The difference is parallel and serial connections. SCSI is the oldest hard drive technology and uses parallel interface, pushing out about 80 MB Bytes per second. ATA (also known as IDE) is also parallel and pumps about 100 MB Bytes per second.
SAS and SATA are the newer technologies, utilizing serial insterfaces. SAS can pump out a whopping 3 GIG Bytes per second. SATA will push out 150 MB Bytes per second.
Hope this helps!
|||Mike Bowers - Marketing Director |||atOmicVPS LTD ||| OnApp Powered Linux & Windows Cloud Hosting ► [Shared] ► [Reseller] ► [VPS] ||| Follow the atOmicVPS Blog
If you are more concerned with performance, stability and ROI, go with SAS (Serial Addressed SCSI). If on the other hand you don't have the budget and want to save money go with SATA. Much as others have said, SAS far outperforms SATA to the point it isn't even worth comparing. SAS drives are available with higher spindle speeds and tend to last quite a bit longer. Sure, they cost 10x more but if performance matters, they are worth it.
ZZ Servers - Business Hosting, HIPAA and PCI Compliant Hosting Solutions - http://www.zzservers.com
Xen Virtual Private Servers | Dedicated Servers | Shared Hosting
Custom configurations, firewall, VPN, load balancers, private networks and more.
As has already been said, if you can fit most of your content in RAM then it won't matter that much how fast your disks are for videos as the content rarely changes (unlike a database) so you can cache it all.
Remember it is not just about the disks you use but also the configuration - the RAID setup you use can make a big difference as well.
There are two main disk IO metrics - raw throughput and IOPS - usually people run out of the second one way before they run out of the first one. In both cases a 15k SAS disk will give you about double the performance of a 7.2k SATA disk.
Comparing spindle speeds isn't really a valid "SATA vs. SAS" comparison. I mean, there are 10k RPM SATA drives, or compare a SATA Intel x-25e, which will destroy any 15k RPM SAS drive in every single benchmark...
Of course, the x-25e is obscenely expensive. But my point is that you guys are doing a SATA vs. SAS comparison, and looking at a very small subset of SAS and SATA drives to justify the differences. Which is completely invalid.
SQLIO random write test
Intel x25-e: 94MB/s
Seagate Cheetah (15k SAS): 5MB/s
The Intel drive is 19x faster on random write IO speeds... More than an order of magnitude. Random reads and writes are one of the places where a good SSD will absolutely destroy any spindle-based disks.
Note the qualifier "good". An SLC SSD from virtually any company will produce excellent random write speeds. A *good* MLC SSD (such as Intel's x25-m) will produce much slower speeds, but still enough to wallop any spindle-based disk. The crappier MLC SSDs (primarily those with JMicron controllers, but not exclusively) are dramatically slower than spindle-based disks.
It seems to me like you're just (way) behind the curve on the state of the SSD market. The slow random write performance that you're probably referencing was an issue with JMicron controllers. Intel, obviously, never used JMicron controllers in their products.
The x25-m got a *lot* of random write benchmarks done of late, and from memory, which could be wrong, it did something like 20MB/s on random write IO. But that wasn't even the same benchmark, so it's not really a valid comparison.
As you can see from the benchmark, 1 X25-E is about 2.5 times faster than 8 SAS 15,000rpm disks in RAID 0.
That means 2 X25-E in RAID 1 would be about 2.5 times faster than a 16 disks SAS 15krpm RAID 10.
Yeah I'd call that "destroyed".
Don't mistaken X25-E for those cheapo MLC SSD you see everywhere. X25-E are SLC SSD specifically made for OLTP workload and is extremely fast on random writes.
Yeah, and as I said, even the higher-end MLC drives (such as the x25-m, which is far cheaper than the x25-e) will still do a good job in the same scenario. Not nearly as good, but still enough to put any SAS drive out to pasture.