One of the key advantages of Serial ATA (SATA) and Serial Attached SCSI (SAS) is that they were designed to work together.
By Mark Ferelli
Set the WayBack machine to 1979. Shugart Associates, led by storage industry pioneer Alan Shugart (who later founded what became Seagate Technology), created the Shugart Associates Systems Interface (SASI). This early ancestor of SCSI was rudimentary by comparison with subsequent interfaces. For example, it supported only a limited set of commands and a slow signaling speed of 1.5MBps. However, for its time SASI was a truly disruptive technology, since it was the first step toward an intelligent storage interface for small computers. The limitations of the interface must be considered in light of the technology level at the time: 8-inch floppies were common, and networked storage was unheard of.
In 1981, Shugart Associates teamed up with NCR and convinced the American National Standards Institute (ANSI) to set up a committee to standardize the interface. A number of changes were made to the interface to broaden the command set and improve performance.
The name was also changed to the Small Computer Systems Interface (SCSI), which made the interface less proprietary, according to many people. The actual standard was published in 1986, and evolutionary changes to the interface have been occurring ever since.
It's important to remember that SCSI is, at its heart, a system interface, as the name suggests. First used for hard disk drives, SCSI is still used predominantly for disk disks and is often compared with the IDE or ATA interfaces. For those reasons, SCSI is sometimes thought of as a hard disk interface. However, SCSI was never tied specifically to hard disks.
The interfaces currently enjoying the greatest mind share and market growth are Serial ATA (SATA) and the newer Serial Attached SCSI (SAS). One drive vendor recently reported that its last quarter's shipments included 2 million SCSI drives, 105,000 SAS drives, and a whopping 9.6 million SATA drives.
SATA then and now
Initially, SATA was positioned as the evolutionary successor to low-end parallel ATA (PATA) devices, and significant enhancements were added in SATA 3G (SATA II) devices. These improvements have made SATA a contender for "enterprise" applications.
In a relatively short period of time, SATA has made significant inroads, not only in desktop applications, but also in server, NAS, SAN, and many diverse applications where cost/capacity and performance intersect.
SATA drives are expected to make up 23% of the overall enterprise disk drive market by year-end and are expected to account for about 34% of the market in 2008, according to International Data Corp. (see figure).
Very high performance is a key advantage of serial technologies such as SAS and Fibre Channel. But "good-enough" performance is available in the current generation of SATA devices.
"Historically, only the highest performance and capacity points were based on serial technology [such as Fibre Channel]," says Steve VonderHaar, vice president of sales and marketing at Ario Data Networks. "Today's SAS and SATA drives remove that distinction by bringing serial technology across all tiers of storage. In addition to the serialization, the capacity points of SATA drives are changing the market drastically. In the early days of ATA versus SCSI, SCSI drives were often selected simply due to the capacity points. The playing field has changed as SATA capacity points are exceeding those of Fibre Channel and SAS drives."
It's not just a matter of capacity anymore. SATA is being used in a wide variety of high-performance applications. For example, Barbara Murphy, vice president and general manager at AMCC, notes that the company's SATA controllers are being integrated in a broad array of applications, including video editing, digital content creation (DCC), broadcast, scientific, and industrial applications "where high levels of sequential bandwidth are required at price points where SCSI, SAS, and Fibre Channel cannot compete."
SATA's performance road map started at 1.5GHz in the original specification, and 3GHz (3Gbps) is now commonplace, as are performance enhancements such as native command queuing (NCQ). "SATA was faster than expected, especially for those who were using Fast ATA," says Tom Treadway, CTO at Adaptec, "and SATA is now available on 10,000rpm drives." The next generation of SATA specifies a 6Gbps transfer rate.
It would be tempting to identify SATA as a SCSI replacement, and indeed many data centers are looking forward to adding less-costly arrays to their infrastructure as their capacity demands grow. However, Harry Mason, director of industry marketing at LSI, believes that SATA is growing more into nearline applications and "is not eating into SCSI volumes." (Mason is also the chair of the SCSI Trade Association.)
Mason also is concerned about end-user expectations. "Some customers misuse SATA," he says. "They bring it into some applications because it's cheaper, but if it's used for applications like OLTP [online transaction processing], users can experience higher failure rates and slower speeds."
This may well change, however, with the emerging category of so-called "enterprise SATA." The new category is characterized by improved mean time between failure (MTBF) and reduced rotational vibration, which Mason says have been long-term historical problems with SATA.
SAS follows a different drummer
About five years ago, it became fairly clear that SCSI was reaching its performance limits. As a result, vendors of chips, drives, and subsystems developed what became the SAS interface. SAS uses a serial, point-to-point topology to overcome the performance barriers associated with storage systems based on parallel bus or arbitrated loop architectures. (In a point-to-point configuration, each device gets full bandwidth.)
