Comments on: Sandisk’s $600 solid-state drive could be a game-changer. 62MB/s, 7,000 IOPS in it’s first rev.

By: Alex Gorbachev

Alex Gorbachev — Sun, 14 Jan 2007 05:52:45 +0000

Michael,
We have exactly the same idea here! It’s not clear so far how they would perform for writes as flash might be really slow on writes. Specs give write performance a bit slower that read but it only gives throughput and not number of writes per second.
I tried to contact SanDisk but email they provided in the brochure has bounced. I will try to follow it up on the first occasion. If you get to test it – please share your results.
Thanks!

By: Alex Gorbachev

Alex Gorbachev — Sun, 14 Jan 2007 05:51:13 +0000

Michael, we have exactly the same idea here! It’s not clear so far how they would perform for writes as flash might be really slow on writes. Specs give write performance a bit slower that read but it only gives throughput and not number of writes per second.
I tried to contact SanDisk but email they provided in the brochure has bounced. I will try to follow it up on the first occasion. If you get to test it – please share your results. Thanks!

By: Michael Friedman

Michael Friedman — Fri, 12 Jan 2007 05:27:18 +0000

Has anyone done any experiments with this for benchmarks?

I’m seriously thinking about suggesting that my customers buy two of these in a mirrored config – most of my customers can fit their DB in 32 gig.

By: Christo Kutrovsky

Christo Kutrovsky — Thu, 11 Jan 2007 21:28:05 +0000

“7,000 IOPS (input/outputs per second) for a 512MB file”

Note the small text, for a 512 Mb file. I wonder if that is still the number for the entire 32gb.

For comparission, IDE drives tend to be twice as fast, if you measure IOs in limited range, within 5% of the drive size.

By: Paul Vallee

Paul Vallee — Thu, 11 Jan 2007 21:08:59 +0000

Interesting review here.

Take aways:

Another touted benefit is faster file access times — a sum of seek time and access latency. SanDisk says it measured access times as low as 0.12ms for the SSD, compared to 19ms for the 1.8-inch, 4200 rpm hard drive.

Umm… wow?

SanDisk’s radical performance claims notwithstanding, NAND flash in general does suffer from a few undesirable performance characteristics, compared to traditional hard drives. For example, data can only be erased in fairly large segments — meaning that especially small write operations could take longer and use more energy.

Hmm… damn?

By: Paul Vallee

Paul Vallee — Thu, 11 Jan 2007 18:58:04 +0000

Another great pun! How many slaves do you think I can drive?

By: Doug Burns

Doug Burns — Thu, 11 Jan 2007 18:55:06 +0000

I have an outstanding email sitting in my Inbox from a vendor who wants to loan me Solid State Disks to repeat my ‘How Many Slaves’ PX tests. I don’t think I’ll get round to it for a while, though, you should meet my new boss – he’s a slave-driver!

;-)

By: Paul Vallee

Paul Vallee — Thu, 11 Jan 2007 18:50:55 +0000

Pete –

> I think I would need to buy rather a lot of them to store the odd terabyte

No doubt. But they’d be fast enough that I could even back putting them in … wait for it … gag … RAID 5!!! :-)

Paul

By: Pete_s

Pete_s — Thu, 11 Jan 2007 18:49:10 +0000

I think I would need to buy rather a lot of them to store the odd terabyte ;-)

But smart design and aligning data to storage speeds to suit need and I could see a new career developing.

By: Alex Gorbachev

Alex Gorbachev — Thu, 11 Jan 2007 18:34:57 +0000

Well, if it sustains 7K random IOs/second than it’s good for much higher volumes per “spindle”. Heh, would you still call it a spindle? :)