Dell and SSD drives

Tue Jan 27 20:33:16 CST 2009

Matthew Geier wrote:

>Stephen John Smoogen wrote:
>> I am looking at our mail servers and trying to get some disk IO
>> contentions down. I would like to know if people have used SSD's in a
>> Linux environment. I am looking at an SSD over a ramdisk for
>> persistence of storage in case of reboots. What I would like to do is
>> put in 2 SSD's into a 2950 and have them deal with the malware and
>> spam scanning directories as that seems to be the contention points.
>>
>>   
>Arn't those SSD units based on flash memory ?. A good solid write load,

>like for example a mail spool, would make them have a rather short life

>before flash cell failure.
>
>Or am I wrong and those SSD modules actually battery backed RAM ?, 
>getting speed and write cycles with out the 'wear' that flash has.

SSD's are NAND-flash based.  No battery backup involved.  There is an
issue with write-cycle limits for the two types of NAND flash used---SLC
(Single-Level Cell) and MLC (Multiple-Level Cell)---however, it's
overblown due to the "wear-leveling" algorithms built into all SSD
controllers.  

Very simplified version:  Each memory cell in an SLC-based device is
good for 100,000 write-cycles on average. MLC-based devices are good for
10,000. That does *not* mean write 10,000 times to the *SSD* and then
the device is junk, however.  It would be true if the same *memory cell*
was written to over and over again. But every time the data in an
allocated page is written to, the cells that make up that page aren't
simply updated.  Instead, the existing page contents and updated values
are written to a different location---selected from off the list of
"least written" pages.  By doing this (again..., very simplified view),
the flash controller ensures an even usage of the NAND flash memory
cells (avoiding "hot-spots" that fail quickly.)

A few things:

1) Write throughput:  Up to this point, SSD's manufacturers have been
all over the map with their real-world performance not matching
published specs. Rules-of-thumb are usually useless generalizations, but
I've found that if the manufacturer doesn't publish "sequential write"
throughput of at least 90+MB/sec, then the drive will likely disappoint.
This is hopefully changing with the entrance of Intel/Micron into the
marketplace.  Some of their SSD's do have great specs.  I've tested
Imation drives as well that are very good. The sector seems to be on the
verge up a big uptick.

2) NAND technology used: In general, SLC-based SSD's should be used in
server applications.  Both for the superior write performance of SLC and
also the 10X greater write-cycles per cell provided. SLC drives are more
expensive because of the lower storage density.  Check the
manufacturer's specs and ask if they aren't published.  Many
manufacturers seem to love hiding the NAND technology used. I'm not sure
what Dell sells...

3) A note on SSD's "wearing-out"... Whether it's an issue or not depends
on the quality of the flash controller used. There's a great deal of
overhead associated with writing to an SSD because a cell must first be
erased before a new value is written and other factors. So calculating
the lifetime of a drive very much depends on such issues of
"wear-leveling" and "write-amplification" overhead. Unless your writes
volumes are in the > 50GB/day level, I wouldn't worry about it for
*current-technology* SLC-flash drives. However, older technology SSD's
may suffer from much greater write overhead. Stick to the SLC-based
drives and it'll be less of an issue. MLC is probably best for laptops.
Again, I know I'm generalizing... <grin>

To Stephen's original question, they can be *very* good
performance-wise. Most systems will eventually be setup to use SSD's for
speed and HDD's where greater storage capacity is needed.  But the
changeover is just starting, so YMMV with different SSD's.  If you're
going down this path, stick with the latest models with the best write
throughput and SLC.

Roger Johnson

</PRE> 
_____________________________________________________ 
This electronic message and any files transmitted with it contains 
information from iDirect, which may be privileged, proprietary 
and/or confidential. It is intended solely for the use of the individual 
or entity to whom they are addressed. If you are not the original 
recipient or the person responsible for delivering the email to the intended recipient, be advised that you have received this email 
in error, and that any use, dissemination, forwarding, printing, or copying of this email is strictly prohibited. If you received this email 
in error, please delete it and immediately notify the sender. 
_____________________________________________________ 
<PRE>