Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

SATA Express marries PCI-e, SATA for faster direct-attached storage

With the speed and power improvements introduced by SATA 3.2 specifications, SATA drives are making a comeback for direct-attached storage.

The Serial Advanced Technology Attachment 3.2 specification from the international SATA-IO organization uses Serial ATA Express, power conservation and other improvements to make direct-attached storage faster and more intelligent.

While centralized, network-reliant storage is front-of-mind for large data centers, direct-attached storage distributed among individual servers or blades enables storage-intensive workloads that otherwise bog down networks, and provides local data caching.

Serial ATA (SATA) devices' low cost and high capacities are luring system designers away from serial-attached SCSI (SAS) for server-side storage. A single SATA drive, such as the Seagate Barracuda family, can easily provide one terabyte of storage. But rotational speeds around 7200 rpm mean latencies are a problem; the traditional SATA data interface is limited to about 0.6 GB per second (GBps).

SATA, PCI-e and SATA Express

SATA 3.2 expands IT administrators' choices, thanks to SATA Express, which supports SATA and PCI Express (PCI-e) protocols simultaneously.

Solid-state drives (SSDs) and hybrid hard drives -- mechanical drives with a large internal cache space -- offer extremely fast data-transfer potential, with solid-state memory accessible at speeds at least an order of magnitude faster than traditional mechanical designs.

SATA and SAS drive interfaces weren't developed to accommodate these devices, which creates a data bottleneck. The PCI-e bus is tailor-made for fast data transfers, roughly 1 GBps per serial lane. But it was not used as a storage interface because drives could not take advantage of PCI-e bandwidth.

The SATA Express specification now recognizes PCI-e as a suitable storage interface for fast storage devices. Conventional drives will continue to use the server's standard SAS and SATA interface operated through the motherboard's South Bridge chip, while new, fast disk drives like SSD use SATA Express through the PCI-e interface.

While this new option helps meet the performance and capacity demands of varied workloads, servers will need SATA Express capabilities and SFF-8639 multifunctional connectors to take advantage.

Smarter storage management

The SATA 3.2 revision adds several features for server and storage management as well.

DevSleep power conservation mode powers down the storage drive almost completely, yet leaves it on and connected. Current power-down modes allow a partial power-down and restoration to full operation within 10 microseconds, or a sleep mode with more power savings and a return to full operation within 10 milliseconds. The DevSleep mode is beneficial for mobile device storage and servers in computing environments that use dynamic workload migration.

Storage devices use power to change modes, so transitional energy reporting can aid storage management at the server level by providing detailed information about the SATA storage devices' mode changes. The server makes better and more efficient decisions about changing storage's state based on the information. For example, servers might perform fewer power conservation transitions than previously expected because the transitions draw more power than leaving storage in a powered-on state does.

Hybrid drive optimizations in SATA 3.2 address solid-state hybrid drives where a large flash memory space is placed as a cache between the interface and platters. New solid-state hybrid drives allow the server to dictate cache preferences to the storage device and improve cache hits. Commands can also be queued to ease the effect on drive performance.

Finally, SATA 3.2's rebuild-assist feature speeds data recovery from a failed RAID drive. Normally, data is reconstructed by processing parity data spread across the other members of the RAID group, which is time-consuming. Rebuild assist determines which data on the failed drive are unreadable and focuses the rebuild effort on that portion alone.

Dig Deeper on Enterprise data storage strategies

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.