Serial ATA (SATA) has been designed as the replacement for parallel ATA, the most common disk drive interface in home and office computers. SATA I drives and controllers are available in many new desktop computers, but the current buzz revolves around improvements to the SATA specification that make SATA more useful in enterprise environments. Right now, companies are announcing plans to ship products built on this improved specification, called SATA II, by the end of 2004.
What's the real difference between SATA I and SATA II? SATA II is better suited for enterprise environments because of three features: port multipliers, port selectors and native command queuing.
Port multipliers: In SATA I, parallel ATA drives had to be configured as master and slave and daisy-chained from each controller. The port multiplier spec from the Serial ATA Working Group allows up to 15 drives to be connected to a SATA controller via a port multiplier. Although this is far less than the number of drives that can be connected using Fibre Channel or Serial Attached SCSI (SAS), this will make it much easier to build disk enclosures using SATA drives.
Port selectors will allow two hosts to be connected to one drive. This is useful because it creates a redundant connection to the disk drive. This way, if one of the hosts has a failure, the second host, acting as a spare, can take over, so that access to the storage is maintained. This type of redundancy is essential for enterprise environments.
Native command queuing will improve the performance and efficiency of SATA II drives.
Normally commands will arrive at a disk to read or write from different locations on the disk. When
commands are executed in the order they arrive, a great deal of mechanical overhead is created
because the read/write head is constantly being repositioned. SATA II drives will use an algorithm
to determine the most efficient order to execute commands, reducing mechanical overhead and
Each of these additions to SATA will make it much more useful to enterprise data centers. However, these improvements don't mean that enterprises will forget about SCSI and Fibre Channel and run into the arms of SATA. SCSI, Fibre Channel and soon SAS will still be the main candidates for mission-critical storage. SATA will be used to solve other problems in enterprise storage, notably nearline storage.
In the past storage was considered either online or offline. Online referred to the data center that operated 24/7 and usually contained high-performance Fibre Channel and SCSI arrays. Offline referred to data that was not immediately accessible, because it was backed up (usually off-site) on tape. As ideas about ILM have gained traction, it has become clear that data needs more than two states.
Often, as data ages, it loses business value; therefore that data is accessed less as time goes on. A file created yesterday is more likely to be accessed today than a file that was created several months ago. As data ages, it doesn't make sense to store it on expensive high-performance drives. It's more logical to take the performance hit and move the aging data onto lower-cost SATA drives, and eventually, to tape. This way, data on SATA drives stays more accessible than it would if it were on tape, but valuable space in the main data center is reserved for the data that really needs to be there.
New SATA drives are being designed to be available 24/7, with longer MTBFs. They are not designed to be as durable as enterprise-class SCSI or Fibre Channel drives. Enterprise drives must be available and running all the time in high I/O environments. SATA drives must be available all the time, but won't get as much traffic in low I/O environments. Because the drives will experience less wear and tear, they don't need to be as robust as enterprise drives, and therefore can be significantly cheaper. SATA drives are priced comparably to parallel ATA drives, and are not nearly as costly as higher end disk drives.
With all these improvements under the hood, SATA will be showing up in many data centers over the next year, and will likely prove an essential tool in managing information as well as the cost of storing it.
About the author: David Woolf is a testing engineer at the University of New Hampshire InterOperability Laboratory (UNH-IOL), a non-profit interop lab hosting industry-wide testing in Fibre Channel, iSCSI, SATA and SAS. He can be contacted at firstname.lastname@example.org.
This was first published in November 2008