SAS subsystems can also accommodate both low-cost, high-capacity SATA disk drives as well as high-availability, high-performance SAS drives. This ability to support two classes of drives offers users, systems/storage integrators, and OEMs new levels of flexibility in designing, scaling, and managing disk arrays.
In fact, the key advantage of SAS is choice. Moving to SAS lets users meet enterprise-level performance, capacity, and reliability requirements, while being able to leverage legacy SCSI software and mix-and-match SAS and/or SATA drives.
Pricing may be an issue with SAS, especially considering that SATA and SAS share the same mechanical characteristics with about the same cost of production. But the nature of the I/O is a differentiator. "SAS and SATA II devices will co-exist in storage hierarchies; however, we do not believe they will be deployed to a great extent within the same box because the nature of the I/O to SAS and SATA drives is so divergent," says AMCC's Murphy. "SAS drives are highly optimized for random, small-block transactions, and SATA drives are optimized for sequential, large-block I/O."
"The [relative] growth potential for SATA and SAS will depend very much on how the drive vendors price the drives," says Ario Data's VonderHaar. "SATA is taking off today due to appealing current and projected price-per-gigabyte points, [but] different vendors have different motives, depending on market shares, and watching the ATA/SATA-focused vendors jockey for position with the vendors that have been historically strong in Fibre Channel and SCSI will be interesting."
SAS, not SATA, seems to be the key SCSI replacement technology. "The initial implementations of SAS are focused on replacement technology for the traditional SCSI server market," says AMCC's Murphy. "SAS drives can be used in the same environments in which SCSI and Fibre Channel are used today-in enterprise environments where higher levels of random performance and redundancy are required. SAS as an infrastructure play will enhance SATA II deployment by enabling support for very high-capacity implementations."
Although SAS is expected to account for only 5% of total enterprise-level disk drive shipments this year, it is expected to garner about 25% of the market by the end of 2008, according to IDC. In that same time frame, IDC expects SATA to represent 34% of the market, with 27% going to Fibre Channel and 14% to parallel SCSI disk drives (see figure below).
In addition to the advantages already mentioned, other features of SAS include dual porting, 16,000 devices per domain, "wide" ports, and rate matching.
The vision for SAS and SATA has always been for the interfaces to work together. In late 2003, Maxtor and LSI were the first manufacturers to prove that data can be exchanged between SATA and SAS drives with the aid of SAS controllers and expanders.
"Generally, SAS and SATA are complementary from a systems perspective, but they serve different market segments and have unique benefits," says Joel Warford, vice president of marketing and business development at SiliconStor.
"In enterprise storage array and server applications, SATA drives can be used in either SAS or adapted Fibre Channel system infrastructures as an alternative to SAS or Fibre Channel drives for the best cost-capacity in applications such as disk-based archiving, nearline applications, and many primary storage applications."
An aggressive growth path for both SATA and SAS is expected, since the need for storage capacity is growing dramatically across the board independent of the drive interface choice. SATA is growing faster as a percentage of overall storage capacity due to the attractive cost-capacity point. SATA is positioned to dominate low-end markets (home/SOHO/small to medium-sized business), some medium-range markets (NAS/SAN), and in "nearline" backup/recovery/archiving applications. And as noted earlier, the concept of "enterprise SATA" will challenge the dominance of Fibre Channel and higher-end SCSI drives as this market becomes more cost-sensitive and SATA becomes "good enough" for these applications.
Although the vision was to have SAS and SATA complement one another, SAS proponents may not have foreseen the impact of SATA II, a higher-performance implementation of SATA that provides features that put it in direct competition with SAS in many areas. For example, SATA II supports hot-swapping. This is significant in the development of enterprise SANs, as well as for smaller RAID subsystems and NAS devices. Because less-expensive SATA II drives can be swapped in and out of arrays, users don't have to go with Fibre Channel, SCSI or SAS to hot-swap drives.
Additionally, SATA II specifies a port multiplier feature that enables connections to as many as 15 SATA or SATA II drives (although four or five drives connected to a single port multiplier provides optimal performance). A port multiplier can be connected internally to a server and deliver as much as 6TB of capacity, depending on the number of drives on a single channel. SATA II also includes an external cabling spec that can connect a server and SAN up to two meters apart. And a single cable can carry four channels of data.
"For more and more storage applications, IT budget constraints are driving the need for lower cost storage where 'good-enough' SATA drives compete with 'better' SAS drives," says SiliconStor's Warford. "SATA may prevail in many purchasing decisions. IT managers have to justify the additional costs for SCSI, SAS, or Fibre Channel drives." SAS was developed to promote choice in the storage interface environment. The choices are there, but the selection process is as challenging as ever. Cost, performance, and reliability are still the key criteria, and the victor will depend largely on which of these criteria comes first.
Mark Ferelli is a freelance writer. He can be contacted at firstname.lastname@example.org.
This article is excerpted from the July 2006 issue of InfoStor. To view the entire article, go to: Making the transition to SATA, SAS.