Conclusion
SAS 6 Gb/s represents a logical and desireable evolution of the SAS architecture. It simplifies management of complex storage deployments, while speeding up individual links from 3 Gb/s to 6 Gb/s.
Performance will probably be the most important reason for people to adopt SAS 2.0, although we have to underline that the second generation really has matured. Zoning and external cabling have now been standardized, removing the complexities and confusion tied to proprietary solutions. SAS 6 Gb/s is now a true competitor to Fibre Channel solutions, even in advanced, mid-range enterprise scenarios.
Where 6 Gb/s Doesn’t Make a Difference
Performance advantages make great headlines, but be aware that this generation of SAS only delivers a benefit in storage environments based on expanders and appliances, where you have a plethora of drives being accessed through relatively narrow data pathways. Even the fastest new SAS 6 Gb/s drive, Seagate’s Cheetah 15K.7, is still limited to 200 MB/s maximum throughput when operated in one-drive configurations.
This can already be handled by SAS 1.1 at 3 Gb/s (300 MB/s). We’ll have to wait at least two more hard drive generations to see an individual drive saturate 3 Gb/s bandwidth. So, there’s not much benefit in focusing on getting 6 Gb/s SAS into systems running only small RAID arrays or individual hard drives. There simply won’t be a noticeable benefit.
The True 6 Gb/s SAS Value
However, things change rapidly once wide ports are used to talk to SAS storage appliances. Without wide ports, the aggregate bandwidth of so many drives can swamp a conventional SAS link. We used 16 flash SSDs to try to taxing LSI’s new 9200-series bandwidth, but imagine what would happen if 20, 30, or even 50 drives were attached through expanders and appropriate storage applications. The SAS link suddenly becomes the bottleneck.
SAS 6 Gb/s = Affordable Backbone for Virtualization
Also consider the SAN scenario. In environments where storage works independently from specific servers (through SAS or iSCSI), the downlink to your storage is key. Enterpises utilize heavy storage appliances to power virtualized system environments so that active operating system partitions can quickly switch from one server to another. iSCSI over Gigabit Ethernet remains the bottleneck, but once SANs are taken over by 10 GbE (or in Fibre Channel environments), SAS 2.0 will present a great opportunity to effectively double storage performance without much additional cost.
Decision makers in the process of finding new solutions should take the time to dig into SAS 6 Gb/s. It may be worth the investment compared to SATA hardware or a workable alternative for expensive Fibre Channel hardware.
Latest Storage Solutions News
Latest Storage Solutions reviews
- 03/09 – Open-E's DSS V6: Storage Software Set Up, Managed, And...
- 01/27 – What Does One Petabyte Of Storage (And $500K) Look Like?
- 10/20 – Does The USB 3.0 Controller On Your Motherboard Matter?
- 08/26 – Not All USB 3.0 Implementations Are Created Equal
- 04/23 – Memory Cards, Part 2: SDHC Cards From 8GB To 16GB
-
Latest Seagate News
- Seagate to Acquire Controlling Interest in LaCie
- Analysts Predict Acceleration in HDD Storage Density Growth
- Seagate Announces Demo of 12 Gb/s Solid-State Drives
- WD, Seagate Recover From HDD Shipment Shortage
- Seagate Reaches Terabit Areal Density Milestone
-
Latest Seagate reviews
- Hitachi's 4 TB Hard Drives Take On The 3 TB Competition
- System Builder Marathon Bonus: Newegg Customer Choice PC
- System Builder Marathon, March 2012: System Value Compared
- System Builder Marathon, March 2012: $2600 Performance PC
- System Builder Marathon, March 2012: $650 Gaming PC
Now I like this, can't wait to see it implemented on EMC & IBM platforms. And on the servers side I'd like to see some numbers on spec's and performance. Good article Patrick Schmid and Achim Roos! Next.... benchmark 2.5 vs 3.5 on enterprise systems and low budget small business ....yes???
SCSI is still not dead?!?!?
