[THIN] Re: OT: Effect of Solid State drives
- From: "Steve Greenberg" <steveg@xxxxxxxxxxxxxx>
- To: <thin@xxxxxxxxxxxxx>
- Date: Thu, 7 Jun 2007 09:26:19 -0700
Aside from the OS support part, haven't we done this already with RAID
controllers that have blocks of RAM on them? It always impresses me how much
faster these controllers are then ones without RAM. Taking that solid state
portion up to multiple gigabytes certainly would make a huge difference for
disk bottlenecks.
BTW- Why does Windows still depend so much on the hard drive when we are so
much RAM available?? Tim, I think this question is for you :-)
Steve Greenberg
Thin Client Computing
34522 N. Scottsdale Rd D8453
Scottsdale, AZ 85262
(602) 432-8649
www.thinclient.net
steveg@xxxxxxxxxxxxxx
_____
From: thin-bounce@xxxxxxxxxxxxx [mailto:thin-bounce@xxxxxxxxxxxxx] On Behalf
Of Joe Shonk
Sent: Thursday, June 07, 2007 8:43 AM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Re: OT: Effect of Solid State drives
I love that term "Half Pregnant", reminds me of the "chubby client" name.
Perhaps we should start calling Microsoft Vista "the dyslexic OS".
The "half pregnant" version is interesting. Combine "Ready Drive" with
Citrix's ASS (Ardence Software-Streaming) and you've got a kickass (pun
intented) solution.
Joe
From: thin-bounce@xxxxxxxxxxxxx [mailto:thin-bounce@xxxxxxxxxxxxx] On Behalf
Of Timothy R. Mangan
Sent: Thursday, June 07, 2007 8:23 AM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Re: OT: Effect of Solid State drives
Another, related technology is the coming "half pregnant" versions. You
will need a Vista or Server 2008 OS to make use of them, but basically they
add a bunch of flash on the drive to provide solid state performance for all
writes and most reads. The latter part is why you need OS support, because
the OS needs to tell the disk what is likely to be desired. It provides for
a more optimal read by directing file access that are likely to be small to
the solid state, while accesses that tend to be larger to pull from the hard
disk. Although latency is the big issue for small accesses (one or two
blocks), transfer speed is important if you are reading an entire file (such
as a remote file access), and solid state devices generally have lower
transfer rates than the hard disk once the heads are in place. Of course,
more parts (even if they are not moving parts) might mean less reliability.
Microsoft calls this "Ready Drive", although I have yet to see a drive
actually on the market (but I have not looked in a while).
"Ready Boost" which uses an external USB stick is of similar idea (but for
reads only) and can prove helpful on Vista machines that don't have enough
memory.
From: thin-bounce@xxxxxxxxxxxxx [mailto:thin-bounce@xxxxxxxxxxxxx] On Behalf
Of Rick Mack
Sent: Thursday, June 07, 2007 6:48 AM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Re: OT: Effect of Solid State drives
Hi Malcolm,
For sequential reads or writes of files that are small enough to fit into
one cylinder (all tracks/heads you can access without head movement),
traditional 15K rpm drives will be faster than SSDs. But that's right now.
With random i/o, latency starts becoming critical. An SSD with .1 msec seek
times or more correctly latency, vs about 4-8 msec average seek times (eg
latest 15K drives with an average seek time of 3.75 msec and an average
rotational latency of 2.0 msec) for hard disks, starts streaking ahead
because it can maintain the maximum data transfer rates largely regardless
of file distribution on the disk.
So SSDs can improve average disk latency by over 30 times, which is kind of
significant.
When you have to physically move a head over a spinning disk there are
things like rotational latency and track to track seek time that just can't
be ignored or avoided. A large on-board cache evens things out but with the
exception of hybrid drives as mentioned earlier, SSDs are going to start
seriously pulling ahead.
They can't crash heads, they don't spin, they don't have any moving parts.
Samsung quote an MTBF of 1 million hours for their SSDs (works out to 100+
years) which is kind of ridiculous because nothing electronic lasts that
long. But it says a lot about what you can expect in terms of reliability
from an SSD.
It's early days for SSDs, and the technology to stripe many SSDs together
doesn't exist outside fairly expensive storage systems. But motherboards
that support 4 SATA drives (often with RAID) are common enough that it'll be
interesting to see what develops.
The point I really wanted to make is that SSDs can solve major disk latency
bottlenecks in servers that really can't be overcome using traditional disk
technology. The other stuff is just speculating what else could happen if
low latency high speed storage subsystems become mainstream.
regards,
Rick
--
Ulrich Mack
Commander Australia
On 6/7/07, Malcolm Bruton <malcolm.bruton@xxxxxxxxxxxxxxxxxx> wrote:
Rick. Do you have any details of the performance? Did a quick Google but
didn't see much. I've read that at the moment there is no real difference
between speed of the flash drives vs. traditional laptop hard disks. I do
expect this to change though in the future. The new 2.5 inch 10,000 RPM SAS
drives are pretty fast vs. the old 3.5 inch 15 K SCSI drives and I know that
HP/Seagate are talking 15K 2.5 inch SAS versions soon as well.
The article was in a PC mag in the UK but I can't find an online version.
The Samsung drive was faster in some benchmarks when compared with a 5400
std notebook drives but not all tests
Malcolm
Other related posts:
