SAS delivers maximum SATA scalability.
But there is another, less immediately obvious benefit to SAS/SATA compatibility: it boosts SATA scalability far beyond the limits imposed by SATA-based infrastructures. Deploying hundreds, even thousands, of SATA drives in a single SAS domain is a straightforward affair, requiring only standard SAS host bus adapters (HBAs) and expanders. Simply put, an enterprise-class SAS infrastructure enables SATA drives to transcend the scalability constraints inherent in their desktop DNA.
SATA: Rooted in the Desktop
The authors of the SATA 1.0 specification wisely focused on those capabilities applicable to desktop environments--incorporating additional features of marginal relevance would only serve to needlessly drive up costs. Scalability, of course, is not a primary consideration among desktop users, who seldom have reason to install more than one additional drive. As such, SATA's limited scalability should not be viewed as an architectural flaw, but rather a logical byproduct of diligent efforts to maximize SATA's cost-effectiveness.
[FIGURE 1 OMITTED]
But as enterprises increasingly turn to SATA disc drives for nearline, backup/restore and other low-availability storage duties, the need for greater scalability has steadily grown. The new SATA II specification addresses this rising demand with its new Port Multipliers (PMs). While SATA 1.0 allows only one drive per host controller port (requiring additional ports to accommodate extra drives), SATA II's PM functions like a hub, enabling each PM-equipped port on the host controller to connect up to 15 drives.
Limitations of Port Multipliers
Port Multipliers clearly enhance SATA's scalability in workstation and high-performance PC environments. Video editing, audio editing, graphic design and video games are just a few of the applications that can benefit from the improved performance and capacity of PC RAID's utilization of PMs. Such RAID configurations can also significantly boost the effectiveness of low-end servers. However, PMs fall short of meeting the enterprise's needs in a data center environment. Specifically, SATA infrastructure (including its PMs) presents the following challenges:
Compatibility: PMs require SATA II host controllers that are PM-aware; legacy SATA 1.0 controllers will need to be replaced.
[FIGURE 2 OMITTED]
Expandability: Unlike conventional hubs, PMs cannot be linked together, severely limiting their flexibility and expandability in a network environment. PMs can quickly become prohibitively complex as drive quantities reach enterprise-class levels. For example, based on the theoretical limit of 15 drives per PM, deployment of 100 SATA drives would require seven PMs. Connecting 100 drives via the more common eight-port PM configuration would require 13 PMs.
Performance: Switches are the de facto enterprise standard because they enable simultaneous communication between multiple initiators and targets. But PMs support only one active host connection, significantly slowing effective throughput. Furthermore, each time communication is initiated with a drive, a time-consuming drive reset must occur.
Data Integrity: Enterprise storage infrastructures typically employ a multitude of disc drives; addressing protects data integrity by assigning a unique address to each drive, thus ensuring data is consistently directed to the correct drive. Desktop environments (entailing minimal drive counts) don't require such addressing capabilities, thus SATA does not support them. Furthermore, PMs don't allow persistent drive connections as the host may only address one drive at a time. The PM dynamically closes the connection to one drive and opens a new connection to another. With each closed connection, drive history (e.g., data source, destination drive, data and command context) is lost; thus with each opened connection the chance of misidentification and sending data to the wrong drive is increased.
Reliability: Port Multipliers offer only passive failover capability vs. the added security of active failover. Should a PM's primary host controller fail, any backup controller must be manually configured to restore PM function. Active-failover devices employ dual ports to ensure uninterrupted service; should one controller fail, the device automatically switches ports to access the remaining controller.
Cabling: Lack of flexibility in PM/drive deployment is further exacerbated by the short cable lengths (one meter) permitted in the SATA specification.
These issues preclude PMs from efficiently meeting the enterprise's demand for greater SATA drive scalability. But there is an elegant alternative, one that delivers both extensive scalability and unmatched flexibility. Not surprisingly, it comes from an interface purposely built for the rigors of enterprise use (high availability, 24 X 7 duty cycles): Serial Attached SCSI.
SAS: Superset of SATA
Leveraging their common serial, point-to-point architecture, SAS encompasses all of SATA's virtues and then surpasses them with a comprehensive range of enterprise-class capabilities far beyond those of its desktop-centric sibling. Specifically designed as a superset of SATA, SAS is able to synergistically interoperate with SATA, significantly enhancing the value of both technologies.
SAS, of course, is optimized for online, high-availability applications in the most demanding enterprise environments. To that end, it incorporates an impressive array of strengths (full-duplex, dual-port operation for maximum transfer rates and failover capability, rock-solid reliability, rich and mature SCSI command set, advanced command queuing, sophisticated verification/error correction) to deliver the throughput and dependability mission-critical environments demand.
