Intel SRCS16 SATA raid?

This is a different thread that the $7k server thread.
Greg Stark started it and wrote:

 "I'm also wondering about whether I'm better off with one of these
 SATA raid                                                          
 controllers or just going with SCSI drives."                      

Rick

pgsql-performance-ow @postgresql.org wrote on 04/15/2005 10:01:56 AM:



broadcast)--------------------­-------

---------------------------(en­d of broadcast)--------------------­-------
TIP 5: Have you checked our extensive FAQ?

               http://www.postgresql.org/docs /faq

Perhaps you didn't read the whole page. It says right at the beginning:

"Because of its simplicity and short test duration, the ATTO Disk
Benchmark is used a lot for comparing the 'peformance' of hard disks.
The tool measures the sequential transfer rate of a partition using a
test length of 32MB at most. Because of this small dataset, ATTO is
unsuitable for measuring media transfer rates of intelligent
RAID-adapters which are equipped with cache memory. The smart RAID
adapters will serve the requested data directly from their cache, as a
result of which the results have no relationship to the media transfer
rates of these cards. For this reason ATTO is an ideal tool to test the
cache transfer rates of intelligent RAID-adapters."

Therefore, the results on this page are valid - they're supposed to show
the cache/transfer speed, the dataset is 32MB(!) and should fit in the
caches of all cards.

I don't, for many (historical) reasons.

Maybe, but with no mention of stripe size and other configuration
details, this is somewhat suspicious. I'll be able to offer benchmarks
for the 8506-8 vs. the 1120 shortly (1-2 weeks), if you're interested
(pg_bench, for example, to be a bit more on-topic).

Regards,
  Marinos

---------------------------(en­d of broadcast)--------------------­-------
TIP 4: Don't 'kill -9' the postmaster

The original thread was how much can I get for $7k

You can't fit a 15k RPM SCSI solution into $7K ;)  Some of us are on a budget!

10k RPM SATA drives give acceptable performance at a good price, thats
really the point here.

I have never really argued that SATA is going to match SCSI
performance on multidrive arrays for IO/sec.  But it's all about the
benjamins baby.  If I told my boss we need $25k for a database
machine, he'd tell me that was impossible, and I have $5k to do it.
If I tell him $7k - he will swallow that.  We don't _need_ the amazing
performance of a 15k RPM drive config.  Our biggest hit is reads, so
we can buy 3xSATA machines and load balance.  It's all about the
application, and buying what is appropriate.  I don't buy a Corvette
if all I need is a malibu.

Alex Turner
netEconomist

On 4/15/05, Dave Held <dave.h @arrayservicesgrp.com> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 7: don't forget to increase your free space map settings

I stand corrected!

Maybe I should re-evaluate our own config!

Alex T

(The dell PERC controllers do pretty much suck on linux)

On 4/15/05, Vivek Khera <v @khera.org> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 4: Don't 'kill -9' the postmaster

I didn't read the whole thing, but it didn't seem that bad to me.

And that's pretty much exactly what the article says.  Even before the
part you quoted.  Not sure what the problem is there.

They weren't pretending that the drive array was serving up data at
that rate directly from the physical media.  They clearly indicated
that they were testing controller cache speed with the small test.

If you take a close look, they pretty much outright say that the Areca
controller does very poorly on the random accesses typical of DB work.
They also specifically mention that the 3ware still dominates the
competition in this area.

Dave

__
David B. Held
Software Engineer/Array Services Group
200 14th Ave. East,  Sartell, MN 56377
320.534.3637 320.253.7800 800.752.8129

---------------------------(en­d of broadcast)--------------------­-------
TIP 8: explain analyze is your friend

And I would be willing to bet that the Atlas 10k is not using the
same generation of technology as the Raptors.

State that in terms of cars.  Would you be willing to pay 300% more
for a car that is 44% faster than your competitor's?  Of course you
would, because we all recognize that the cost of speed/performance
does not scale linearly.  Naturally, you buy the best speed that you
can afford, but when it comes to hard drives, the only major feature
whose price tends to scale anywhere close to linearly is capacity.

So get a Raptor for your WAL partition. ;)

Just keep in mind the points made by the Seagate article.  You're
buying much more than just performance for that $500+.  You're also
buying vibrational tolerance, high MTBF, better internal
environmental controls, and a pretty significant margin on seek time,
which is probably your most important feature for disks storing tables.
An interesting test would be to stick several drives in a cabinet and
graph how performance is affected at the different price points/
technologies/number of drives.

__
David B. Held
Software Engineer/Array Services Group
200 14th Ave. East,  Sartell, MN 56377
320.534.3637 320.253.7800 800.752.8129

---------------------------(en­d of broadcast)--------------------­-------
TIP 3: if posting/reading through Usenet, please send an appropriate
      subscribe-nomail command to majord @postgresql.org so that your
      message can get through to the mailing list cleanly

True, but that's a one time expense (300%) for a 44% gain ALL the time.
  '44% better' is nothing to sneeze at.  I'd easily pay the price for
the gain in a large server env.

