numbers,
>> there are some contradicting numbers out there that show the drives
>> offering pretty similar performance.  For example, look here:
>> http://www.anandtech.com/show/4902/intel-ssd-710-200gb-review/4  and
>> you can see that 4k aligned writes are giving quite similar results
>> (14k iops) even though the 710 is only rated for 2700 iops while the
>> 320 is rated for 21000 IOPS.  Other benchmarks also show similar
>> results.
>
> Actually I said the same thing you're saying : that the two series will
> deliver similar performance.
>
> The spec numbers however would be for worst case conditions (in the case of
> the 710).
> I'm not convinced that those tests were exercising the worst case part of
> the envelope.

Yeah -- you might be right -- their numbers are based on iometer which
looks like it runs lower than other tests (see here:
http://www.storagereview.com/intel_ssd_710_series_review_200gb).   I
still find it interesting the 320 is spec'd so much higher though.   I
guess I spoke to soon -- it looks it has to do with the life extending
attributes of the drive.  Benchmarks are all over the place though.

merlin

On 05/16/2012 01:01 PM, Merlin Moncure wrote:
> Although your assertion 100% supported by intel's marketing numbers,
> there are some contradicting numbers out there that show the drives
> offering pretty similar performance.  For example, look here:
> http://www.anandtech.com/show/4902/intel-ssd-710-200gb-review/4 and
> you can see that 4k aligned writes are giving quite similar results
> (14k iops) even though the 710 is only rated for 2700 iops while the
> 320 is rated for 21000 IOPS.  Other benchmarks also show similar
> results.

I wrote something talking about all the ways the two drives differ at
http://blog.2ndquadrant.com/intel_ssds_lifetime_and_the_32/

What the 710 numbers are saying is that you can't push lots of tiny
writes out at a high IOPS without busting the drive's lifetime
estimates.  You can either get a really high IOPS of small writes (320)
or a smaller IOPS of writes that are done more efficiently in terms of
flash longevity (710).  You can't get both at the same time.  The 710
may ultimately throttle its speed back to meet lifetime specifications
as the drive fills, it's really hard to benchmark the differences
between the two series.

--
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com