A primer on multiactuator HDD technology

Multiactuator technology makes it possible to build HDDs that offer both high capacities and high performance.

Seagate has been at the forefront of this effort, followed recently by Western Digital. Their efforts represent only the beginning. If their drives work out, the new technology could change how many enterprise HDDs get built in the foreseeable future. To be prepared, storage admins should understand how multiple actuators work together, as well as the benefits and challenges ahead.

How do multiactuator HDDs work?

A multiactuator HDD is a hard drive that contains more than one actuator. These drives are often referred to as dual-actuator HDDs because the current models contain only two actuators. It’s conceivable that more will be added in the future if the current dual-actuator technology takes off.

In the past, HDDs contained only one actuator, and most HDDs are still built this way. An actuator is a physical assembly within an HDD that moves read/write heads over the surfaces of the drive’s spinning platters.

Today’s HDDs commonly include multiple platters, actuator arms and read/write heads. Despite all these components, however, an HDD with a single actuator can read/write data from only one location at a time, which means that only one head is active at a time.

A drive’s performance is determined, in large part, by how efficiently the actuator can access the data at different locations on the platters. The greater the density, the more difficult this process becomes. Even the fastest and most efficient actuator can get bogged down when the density becomes too great.

To overcome the limitations of a single actuator, both Seagate and Western Digital have come out with dual-actuator HDDs that, in effect, split the single actuator into two. In this configuration, the two actuators share the same axis, but each one has its own set of actuator arms, read/write heads and voice-coil motor. The actuators operate independently of each other, with each one controlling half of the drive’s actuator arms.

Because of their independence, the two actuators can access data from different locations at the same time, as if there were two separate drives. This independence makes it possible for the actuators to read/write data in parallel, thus increasing the data flow to or from the drive.

For drives that use the SAS interface, the host system sees the HDD as two independent LUNs with equal capacity. For example, an 18 TB dual-actuator drive appears to the host system as two LUNs with 9 TB of capacity each. For drives that use the SATA interface, the host system sees only one 18 TB logical device.

Why so much interest in multiactuator HDDs?

Multiactuator HDD technology holds promise for certain future workloads. Such workloads will require high capacities and strong performance but not necessarily the level of performance that justifies the investment in SSDs.

Over the years, advancements in HDD technology have helped to improve both capacity and performance, greatly benefiting these types of workloads. The demand for capacity and performance continues unabated, however — a trend that will likely carry forward in the foreseeable future, especially as data volumes grow unchecked. Already, enterprise HDDs are starting to face their practical limitations.

An HDD’s mechanical components, particularly the actuator, can affect both sequential and random performance. If HDD capacities continue to grow without increasing the data rate, a single actuator will no longer be able to get to all the data quickly enough to keep up with the workload demand, causing performance to suffer. As capacities continue to grow, the hit on performance will suffer even more. These increased capacities also mean that it will take an inordinate amount of time to rebuild a drive in the event of failure.

Multiactuator HDDs could prove instrumental in supporting workloads such as mail servers, video streaming, content delivery networks and Hadoop clusters.

As the industry moves toward 30 TB and 40 TB drives, hyperscale data centers will require HDDs that can deliver the capacities they need without sacrificing performance. At the same time, the data centers must still be able to store their data at an affordable cost per terabyte, especially large stores of frequently accessed warm data. This is where multiactuator HDDs come into play. They offer a solution for maintaining or even improving IOPS, while still moving toward greater capacities.

Multiactuator HDDs are not a replacement for SSDs. They fill the need of hyperscale cloud providers and other data centers that require storage systems capable of scaling to the capacities they need without breaking the bank. SSDs can’t be beat when it comes to performance, and they’ve displaced HDDs in several situations. Yet, HDDs can still support many workloads now and in the future, while offering a better cost per terabyte than SSDs.

Multiactuator HDDs could prove instrumental in supporting workloads such as mail servers, video streaming, content delivery networks and Hadoop clusters.

Benefits, challenges of multiactuator HDDs on the market

Seagate recently introduced the 18 TB Exos 2X18 drive. The multiactuator HDD delivers a maximum sustained transfer rate of 554 MBps, with random read/write rates up to 304 IOPS and 560 IOPS, respectively.

The differences between Exos 2X18 and Exos X18, an 18 TB single-actuator drive, showcase the improvements in performance. X18 provides a maximum sustained transfer rate of 270 MBps — less than half of that offered by 2X18. The read rates are also significantly lower, coming in at only 170 IOPS, a little better than half that of 2X18. Random writes fare a little better. The random write speed for X18 is 550 IOPS, compared to the 560 IOPS of 2X18.

Western Digital recently launched Ultrastar DC HS760, a dual-actuator drive that provides 20 TB capacity. Western Digital has yet to publish detailed specifications for its new drive, although the vendor claims it doubles sequential throughput and increases random performance 1.7 times, compared to the 20 TB Ultrastar DC HC560 HDD.

A prototype of a dual-actuator drive Western Digital demoed at the 2019 Open Compute Project Summit.

The Exos dual-actuator drives don’t offer better latency rates than the single-actuator drives. 2X14 and 2X18 have an average latency rate of 4.16 milliseconds, the same as X18 and X20. As for the Western Digital drives, until specifications are available for Ultrastar DC HS760, it is uncertain how latency might compare to the vendor’s single-actuator drives.

Incorporating multiactuator HDDs into the data center might come with a few growing pains. For example, it could be necessary to modify host software to properly handle file placement between the two logical volumes. The LUN behavior across the two partitions could lead to confusion because it’s not always apparent which commands affect the individual partition and which affect the drive as a whole.

The industry might need time to catch up to multiactuator HDDs. For example, some host bus adapters might not provide out-of-the-box compatibility with the current dual-actuator drives. IT teams could run into issues around vendor lock-in because the market is still so small, although the introduction of the Western Digital drive should help alleviate some of that concern.

Both Seagate and Western Digital have introduced their multiactuator HDDs with relatively little fanfare. Perhaps they’ve taken this approach to see how the technology performs with current workloads and to get a better sense of what the market will bear.

It might also be why vendors such as Toshiba have yet to toss their hats into the multiactuator ring. They, too, want to get a better sense of which way the wind blows before going all out for these new HDDs.

Robert Sheldon is a technical consultant and freelance technology writer. He has written numerous books, articles and training materials related to Windows, databases, business intelligence and other areas of technology.

You may also like...