Skip to main content
Start a Conversation

Unsolved

This post is more than 5 years old

doof1

45 Posts

1100

June 26th, 2008 01:00

SSD benchmark

Hi

does someone test the new flash drive (ssd) in DMX4 (benchmark) ?

thanks

RRR

2 Intern

 • 

5.7K Posts

June 26th, 2008 05:00

I haven't heard of anyone actually having bought that thing. I looked up the price though ;)
By the way: the new clariion is supposed to support the SSD as well. I'm very curious when the new CX4 will become GA.

bodnarg

385 Posts

June 27th, 2008 07:00

With RRR - we've drooled over these for a few applications which might really benefit, but the cost is still just too high right now.

Saw some interesting presentations at EMC World and got the message beaten into my head that the real improvement of these drives comes from workloads with more random read misses since regular cache fed writes/reads are still faster than the SSD drives.

Never seen any benchmarks posted in the forum - not sure the moderators would allow that or not.

RRR

2 Intern

 • 

5.7K Posts

June 27th, 2008 07:00

Furthermore: you need at least 5 SSD's I think, because you want them in a RAID5 (3+1) with a hot spare and besides that, you need to place them in a 2Gb Quadrant of you DMX4, leaving you with a "slower" backend, just because you're having these SSD's in there.

My guess is that you'd be better off waiting just a bit longer. Prices will drop and perhaps the interface will go up to 4Gb, so you can place them in a regular 4Gb quadrant.

bodnarg

385 Posts

June 27th, 2008 09:00

Based on the presenations and discussions at the conference the 2Gb loop really was not a limit. The benefit is the 1-2ms response time on read-misses which pretty much mean you have much closer to uniform I/O response time for every type of I/O time driving down your overall response time. These are targetted to transactional systems that want faster response time not huge bandwidth which is where you could have the drives become "bottlenecked" by the back-end system.

Even for an application that could sustain enough I/O to choke on the bandwidth everything is relative - hitting your bottleneck after improving your throughput by 5-10x is certainly better than bottlenecking earlier than that. Performance is a wonderful thing - you are always waiting or bottlenecked on something ;)

xe2sdc

2 Intern

 • 

2.8K Posts

June 27th, 2008 10:00

Performance is a wonderful thing - you are always waiting or
bottlenecked on something ;)


The best phrase I heared on performances: "It depends" :-)

xe2sdc

2 Intern

 • 

2.8K Posts

June 27th, 2008 10:00

Never seen any benchmarks posted in the forum - not
sure the moderators would allow that or not.


Not sure .. however I think you'll never see benchmarks from anyone in EMC since only speed gurus are allowed to talk about speed and performances ]:)

xe2sdc

2 Intern

 • 

2.8K Posts

June 27th, 2008 10:00

Hmmm Not yet .. however it's a neat idea ;-)

Allen Ward

2.1K Posts

June 27th, 2008 10:00

Didn't EMC copywrite that phrase?

Allen Ward

2.1K Posts

June 27th, 2008 11:00

I guess I just assumed. Maybe I can copywrite it and incorporate and they'll buy my company :-)

ROTFLMAO

RRR

2 Intern

 • 

5.7K Posts

June 30th, 2008 03:00

It depends .... on what you mean, LOL

View More

View All

No Events found!

Top