Author: Samuel J. Barnett
Product Line Manager, Serial Attached SCSI Products
Mention Serial Attached SCSI (SAS) technology to a room full of IT professionals and the air is immediately charged with excitement. Take the dialogue one step further and associate SAS with Fibre Channel front-ends/fabrics/switching or a SAN and the mood turns noticeably somber.
While Serial Attached SCSI offers a wealth of benefits to the enterprise server and enclosure customer (high reliability and performance, mixed enterprise/desktop drive support, and improved economies of scale), the thought of yet another SAN technology initially concerns the majority of the end-customer community. Legacy deployments of 1, 2, and now 4Gbps Fibre Channel in conjunction with costly director-class switches make a wholesale technology shift difficult at best, while other promising technologies like iSCSI further divide the end-user community.
A recent IT end-user survey published by Ashton, Metzler & Associates outlines the overwhelming shift to SAN-based consolidated (following a CORE-EDGE architecture) storage facilities from “SAN islands” or server-centric direct attached storage (DAS) environments. It is reasonable to assume that this increasing rate of “SAN adoption” could open the door to other technologies easily adapted to deliver a given level of service at a given price point.
In fact, when asked about new storage technology adoption criteria, end-users tended to agree that reliability, availability, interoperability, and scalability would top the list with management ease, performance, and total cost of ownership filling in the critical ownership criteria matrix. Oddly enough, initial purchase price was not a factor in a new technology deployment decision. Below, we examine each adoption criteria outlined by the end-user community in more detail.
Reliability and Availability
Serial Attached SCSI is an enterprise-class storage technology. End-user reliability and availability concerns are addressed through features that include dual-port drives, enterprise drive electronics/mechanics (not desktop electronics/mechanics), support for redundant connection/fail-over schemes (using either SAS or SATA targets), and hot plug-ability.
IT end-user interoperability concerns are being addressed head-on by the SCSI Trade Association and its member companies. Numerous public and private interoperability/test events (called plugfests), and direct consortium support through the University of New Hampshire InterOperability Laboratory, guarantee SAS device interoperability and ecosystem stability. With multi-vendor/OEM support, SAS is on-track to be the most tested storage technology in history.
SAS was designed to maximize the ease with which drives (or whole subsystems) can be added to boost both capacity and throughput. With the use of “expander” technology, aggregation of up to 16,384 devices can be achieved in a single SAS domain, making SAS second only to Fibre Channel in single-domain theoretical scalability limits.
SCSI Command Descriptor Block (CDB) reuse/support in SAS guarantees that management facilities developed for parallel SCSI systems will continue to be effective. New management applications and plug-ins for existing management software will evolve to support new features as SAS technology deploys.
Point-to-point, full-duplex, connection-oriented transfers at speeds of 3Gbps (1.5Gbps for SATA targets based on the SAS 1.0 standard), combined with low device arbitration and transit latency, make SAS a superior performer to Serial ATA and 2Gbps Fibre Channel. Through the use of SAS “wide-ports,” effective aggregate bandwidth between devices grows in an additive fashion, allowing system OEMs to tailor systems to particular application needs.
Total Cost of Ownership
As the evolutionary replacement of parallel SCSI technology, no significant changes are anticipated in the total cost of ownership (TCO) of SAS systems when compared to their legacy counterparts. SAS as a fabric interconnect solution would most likely alter these dynamics.
Since SAS was conceived as a DAS interconnect technology, it does not immediately fit the bill as a “fabric technology,” however, with an improved credit schema, connectionless transfer mechanism, routing summarization, and an optical interconnect, SAS could evolve into something much different than it is today.
Summary and Conclusion
Ultimately, the future of storage and storage networking is contingent upon the evolution of SAN and NAS architectures, the distribution model for storage, and advances in transparent protocol communication techniques. As with any new technology development, whether revolutionary or evolutionary, one size will never fit all. There will be complementary technologies that address different market segments, and the “proper solution” will differ by application, connectivity requirements, scalability, performance, and price sensitivity. Hold on for the ride!