But the very strengths that make SAS drives an ideal performance solution renders them a relatively over-engineered (and costly) choice for nearline, bulk storage chores. Conversely, SATA drives (optimized for enterprise nearline applications) are ideally suited to such duties, where maximum capacity per dollar supersedes other factors.
The Serial Attached SCSI standards committee well understood the complementary nature of these two storage technologies, and the synergies (both fiscal and physical) that would result if SAS and SATA drives could share a common storage infrastructure. To effectively serve both interfaces, such an infrastructure must seamlessly blend scalability, flexibility and affordability. SAS infrastructures satisfy all of these criteria via devices known as expanders.
Expanders: The Key to SATA Scalability
SAS expanders are high-speed switches that enable a single SAS domain to contain over 16,000 drives (SAS and/or SATA). There are two types of expanders: edge expanders, capable of connecting up to 128 drives; and fan-out expanders, one of which can aggregate up to 128 edge expanders in a single SAS domain.
Unlike PMs, SAS edge expanders can be linked--either to a fan-out expander (which employs table routing) or to another edge expander (which employs subtractive routing, and optionally, table routing). Increasingly, SAS vendors are recognizing the value of incorporating table routing into their edge expanders, thus eliminating the need for a fan-out expander and allowing multiple edge expanders to be cascaded together in daisy-chain fashion.
In addition to expanders, there is another key component of SAS/SATA scalability--the SATA Tunneling Protocol (STP). STP enables SAS HBAs to identify and communicate with SATA devices. When data is directed to a SATA drive that's connected to a SAS backplane with an edge expander, an STP connection is immediately opened to enable SATA frames to pass through the connection to the drive. STP operates transparently in the background, with virtually no impact on system throughput.
Expanders = Efficient Scalability
SAS infrastructure addresses many of the SATA infrastructure challenges previously discussed, ensuring optimal scalability for both SATA and SAS drives:
Compatibility: Offering an unprecedented degree of compatibility and efficiency, SAS HBAs and expanders enable deployment of both high-performance (SAS) drives and high-capacity (SATA) drives in the same infrastructure. By eliminating the need for separate and redundant infrastructures, SAS reduces both hardware and IT management costs.
Expandability: Each SAS edge expander is capable of connecting up to 128 devices (SAS HBAs, SAS and/or SATA drives and other SAS expanders). Thus, a single SAS HBA port connected to a SAS edge expander is theoretically capable of addressing over eight times as many SATA drives as a single SATA host controller port connected to a 15-port PM. Should additional drive ports be necessary, another edge expander can simply be cascaded off the first expander, maximizing the value of each HBA port.
SAS delivers exceptional expandability and flexibility by enabling direct cascading of multiple edge expanders (each expander connecting up to 128 devices). For still greater scalability, a single fan-out expander can aggregate up to 128 edge expanders, yielding a theoretical total of over 16,000 devices in a single SAS domain (Figure 1).
The dual-port architecture of SAS also facilitates expandability with the improved performance of bandwidth aggregation. Constructing wide links from dual ports enables transmissions from several drives to be aggregated over a single large pipe to the host, eliminating the performance bottleneck that can arise when too many drives are connected to a single host port (Figure 2).
Performance: SAS edge expanders and fan-out expanders are true enterprise-class, high-speed switches that enable simultaneous communication between multiple initiators and targets. Expanders also benefit from SAS's full-duplex architecture, allowing a drive to transmit data to the host while the host is sending additional commands to the drive. Separate interface sessions aren't needed to send commands, freeing more bandwidth for data transmission.
Data Integrity: SAS expander route tables contain addresses for all attached SAS and SATA drives, ensuring they can be readily located and sent data regardless of their location in the SAS domain. This greatly reduces the risk of losing data by sending it to the wrong drive, a key consideration as drive counts continue to climb. SAS further ensures accurate data transmission with more comprehensive header information, which includes the source, destination and context of each command associated with the actual data transmitted. This effectively "logs" host/drive activity in a SAS domain, significantly improving connection consistency and thus data integrity. For SATA drives in SAS infrastructures, this enhanced header information passes between SAS HBAs and expanders. For SAS drives this header information travels directly between HBAs and the drives themselves.
Reliability: SAS's dual-port architecture also enables seamless, active failover capability to ensure reliability under intense enterprise traffic. Dual ports allow SAS devices to be connected to multiple hosts; should one controller fail, the SAS device will automatically switch to another available controller.
Cabling: Maximum cable length per discrete connection between two SAS devices is eight meters (total SAS domain cabling distance can run into thousands of feet). This offers exceptional flexibility to locate servers and storage arrays in the most cost-effective, space-efficient configurations possible.
