Friday, January 7, 2011

Six reasons why the 2½” FF is likely to replace the 3½” FF as the HDD of choice in performance array’s.

Part 2 of my blog crystal balling the adoption trend and marketplace momentum for 2½” FF HDD’s as a legitimate enterprise storage option. This continuation of my December 28th post will auger into a bit more detailed as it focuses on the six key areas that I believe are propelling this next generation design into a leadership position in enterprise storage.

1. Drive Capacity: 2½” drives now support 1TB. Not quite on par with the 3½” drive but impressive when looked at in terms of volumetric density.

2. Volumetric Density: How much storage capacity is available per unit volume? Using the efficiency ratios detailed in paragraph 3 and based on maximum current drive capacities of 1TB for , 2½” drives and 2TB for 3½” drives the volumetric storage density advantage for the 2½” form factor is 2.9 and 1.88 respectively for consumer and enterprise class drives.

3. Volumetric Efficiency: How many drives can be packaged per unit volume? Each 3½” drive occupies 391³ cm compared to the slimmer 2½” form factor at 66.6³ cm for a consumer class and 104³ cm for the enterprise class drive. This translates to a 5.8 and 3.76 space utilization advantage respectively meaning more drives hence more spindles per unit volume. The lower mass of the 2½” drive will minimize rotary random-operating vibration (RROV) enabling much denser packaging. However, limitations will temper the magnitude of the actual advantage realized in packaging density.

4. Access Density: Access Density dictates the speed at which an end user can access data. It is defined as the ratio of drive or system I/O performance over data stored, (IOP/GB). If capacity doubles and performance doubles then access density remains unchanged. This would be a good thing, a good theory, but unfortunately in the evolution of hard drives that desired relationship is never the reality.

The problem was well articulated in April 2000 by Fred Moore wrote in CTRthat while the capacity of a disk drive has increased 6000 times since 1964 the raw performance, seek, latency and transfer rate has only increased by a factor of 8”. A massive imbalance and points to the issue that scaling disks is more than just capacity and despite techniques to counter the problem such as larger cache and actuator level buffers the imbalance remains.”.

In short as areal density has increased, more capacity sits under each actuator, increasing drive latency and causing the access performance of these drives drop. This access performance (access to data) hit extends to the subsystems that use these drives. So the obvious answer when driving data latency improvements is to have less data under each actuator and the 2.5” form factor enables more spindles in the same or better capacity footprint which will mitigate the data access issue.

Simply put, greater volumetric density translates to more drives per unit volume which equates to an increase in spindle density driving greater access density meaning faster access to data.

5. Energy Efficiency: How many TB can be supported by 1W of energy? While there are many variations the idle power for the Seagate Momentus®, 2½” consumer class drive is rated at 0.96W with an average power rating of 2.5W. For the Seagate Constellation.2™, an enterprise SATA drive. the idle power is 3.31W with an operating power of 5.21W. In comparison a 3½” drive consumes about 8W on idle and 12W on average. These drives all have a rotational speed of 7200RPM; 5400RPM drives are, as you would expect, even more power efficient. However 3½” drives have greater maximum capacity, 2TB as against 1TB for the smaller 2½” form factor. I know 3TB, 3½” drives are on the horizon but to my knowledge they are not yet available in enterprise storage arrays. To normalize the capacity difference the power consumption numbers were compared for a 5PB storage array using different drive types and while the power consumption ratios were reduced they were still significant particularly when considering that drives can be as much as 80% of the power hogs in a storage array.Perhaps a sharper comparison is to compare the GB/W metric.
    1. A 3½” 7200 enterprise drive, 2TB, supports approximately 164GB/W.
    2. A 2½” 7200 enterprise drive, 1TB, supports approximately 192 GB/W.
    3. A 2½” 7200 consumer drive, 750GB supports approximately 300GB/W.
Compare these numbers to a high performance 15K drive such as a Seagate Cheetah, 15K drive will support a mere 37GB/W. The Seagate Savvio drive is a high performance, low capacity 2½” drive with a rotational speed of 15K. While the Savvio offers a very miserly 22GB/W (depending on IOP) it can deliver 2.5 times the I/O performance of its 3½” “Cheetah” counterpart. So depending on the need the Savvio could efficiently replace the Cheetah particularly if the ‘Cheetah is subject to the highly inefficient practice of short stroking. In some circumstances it might even be a better performance option than SSD.

6. Performance: As already noted the greater the number of spindles the greater will be the I/O performance. However, bandwidth is another positive beneficiary of a higher spindle count to capacity ratio and when combined with miserly power demands (Savvio excluded), 2½” architectures can be a very power conscious, bandwidth solution.

The bottom line is that a storage array based on 2½” drive technology can deliver higher storage density, more IOPS, higher aggregate bandwidth and with a lower energy consumption than an equivalent solution using current 3½” technology. The remaining challenges however are cost and and to an increasingly lesser extent, reliability. While the cost of the 2½” is rapidly declining its price point continues to exclude it from applications where capacity is the driving requirement, particularly now that 3TB capacities are availability horizon of the 3½” form factor. However, if high performance and capacity are the requirements then 2½” technology is a very viable option.

No comments: