xSATA and SATA 6G news

[LiamC]

Charlie "Bunny" over at Le Inq has news on xSATA and SATA 6G

http://www.theinquirer.net/default.aspx?article=34712


xSATA has an 8m (26') external cable while 6G is/has a transfer rate of 6 Gigabits/s

 

[P5-133XL]

Few will end up dealing with the XSATA multiplexers and we really don't need 6mbps when drives are still constrained to a maximum of 100MBps.

What I want is for SAS to replace SATA totally. SATA is simply a subset of SAS and that means SAS controllers can operate either SATA or SAS drives eliminating the the need for seperate SAS controllers. If all the MB's put in SAS controllers rather than SATA controllers, the computer world would be in a better place.

 

[Pradeep]

Probably going for 6 Gbit/sec to compete with 4 Gbit/sec Fibre Channel.

 

[Explorer]

Probably going for 6 Gbit/sec to compete with 4 Gbit/sec Fibre Channel.

ummm... not really.


iSCSI (IP-SAN) is by far Fibre-Channel's biggest competitor. IP-SAN technology has been doing an increasingly good job of grabbing new entry-level SAN deployments.

As we all know, the SCSI and ATA/IDE camps were enemies for years. Then, ATA/IDE began transforming itself into SATA. Shortly thereafter, SCSI and SATA combined forces to stay relevant as a whole -- with SCSI becoming SAS, and SAS taking on compatibility with SATA to help SATA get into the datacenter.


Then, not much later, SAS combined forces with Ethernet to take on SCSI's old arch-enemy head on -- Fibre Channel -- with iSCSI technology to build inexpensive IP-SANs.


At the high end of the storage universe, InfiniBand has been slowly eating away at Fibre-Channel on the bleeding upper edge where supercomputers and large datacenters want a fabric with massive throughput.


Just recently, Fibre-Channel's worst nightmare came true: InfiniBand technology will now incorporate iSCSI technology to help simply InfiniBand-based SANs, and iSCSI will gain InfiniBand's hardware's blistering throughput and advanced RDMA (Remote DMA) data transfer methods. The result is iSER (iSCSI RDMA) technology.


So, you will soon be able to build iSER fabrics, where, for example, you have ultra-fast InfiniBand core switches and inter-switch links connected to less expensive Gb Ethernet interfaces and switches for an edge fabric. The whole fabric will talk iSER, even though the edge is Ethernet and the core InfiniBand. You could also build an iSER fabric without any InfiniBand hardware or an iSER fabric without any Ethernet (iSCSI) hardware.



Press release in PDF format:

http://www.infinibandta.org/newsroom/IBTA_iSER_press_release_FINAL.pdf

 

[sechs]

What I want is for SAS to replace SATA totally. SATA is simply a subset of SAS and that means SAS controllers can operate either SATA or SAS drives eliminating the the need for seperate SAS controllers. If all the MB's put in SAS controllers rather than SATA controllers, the computer world would be in a better place.

For those of us who know them both and see SATA has SAS with separate data and power connectors, this makes a lot of sense. For those folks who see SCSI as some strange ivory tower interconnect, it's like putting oranges in the apple pie.

SATA has been, is, and will be, for the forseeable future, cheaper to implement than SAS. For the desktop, this is important; and, as long as it stays in the box, I see no real reason to bother. The vast majority of people will never know the difference.

On the other hand, external SATA implementations all appear to be half-baked versions of what SAS does. This is surprising in that it's largely the same folks who are determining both standards. And it really seems like they're wasting a lot of money on these.

 

[Pradeep]

I noticed that the most recent SATA spec includes daisychaining e.SATA external boxes. This then appeared as a seamless increase in capacity to the O.S.

Perfect for daisychaining external SATA hard drives off the HD DVR.

 

[Handruin]

Why is SAS any more expensive than SATA? It doesn't seem to be anything wondefully exceptional. It seems like the price was set high from the start, so SAS becomes this enterprise solution when it could have simply been the one and only solution. Make all hard drives move to SAS and stop fussing with SATA. Once SAS is spec'ed out, how much harder is it to make SAS controller chips compared to SATA? It seems like some silly way to profit from enterprise level systems because big corporations have more money?

 

[Pradeep]

Don't you work for EMC? The money people pay for "enterprise" level storage is always staggering to me.

 

[Handruin]

Yes,...and it amazes me too.

 

[Buck]

I was just looking and SIIG has some nice PCI-E 1x slot cards that support a variety of features and interfaces, including eSATA, SATAII, SATAII RAID, and 1394b. That certainly is a nice way to use up that lonely slot on a new motherboard.

 

[sechs]

Once SAS is spec'ed out, how much harder is it to make SAS controller chips compared to SATA?

You'll have to ask LSI about that one, but SAS is more complicated to implement, especially since it carries SATA as a subset.

Keep in mind the kind of incredibly crappy silicon that has come out for SATA versus what comes out for SAS.

 

[Splash]

Why is SAS any more expensive than SATA? It doesn't seem to be anything wondefully exceptional... Once SAS is spec'ed out, how much harder is it to make SAS controller chips compared to SATA?...

Low volume enterprise product always costs more. Early adopter product always costs more -- though I believe the "early adopter" period for SAS technologies is finally receding rapidly into the past. However, there is more to the premium pricing for SAS than relatively poor supply/demand ratio or whatever price the enterprise/technical marketplaces will bear.

Unfortunately, there is a significant amount of cost in licensing for SCSI and SAS. Trade organisations like the SCSITA (www.scsita.org) help reduce or eliminate these costs by encouraging cross-licensing or putting patents into a patent pool. But, not all intellectual property licensing falls into this convenient realm.

Second, the SAS is definitely more sophisticated and complex than SATA. SAS requires deep buffers that can operate in full-duplex mode along with the significantly more complex SAS microcontroller circuitry (and commmand set) in all SAS devices. SAS hardware undergoes longer, tougher, more-expensive testing for the enterprise/techical markets than SATA does for the home / office / gamer markets (there is also a burgeoning "enterprise" SATA market).