SATA Targets Enterprise Storage

The Promise: SATA II will provide highly reliable storage in SAN, NAS, and other storage devices at prices considerably below those of FC, SCSI, or SAS. Simplified connections and new components will enable the use of more drives and provide host redundancy.

The Players: Maxtor, Seagate Technology, Western Digital, Intel, Silicon Image, and others.

The Prospects: Judging from initial implementations of SATA drives in networked storage devices, it seems clear that SATA II's added capabilities will further position these drives as lower priced (but also initially lower performance) alternatives to FC and SCSI drives. As drive manufacturers boost the performance of SATA II drives, the choice between FC and SATA II will boil down to pure price/ performance differences.

To drive makers, network architects, and storage system manufacturers, the difference between enterprise-class drives and desktop and workstation drives is clear: The former are either SCSI- or Fibre Channel (FC)-based, while the latter are populated with ATA drives. But Serial ATA (SATA) is changing that.

SATA drives have already been incorporated into SAN and NAS boxes and are now being integrated into devices that previously only used SCSI or FC drives. SATA is also being integrated into system and workstation motherboards and will almost inevitably become the major type of drive included in such devices in the future.

It's also likely that the storage system you buy next year will be populated not with SCSI or FC drives, but with SATA drives. The reason for this is simple: Although SATA drives may not be as fast as Serial-Attached SCSI (SAS) or FC, additional buffering on the system controller will make the perceived performance of a SATA-based storage device comparable to that of a pure FC SAN--but at a considerably lower price. Because SANs using SATA drives are fast becoming a suitable choice for most computing environments, SANs based on FC drives will become more of a niche product than they are today, relegated to Online Transaction Processing (OLTP) or large database applications where high performance and reliability are critical.

WHY SATA?

Developers of SATA didn't have much choice when it came to designing a next-generation drive interconnect standard. The fact is that Parallel ATA (PATA) had reached the end of its useful life. The industry was using an 80-pin cable for a 40-pin signal so that crosstalk between channels could be mitigated, and performance couldn't be pushed far beyond PATA's 133Mbits/sec. Additionally, PATA drives use 5V signaling. Because most chip foundries now make silicon running at 3.3V or lower, it's become increasingly hard to find the control chips necessary for PATA.SATA overcomes the limitations of PATA and provides bus performance mapped to at least 600Mbits/sec in the near future. Storage device manufacturers are especially interested because the drives are easier to connect, as the new connectors can be used on 2.5-inch or smaller drives. The smaller connectors allow drives to be mounted closer together so that more drives can be packed into a SAN or NAS than was possible using PATA. (Keeping a denser packed box of drives cool will be a challenge, however.)

What's more, SATA delivers the headroom necessary for continued improvement in drive performance for the desktop. SATA drives retain full command compatibility with OSs that support PATA, making them fully compatible with systems supporting the older specification.

Hard drive makers can typically make two types of drives: enterprise-class and desktop. The enterprise-class drives are higher performance, spinning at 7,200rpm, 10,000rpm, or faster. They have larger data buffers and higher quality components that are designed for high reliability. The electronics inside the drives are also more sophisticated. Desktop drives aren't built with as much attention to performance and reliability. They typically run at 5,400 or 7,200rpm and have smaller buffers, and their internal components aren't designed for continuous use.

In addition to building the basic drive, manufacturers must install the logical interconnect. Enterprise-class drives typically get an FC, SAS, or SCSI interconnect, while desktop drives get SATA. However, enterprise-class drives are now getting SATA and SATA II interconnects, making them suitable for use in many enterprise-class storage devices.

WHAT ABOUT SAS? SAS is a next-generation SCSI interface. Like SATA, it overcomes the severe limitations posed by a parallel interface architecture, and even greater problems resulting from awkward cabling and touchy connectors.

The SAS specification was built with SATA compatibility in mind. The cabling and connectors for SAS and SATA are identical, for instance. In fact, a SATA drive can be connected to a SAS interface and run. When a SATA drive is connected to a SAS connector, logic inside the SAS connector allows SATA commands to pass through to the system transparently. It doesn't work the other way around, however. Even though a SAS drive can be physically connected to a SATA interface, the drive won't work because the SATA interface doesn't support SAS.

The original idea behind providing SAS with the ability to connect to SATA devices was to enable storage device manufacturers (or even enterprise users with SAS installed on their servers) to use either SATA or SAS drives. The theory is that in some networks, where high performance and reliability aren't essential, a SATA drive could be used. Where higher performance or increased reliability is required, the SATA drive can be removed and replaced with a SAS drive.

A NAS manufacturer may offer two different NAS systems: a lower priced version with SATA drives, and a higher priced version with SAS. The buyer would be offered the option of upgrading the NAS system by adding more drives, or by replacing the SATA drives with SAS drives.

However, as with the relationship between PATA and parallel SCSI, the SAS silicon is considerably more complex than SATA and thus costs more to create. This cost is extended into a higher price for SAS drives and a higher price for SAS controllers and the devices they're installed on. SAS does have some advantages, however: It has a larger command set than SATA, providing better communications between the drive and the unit requesting data. It's also designed to transfer data more rapidly than SATA drives. Moreover, SCSI is often installed inside SANs that feature FC connections to servers, so FC SANs could conceivably use SAS drives instead of the even more expensive FC drives. In most cases, however, they're not--they're using SATA drives.ENTER SATA II

The developers of SAS may not have fully anticipated SATA II, a higher performance iteration 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 obviously important for enterprise SANs, as well as for smaller RAID arrays and NAS devices. Because less-expensive SATA II drives can be swapped into and out of arrays, this eliminates the need to go to SCSI (or SAS) in order to hot-swap drives.

Additionally, SATA II specifies a port multiplier that enables connections to up to 15 SATA or SATA II drives. However, 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 6Tbytes of storage, assuming 15 400Mbyte drives on a single channel. However, there's no technical limit on the number of drive channels that can be installed on a server or used inside a SAN or NAS system.

SATA II also comes with a new external cabling specification that lets a server and SAN connect over distances of up to 2 meters. A single cable can carry four channels of data. Connecting one cable to a device with four port multipliers can theoretically provide 24Tbytes of storage using the above assumptions. Realistically, a SAN with 20 drives (five on each channel) can deliver 8Tbytes of storage.

Host redundancy can be achieved via a port selector. The port selector can be attached to two hosts so that when the primary host fails, the port selector can automatically switch control to the secondary host, thus enabling rapid failover.Some observers argue that ATA--either parallel or serial--will never have the reliability or performance of FC or SCSI. However, SATA and SATA II are merely drive interfaces. It's up to the drive makers to come up with enterprise-class SATA drives, and that's exactly what they're doing. Hitachi Data Systems is a major seller in the high-end storage marketplace, providing enterprise-class drives used by Nexsan Technologies in its ATA-based SAN systems. Seagate Technology will ship its enterprise-class SATA drives this month.

It's clear there's a split going on in the SATA drive marketplace. Drives intended for use in network storage are being built to enterprise-class specifications, with faster spindles, larger caches, and more robust components than the lower priced consumer-class drives. It's likely that consumer drives will continue to be built with the SATA interface, while enterprise-class drives will be SATA II.

Mark Brownstein can be reached at [email protected]. Send comments to [email protected].