Performance differences between 2.5" vs. 3.5" SAS
What real performance gains/losses are there on 2.5" vs. 3.5" 15k SAS drives. Are folks really seeing seek differences, &c?
HP's published drive specs on their SAS drives seems to indicate there is a difference in the seek times on the drives:
HP 146GB 3G SAS 15K 3.5" DP ENT HDD
Transfer Rate Synchronous (Maximum) 3 Gb/sec
Seek Time (typical reads, including settling)
Single Track 0.57 ms
Average 3.5 ms
Full-Stroke 9.0 ms
HP 146GB 3G SAS15K SFF (2.5") DP ENT HDD
Transfer Rate Synchronous (Maximum) 3 Gb/sec
Seek Time (typical reads, including settling)
Single Track 0.14 ms
Average 2.58 ms
Full-Stroke 4.85 ms
So 2.5" seems quite a bit quicker on the seek times. You also wanna be mindful that the new drives come with the 6G SAS interfaces which are more rapid then the 3G interfaces, but the disk controller has to support it (such as the P410's you get in HP DL G6 servers after a firmware upgrade).
2.5" hard drives have a much smaller platter diameter and thus there is a smaller mass that must be moved while friction is reduced as well. On the other hand, the operating range of the slider (which carries the heads) is smaller, which has a positive impact on access time if you compare 3.5" and 2.5" at equal rotation speeds and mechanical components. A 2.5" server hard drive delivers somewhat slower transfer rates than 3.5" drives do, but they beat them in access time.
A 2.5" SAS drive currently reaches capacities of 147 GB at 10,000 RPM rotation speeds, or 73 GB if you want a 15,000-RPM drive. At the same time, 3.5" SAS and SCSI hard drives have been available in capacities of 300 GB for a while. If you come across other "enterprise-class" 3.5" drives you can be sure that these have their origin in the desktop space, but pass different validation and qualification processes to ensure the requirements of 24/7 operation are met. These drives only come with a SATA interface, but none of them supports SAS.
However, 2.5" SAS drives boost so-called storage density, because their required space is much lower when compared to 3.5" drives. This definition of storage density refers to storage capacity per volume, but it is also possible to define storage density as storage performance per volume. Popular storage appliances with only a single height unit (1U) either accept four 3.5" hard drives or as many as 10 2.5" drives - a good example is the Infostation by StorCase. Up to 10 do not require substantially more energy than four 3.5" drives, but they will outperform the four 3.5" drives by two times or sometimes even 2.5 times. If you decide to use 4-6 2.5" drives, you will receive at least equivalent performance at clearly reduced power consumption.[1]
[1] http://www.tomshardware.com/reviews/sas-hard-drives,1702-2.html
To add a little bit to Stephen Thompson's answer which was good.
Currently the largest SAS 2.5" drives are 146 GB for 15k, and 300 GB for 10k.
One reason why the SAS 3.5" disks have relatively poor density, is that the platters themselves are far from filling the case, closer to 2.5" size actually. This is AFAIK due to problems with larger platters wobbling slightly at high rpm.
Many server applications are IOPS limited rather than storage space limited, so the lower capacity of the 2.5" drives are more than compensated for by the fact that you get a lot more IOPS per U of rack space.
For drives that are intended to be performance based, I have yet to see a major difference in actual form-factor.
However, I would question the reliability of the 2.5" drives from experience. 3.5" drives are still more common -- and therefore more understood.