--
Until later, Geoffrey

---------------------------(en­d of broadcast)--------------------­-------
TIP 2: you can get off all lists at once with the unregister command
    (send "unregister YourEmailAddressHere" to majord @postgresql.org)

no! 3ware was rather early in this business, but there are plenty of
(IMHO, and some other people's opinion) better alternatives available.
3ware has good Linux drivers, but the performance of their current
controllers isn't that good.

Have a look at this: http://www.tweakers.net/review s/557/1

especially the sequential writes with RAID-5 on this page:

http://www.tweakers.net/review s/557/19

We have been a long-time user of a 3ware 8506 controller (8 disks,
RAID-5) and have purchased 2 Areca ARC-1120 now since we weren't
satisfied with the performance and the 2TB per array limit...

-mjy

---------------------------(en­d of broadcast)--------------------­-------
TIP 4: Don't 'kill -9' the postmaster

No offense to that review, but it was really wasn't that good, and
drew bad conclusions from the data.  I posted it originaly and
immediately regretted it.

See http://www.tweakers.net/review s/557/18

Amazingly the controller with 1Gig cache manages a write throughput of
750MB/sec on a single drive.

quote:
"Floating high above the crowd, the ARC-1120 has a perfect view on the
struggles of the other adapters. "

It's because the adapter has 1Gig of RAM, nothing to do with the RAID
architecture, it's clearly caching the entire dataset.  The drive
can't physicaly run that fast.  These guys really don't know what they
are doing.

Curiously:
http://www.tweakers.net/review s/557/25

The 3ware does very well as a data drive for MySQL.

The size of your cache is going to _directly_ affect RAID 5
performance.  Put a gig of memory in a 3ware 9500S and benchmark it
against the Areca then.

Also - folks don't run data paritions on RAID 5 because the write
speed is too low.  When you look at the results for RAID 10, the 3ware
leads the pack.

See also:
http://www20.tomshardware.com/ storage/20041227/areca-raid6-0 6.html

I trust toms hardware a little more to set up a good review to be honest.

The 3ware trounces the Areca in all IO/sec test.

Alex Turner
netEconomist

On 4/15/05, Marinos Yannikos <m @geizhals.at> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 2: you can get off all lists at once with the unregister command
    (send "unregister YourEmailAddressHere" to majord @postgresql.org)

Dave wrote "An interesting test would be to stick several drives in a
cabinet and
graph how performance is affected at the different price points/
technologies/number of drives."

From the discussion on the $7k server thread, it seems the RAID controller
would
be an important data point also.  And RAID level.  And application
load/kind.

Hmmm.  I just talked myself out of it.  Seems like I'd end up with
something
akin to those database benchmarks we all love to hate.

Rick

pgsql-performance-ow @postgresql.org wrote on 04/15/2005 08:40:13 AM:



---------------------------(en­d of broadcast)--------------------­-------
TIP 6: Have you searched our list archives?

               http://archives.postgresql.org

Nice research Alex.

Your data strongly support the information in the paper.  Your SCSI drives
blew away the others in all of the server benchmarks.  They're only
marginally better in desktop use.

I do find it somewhat amazing that a 15K SCSI 320 drive isn't going to help
me play Unreal Tournament much faster.  That's okay.  I suck at it anyway.
My kid has never lost to me.  She enjoys seeing daddy as a bloody smear and
bouncing body parts anyway.  It promotes togetherness.

Here's a quote from the paper:

"[SCSI] interfaces support multiple initiators or hosts. The
drive must keep track of separate sets of information for each
host to which it is attached, e.g., maintaining the processor
pointer sets for multiple initiators and tagged commands.
The capability of SCSI/FC to efficiently process commands
and tasks in parallel has also resulted in a higher overhead
???kernel??? structure for the firmware."

Has anyone ever seen a system with multiple hosts or initiators on a SCSI
bus?  Seems like it would be a very cool thing in an SMP architecture, but
I've not seen an example implemented.

Rick

Alex Turner <arm @gmail.com> wrote on 04/14/2005 12:13:41 PM:

typeID=10&testbedID=3&osID=4&r­aidconfigID=1&numDrives=1&devI­D_0=232&devID_1=40&devID_2=259­&devID_3=267&devID_4=261&devID­_5=248&devCnt=6



com/content/docs/pdf/whitepape­r/D2c_More_than_Interface_ATA_­vs_SCSI_042003.pdf



broadcast)--------------------­-------
broadcast)--------------------­-------
---------------------------(en­d of broadcast)--------------------­-------
TIP 3: if posting/reading through Usenet, please send an appropriate
      subscribe-nomail command to majord @postgresql.org so that your
      message can get through to the mailing list cleanly

Just to clarify these are tests from http://www.storagereview.com, not
my own.  I guess they couldn't get number for those parts.  I think
everyone understands that a 0ms seek time impossible, and indicates a
missing data point.