While initial expenditure on SAS infrastructure will surpass that of a comparable SATA deployment (preliminary industry estimates put SAS HBAs and RAID backplanes at rough price parity with equivalent parallel SCSI pieces), the true cost-effectiveness of any storage infrastructure goes well beyond its initial expense. What is the cost in manpower and downtime/lost productivity when a drive fails and a RAID volume must be rebuilt? How many additional drives must be purchased to ensure an adequate supply of spares? How well does the infrastructure adapt and scale as needs inevitably change? Is it labor-intensive to deploy and administer? How many SKUs must be qualified, purchased and inventoried? The answers to such questions determine the long-term value of any storage infrastructure. Consider these scenarios for SATA deployment in the enterprise:
Scenario One: Raw Capacity
To achieve maximum storage capacity at minimum cost, one approach might be to connect a SATA host server to an external SATA enclosure utilizing SATA RAID backplanes (and multiple PMs). As more SATA drives are needed, additional backplanes (and PMs) could be added to the enclosure.
Because PMs cannot be cascaded, each additional PM requires another host controller port. As drive (and backplane/PM) quantities escalate, the host controller must be upgraded to a more costly unit with more ports. Before long the maximum port count for host controllers (up to 32 in theory, 8 or 16 in common practice) is reached.
By contrast, a SAS infrastructure easily handles a multitude of SATA drives with a single SAS HBA port. As more SATA drives and SAS RAID backplanes are added to the enclosure, additional SAS expanders (incorporating table routing) on those backplanes can be cascaded together.
This SAS-based approach to SATA storage requires only a minimal number of ports on the HBA, and with more ports available on the expanders (theoretical maximum of 128 vs. 15 maximum on PMs), a SAS infrastructure becomes increasingly cost-effective as drive count grows.
Scenario Two: Performance and Capacity
For many enterprises, the mix of online storage and nearline storage dynamically changes as business needs evolve. SAS is particularly appropriate for such environments, offering the compatibility and flexibility to bridge these two distinct storage applications.
As noted in Scenario One, investing in SATA infrastructure to meet bulk storage needs becomes increasingly inefficient as capacity and drive counts grow. But when both capacity and performance requirements must be met, the value proposition of SATA infrastructure immediately plummets. SATA is simply not designed for online, high-availability enterprise storage duty; when greater need for such storage arises, SATA infrastructure must be augmented by additional investment in SAS infrastructure, an unnecessary and costly redundancy.
Using the same server/enclosure model as Scenario One, this redundancy not only entails added expenditures on SAS HBAs and SAS backplanes, it also requires purchasing additional enclosures. In effect, infrastructure costs are almost doubled when SATA is initially installed and then followed by SAS deployment. This also increases the workload on IT departments that must administer two separate infrastructures.
Employing SAS infrastructure to support SATA drives and meet current capacity needs delivers the benefits noted in Scenario One, while ensuring the flexibility to deploy SAS drives by merely plugging them into the existing infrastructure. Thus a single enclosure equipped with a SAS RAID backplane can address both capacity (nearline) and performance (online) applications.
Beyond the obvious efficiency of specifying the optimal drive for a given application, standardizing on the SAS platform will significantly reduce the cost and complexity of the data center by minimizing the number of individual components that must be qualified, purchased, inventoried and maintained. Such component rationalization also results in a smaller data center footprint, and places fewer demands on management resources and support staff.
Greater storage efficiency continues to be an urgent priority for every enterprise, and SAS is uniquely positioned to facilitate this key goal. SAS infrastructure's compatibility with SATA drives enables IT managers to select the most cost-effective storage solution for any task, and it vastly improves SATA scalability. In terms of both maximum drive quantities and long-term deployment flexibility, SAS infrastructure simply outperforms its SATA counterpart. SAS also eliminates the costly redundancies of purchasing and maintaining separate infrastructures for high-performance (SAS) and high-capacity (SATA) storage applications. Clearly, SAS establishes a new paradigm for efficient enterprise storage.
Overview: Expanders vs. Port multipliers Device Maximum # Drives Notes Connections Supported SATA II Port 15 drives SATA only Requires SATA II, Port Multiplier Multiplier-aware host controller SAS Edge 128 drives; one edge SAS and *Multiple edge Expander expander*, one fan-out SATA expanders can be linked expander if table routing incorporated SAS 128 edge expanders SAS and Maximum one fan-out Fan-out and/or drives SATA expander, 16,384 SAS Expander devices in a single domain
Franco Castaldini is senior product marketing manager for Seagate Technology (Shakopee, MN)
|Printer friendly Cite/link Email Feedback|
|Publication:||Computer Technology Review|
|Date:||Feb 1, 2005|
|Previous Article:||Beyond storage consolidation: the benefits of iSCSI SANs.|
|Next Article:||Is SMB disaster recovery really within reach?|