Joking aside, I wonder if any motherboard manufacture like ASUS will get this on their high end WS models (see P6T WS). And I wonder if they will make any SSDs using this for the server sector.
[citation][nom]Shadow703793[/nom]SCSI is still not dead?!?!? citation]
no its not -- its a long death cause most scsi drives run forever. I have a bunch of scsi drives running on my systems at home and as much as Id like the performance boost of SAS or even Sata, the things just wont die.
but 6g/b SAS looks to be the next upgrade step
6Gb for SAS is kind of pointless considering SAS drives only peak 100MB
On this page (http://www.tomshardware.com/reviews/sas-6gb-hdd,2392-8.html), I see that you stated "SAS 1.1 at 3 Gb/s (300 MB/s)"
Well this has always confused me. If SAS 1.1 is 3 Gb/s, that would be 375 MB/sec (3 x 1000 to convert to 3000 Mb/sec and then 3000 \ 8 to convert to 375 MB/sec.)
On the SAS Wiki site (http://en.wikipedia.org/wiki/Serial_Attached_SCSI) it shows 3 Gb/sec. However on the list of device speeds Wiki site (http://en.wikipedia.org/wiki/List_of_device_bandwidths) it shows 2.4 Gb/sec with a 300MB/sec. The only maximum that appears to stay the same is MB, which is 300. So is it really 2.4 Gb/sec or is it 3.0 Gb/sec really for SAS 1.1?
Thanks!
@kittle -- Not to mention that SCSI is a protocol, not just a physical/electrical interface. SAS simply changes the physical/electrical interface, but the good old SCSI protocol is still in there (just like SATA still has ATA underneath).
@falchard -- No, it's not pointless. A dedicated controller channel for every drive would be prohibitively expensive for large drive arrays (not to mention the cabling nightmare), which is why SAS (unlike SATA) allows for more complex topologies, not just point-to-point. When you put multiple drives/expanders on a channel, you can quickly hit SAS bus speed limits.
@rhodenator -- SAS (like SATA) uses 8b/10b encoding. That is, 8 data bits end up as 10 bits on the wire (typical of high speed serial buses), so: 3.0Gbs wire * 8/10 encoding = 2.4Gbs data = 300MBs data.
I bet the hardcore gamers are drooling at the prospect of SAS 6Gb/sec. Too bad they'll have to wait.
Actually, LSI was not 1st to market w/ 6g controllers. ATTO Tech was shipping their H6xx 6G HBAs months ago based on the PMC Sierra chip. Also HP has been shipping their 6G RAID controllers - P212 & P410 (also PMC Sierra chip) for a couple of months also. Granted LSI is the industry leader w/ SAS but has stumbled in execution of releasing 6G product.
Actually, LSI was not 1st to market w/ 6G SAS controllers. That distinction goes to ATTO Tech with their H6xx 6G HBA line up and HP for their 6G RAID controllers - P212 & P410. All of these are based on the PMC Sierra chipset. LSI has been stumbling to get 6G products out the door.
re: rhodenator. As mentioned in the article; 8b/10b encoding means that 8bits data is encoded using 10bits accross the link. 3000Mb/10b = 300MB. Command and control overhead and processing delay further limits real world performance to about 270MB/s. I think someone managed over 580MB/s using the new 6Gb/s signalling and more data per transfer to reduce overhead (was that using SATA or SAS ?).
re: falchard "..kind of pointless.." NO.
"Build it and they will come". You have to start building the infrastructure and plan for the future otherwise it's pointless making the drives go faster. I want a server that in a few years time can take additional drives or replacements where performance boost is not nullified by slow IO bandwidth.
It will take 12months after the standard's release to be more common as HBA, motherboard and drive designs are refreshed.
SATA is no longer a 1-to-1 link and burst bandwidth is close to current limit.
600MB/s means enough bandwidth to aggregate 4 of today's SAS HD per SAS port or maybe 2 SSD's. Within 2 years a single HD will exceed current limit and within 1 year SSD's will reach the new limit.