Thanks,

Alex Turner
netEconomist

On 4/14/05, Dave Held <dave.h @arrayservicesgrp.com> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 6: Have you searched our list archives?

               http://archives.postgresql.org

Looking at the numbers, the raptor with TCQ enabled was close or beat
the Atlas III 10k drive on most benchmarks.

Naturaly a 15k drive is going to be faster in many areas, but it is
also much more expensive.  It was only 44% better on the server tests
than the raptor with TCQ, but it costs nearly 300% more ($538 cdw.com,
$180 newegg.com).  Note also that the 15k drive was the only drive
that kept up with the raptor on raw transfer speed, which is going to
matter for WAL.

For those of us on a budget, a quality controller card with lots of
RAM is going to be our biggest friend because it can cache writes, and
improve performance.  The 3ware controllers seem to be universally
benchmarked as the best SATA RAID 10 controllers where database
performance is concerned.  Even the crappy tweakers.net review had the
3ware as the fastest controller for a MySQL data partition in RAID 10.

The Raptor drives can be had for as little as $180/ea, which is quite
a good price point considering they can keep up with their SCSI 10k
RPM counterparts on almost all tests with NCQ enabled (Note that 3ware
controllers _don't_ support NCQ, although they claim their HBA based
queueing is 95% as good as NCQ on the drive).

Alex Turner
netEconomist

On 4/14/05, Dave Held <dave.h @arrayservicesgrp.com> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 1: subscribe and unsubscribe commands go to majord @postgresql.org

Well I have never even heard of it. 3ware is the defacto authority of
reasonable SATA RAID. If you were to
go with a different brand I would go with LSI. The LSI 150-6 is a nice
card with a battery backup option as well.

Oh and 3ware has  BBU for certain models as well.

Sincerely,

Joshua D. Drake

---------------------------(en­d of broadcast)--------------------­-------
TIP 8: explain analyze is your friend

I think it's a little misleading that your tests show 0ms seek times
for some of the write tests.  The environmental test also selects a
missing data point as the winner.  Besides that, it seems to me that
seek time is one of the most important features for a DB server, which
means that the SCSI drives are the clear winners and the non-WD SATA
drives are the embarrassing losers.  Transfer rate is import, but
perhaps less so because DBs tend to read/write small blocks rather
than large files.  On the server suite, which seems to me to be the
most relevant for DBs, the Atlas 15k spanks the other drives by a
fairly large margin (especially the lesser SATA drives).  When you
ignore the "consumer app" benchmarks, I wouldn't be so confident in
saying that the Raptors "put on a good show".

__
David B. Held
Software Engineer/Array Services Group
200 14th Ave. East,  Sartell, MN 56377
320.534.3637 320.253.7800 800.752.8129

---------------------------(en­d of broadcast)--------------------­-------
TIP 8: explain analyze is your friend

sorry, don't remember whether it's SCSI or SATA II, but IIRC
the Areca controllers are just stellar for things.

If you do get SATA for db stuff..especially multiuser...i still
haven't seen anything to indicate an across-the-board primacy
for SATA over SCSI. I'd go w/SCSI, or if SATA for $$$ reasons, I'd
be sure to have many spindles and RAID 10.

my 0.02. I'm surely not an expert of any kind.

I have put together a little head to head performance of a 15k SCSI,
10k SCSI 10K SATA w/TCQ, 10K SATA wo/TCQ and 7.2K SATA drive
comparison at storage review

http://www.storagereview.com/p hp/benchmark/compare_rtg_2001. php?typeI...

It does illustrate some of the weaknesses of SATA drives, but all in
all the Raptor drives put on a good show.

Alex Turner
netEconomist

On 4/14/05, Alex Turner &lt;arm @gmail.com> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 4: Don't 'kill -9' the postmaster

I have read a large chunk of this, and I would highly recommend it to
anyone who has been participating in the drive discussions.  It is
most informative!!

Alex Turner
netEconomist

On 4/14/05, Richard_D_Lev @raytheon.com <Richard_D_Lev @raytheon.com> wrote:



---------------------------(en­d of broadcast)--------------------­-------
TIP 6: Have you searched our list archives?

               http://archives.postgresql.org

Greg,

I posted this link under a different thread (the $7k server thread).  It is
a very good read on why SCSI is better for servers than ATA.  I didn't note
bias, though it is from a drive manufacturer.  YMMV.  There is an
interesting, though dated appendix on different manufacturers' drive
characteristics.

http://www.seagate.com/content /docs/pdf/whitepaper/D2c_More_ than_Inte...

Enjoy,

Rick

pgsql-performance-ow @postgresql.org wrote on 04/14/2005 09:54:45 AM:




---------------------------(en­d of broadcast)--------------------­-------
TIP 9: the planner will ignore your desire to choose an index scan if your
      joining column's datatypes do not match