Western Digital accelerates SATA hard drives to 500 MByte/s

So far, hard drives have achieved around 300 MByte/s in their outer zones, which is very slow compared to PCIe SSDs: the fastest M.2 SSDs reach almost 14,000 MByte/s when writing. However, many SSDs used in data centers forgo an accelerating SLC cache and write data directly into the flash cells; as a result, the write speeds of QLC SSDs can drop to well below 1000 MByte/s.

Western Digital (WDC, Western Digital Corporation) now wants to bring its data center hard drives to a read and write speed of 500 MByte/s. No hardware changes are necessary for this, said the company's product manager, Ahmed Shihab, in an interview with heise online. The basic idea is immediately obvious: instead of using just one head to write or read a track, WDC uses multiple heads and simultaneously writes to or reads data from multiple platters.

So far, no hard drive manufacturer has commercially implemented this technology. The tracks on modern hard drives are only about 30 nanometers wide, and tracks lying "on top of each other" on different platters are likely never truly in the same position. Only by using a very finely adjustable joint at the head tip can this be brought into the correct position – the company relies on the Triple Stage Actuator introduced a few years ago for this purpose (WDC explains its function in a YouTube video).

According to Shihab, such drives are already in use with pilot customers. He said no changes to the interface or housings are necessary. Customer reactions have been positive, with customers asking about commercial availability and even higher speeds.

Western Digital calls these drives High-Bandwidth Drives. The primary goal is to fully utilize the SATA interface. In principle, however, much higher speeds are possible, for example with the proven SAS interface or even with a PCIe interface. Shihab emphasized that developments are based on concrete customer requirements. The goal is to reinvent the hard drive for the AI era without breaking up existing customer infrastructures. We do not expect a later upgrade of existing drives with support for parallel operation.

Among the developments driven by customer requirements is a reduction in the power consumption of the drives by around 20 percent through reduced rotational speed. Such drives could be used, for example, in a more affordable storage tier; despite the 20 percent energy saving, their performance is only 7 percent lower. Shihab did not announce specific products.

More Hard Drive Capacity

The High-Bandwidth Drives are not yet ready for mass production. Shihab announced another capacity level for traditional server drives: in the second half of the year, an UltraSMR hard drive with a capacity of 40 TByte is expected to be released. Like the 32 and 36 TByte models, it uses write assist, a technology WDC calls Energy Assisted Perpendicular Magnetic Recording (ePMR).

UltraSMR is an extension of normal SMR hard drives, which write data overlapping (Shingled Magnetic Recording) and thus offer 10 to 15 percent higher capacities. UltraSMR provides more than 20 percent capacity increase through larger SMR areas and improved error correction, but also requires appropriate control by the host controller or storage software. Such drives cannot be used on a standard PC.

ePMR is expected to be sufficient for capacities up to 50 TByte, but no launch dates for such drives are known yet. In parallel, WDC is further developing HAMR technology, which is intended for drives with well over 100 TByte. 100 TByte drives are already possible in the lab, and the physical principles for 200 TByte drives are also known.

Empfohlener redaktioneller Inhalt

Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.

Preisvergleiche immer laden Preisvergleich jetzt laden

Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.

(ll)