Last post Nov 21, 2009 04:49 PM by ccondo
Jul 08, 2008 10:03 PM|BladesAway|LINK
We are anticipating installing Exchange 2007 in the Hosted Environment. We expect to support about 3000 mailboxes per server. We are looking for advice on SAS vs SATA vs SAS Near-Line storage. Our question is .. can SATA or SAS Near Line be used in such
a mission-critical, always on environment or is conventional SAS really the only way we should go?
Jul 09, 2008 08:47 PM|paulsilverstein|LINK
We have no problem running SATA
Jul 24, 2008 06:17 AM|RevCrispy|LINK
There's no definitive answer really, as it alldepends on budgets and service levels. I normally though stick with the following;
If application requires a cheap large storage and isn't going to need to be performant (is that a word?) then SATA may be ideal, good for file & print / archiving solutions.
If you have things like Exchange DB's and SQL DB's then SAS is generally a better option as it offers better performance and reliability compared to SATA.
Near-line are you meaning DAS (direct attached storage)? if so again it depends on the hardware you use e.g. decent RAID controllers (with cache) and what RAID configuration you plan to use. As an example a RAID 5 with a single hot-spare you may decide on
SATA as you could deal with two disk failures before data loss. But this may turn up performance issues compare to SAS under heavier loads.
You may decide on a combination as well, use SAS for the mail stores and SATA disks for recovery partions and maybe backups before storing on tape.
We have a fair amount of SAN infrastructure so it's not too much more for HBA's and to utilise the benefits of our existing infrastructure.
I know it's not a strict answer, but storage strategy never is black and white, but I hope it helps.
Jul 30, 2008 04:02 AM|filippg|LINK
under http://msexchangeteam.com/archive/2008/07/28/449379.aspx there's a comparision of a SATA and a USCSI-System. It's for the Hub-Server, so it's not exactly what you're looking for. But maybe it's interesting.
Nov 14, 2009 09:39 AM|Kiran Ramesh|LINK
Hi In partial answer to the original question, regarding the difference between SAS and SATA drives: I have found it very difficult to get any definitive information on the differences, especially since manufacturer information now tends to be less definitive
in their documentation, and different manufacturers give different statistics, making comparison difficult. In the past, SAS (or SCSI or FC) drives used to be for main data use, and SATA (or ATA or FATA) for auxiliary or backup use. This came down to three
factors: the interfaces in SCSI drives was more efficient, the seek times and rotational speeds for SCSI drives tended to be better, and the reliability of the mechanism for SCSI drives was better. The last of these was typically documented in terms of the
duty cycle – how much of the time was the drive expected to be active – some were as low as 25%. Thus, both a SCSI and an ATA may have a similar MTBF, but if the ATA was used 100% of the time, then its expected MTBF would be a lot lot shorter. At that time,
there used to be a large difference in price between SCSI and ATA for similar sized disks. Unfortunately, it looks like some of the price differential has also reduced. Nowadays, it seems that the mechanisms for ATA and SCSI seem to now be more similar, and
assuming they have been specified for 100% duty cycle, the MTBFs can be similar, so the reliability concern is less of an issue. However, you still have to be careful, since SATA drives are more likely to have a lower MTBF (or average failures per annum as
may nowadays be documented). For example, compare the ENT and MDL HP drives. As an example of overall performance, several years ago, I did some testing on drive arrays for backup, (using an EMC Clariion array) with a disk array of either FC or FATA drives.
What I found was that, if I directed a single backup stream to them, they both performed about the same. However, as I increased the number of backup streams, the total throughput on the FC disks was more than double after four streams, but the total throughput
on the FATA drives hardly changed at all, showing that the FC disks performed much better than the FATA disks, presumably because they handled [semi] random writes better. It would be interesting to know if a similar test today would give similar results.
Nov 21, 2009 04:49 PM|ccondo|LINK
I work at
www.finalasp.com and we do Exchange hosting for thousands of users. We currently use an EMC SAN with fibre channel drives. We invested in a shelf of SATA storage thinking the cache in the SAN would make it usable for information stores. Exchange smoked
the storage and the latency was terrible. Stick with SAS, SCSI, or fibre channel.
exchange 2007 hosting