Media and Cloud Usage Drive HDD Business
Oct 20, 2011 - Santa Clara, CA - At the IDEMA’s Diskcon event, there were two main topic for the presentation - the technology behind increasing areal density for rotating media (HDDs) and the consumption appetite of the cloud and media content as the new the driver of the marketplace. With storage and compute in the cloud, the main business shift of large scale storage being part of an OPEX budget rather than a CAPEX budget is in place. This has the consequence of the HDD itself not being an object of recommendation and purchase, rather the storage capacity that is available is being purchased on a dynamic “pay as you go” basis. This is causing the large storage providers to clamor for increased areal density, to put more capacity in the same power and footprint. In this use model, the cloud storage application masks the IOPS advantage of SSDs over HDDs. However, HDDs lead with the capacity and reliability of the data for longer term storage and known error correction methods.
Current areal densities are near 1TB/sq in. The research in progress is to achieve densities of 2TB/sq in. The keys for achieving these densities focus on surface prep, overcoat technology and the method of magnetic recording used. Historically the format shift that allowed the 1TB/sqin densities was PMR (perpendicular magnetic recording). The new methods being explored include NCMR (NanoContact Magnetic Recording), and SMT (Spin Miss Tracking). SMT is based on tracking the electron rotation that is not in sync with the spin of the magnetic material. Another method is SMR (Shingled Magnetic Recording) this uses a corner based read head to allow for overlapping of the tracks. (See illustration from - article 0018-9464 © 2009 IEEE - IEEE Transactions on Magnetics, Vol. 45, No. 2, February 2009). SMR has been theorized to support up to 10TB/sq in, and is the leading method for short term increase of areal density, as well as being a postulized alternative to tape for long term storage. Other methods discussed included BPM (Bit Patterned Media) and HAMR (Heat Assisted Magnetic Recording).
These large drive densities are needed to support the growing need of media content and user generated content being put on the cloud. Cloud storage has the advantage of allowing multiple content views per stored object than individual local storage. As as a result, large storage subsystems (tiered storage of DRAM, SSD, and HDD) can be configured and managed (backup, de-duplication) than distributed local storage. These enterprise drives are still dominated by the SAS and SATAII interfaces.
There is a transition in place to higher speed 6Gb/s SATA III drives using the new 4K sector Advanced Format (AF) and these should be a measurable percentage of the installed base by 2013. Another high speed interface that is targeting the enterprise community is SOP (SCSI over PCIe). To address some of the challenges of the BER issues and create a common interface, there is a SOP-PQI (PCIe Queuing Interface) that is addressing packet optimization and buffering issues for the data. These high speed interfaces are attempting to get higher data throughput from a mechanical systems (rotating media and read head) that currently is maxed out at ~500Mb/s from the heads on a 15K RPM drive. Other methods to increase this data interface are DRAM creep (enterprise drive now has 32MB-64MB of cache) and RAID. A discussion of Thunderbolt as an interface, places the connection format as an endpoint storage mechanism that is targeted as an IEEE 1394 replacement, and is not really being looked at for enterprise applications.