Originating Author: Fred Moore
Magnetic disk drive capacities are growing between 35-50% annually and development roadmaps now outline a clear path to a 10+ terabyte (TB) drive capacity over the next 10 years until 2017. However, disk drive performance capabilities have not kept pace with the continual growth in disk capacities. Disk drive performance improvements are well below 10 percent annually failing to offset the capacity increases. Raw disk drive performance is normally measured in total random I/Os per second and can surpass 100 I/Os per second for an average access time (defined as average seek, latency and data transfer) of about 10ms per I/O on select high performance drives.
Performance Issues
Continual increases in capacity, without corresponding performance improvements at the drive level, create a performance imbalance that is defined by the ratio called disk Access Density. Access Density is the ratio of performance, measured in I/Os per second, to the capacity of the drive, currently measured in gigabytes (Access Density = I/Os per second per gigabyte). This diverging trend is now a growing concern for end-users as actual allocation levels often decline as disks get larger to maintain acceptable performance levels. Poor utilization leads to increased operating expense. The impact of reduced access density includes:
- end-user allocates less space on the drive to maintain performance levels
- add more drives gaining more actuator arms to support increased activity levels and capacity
- increased time storage administrators spend tuning disk drives
All of these add costs to either the infrastructure or to operational expenses, or both.
Disk drive suppliers driving density and cost
The primary focus for the six remaining data center disk drive manufacturers (Seagate, Western Digital, Hitachi GST, Samsung, Toshiba, Fujitsu) disk manufacturers remains on increasing drive capacity and lowering the purchase price per gigabyte. The “price per gigabyte is everything” mentality is creating additional storage management issues for IT users. Note that the standard 500GB, 750GB and 1,000GB (one terabyte) drives all have the same performance specification, just more platters. The access density for the 1,000GB drive is one half that of the 500GB drive.
Capacity | Access Time | Rotational speed | IO's /sec | Access density |
---|---|---|---|---|
500 GB | 8.9 ms (read),9 ms (write) | 7,200 rpm | 80-90 | .16-18 |
750 GB | 8.9 ms (read),9 ms (write) | 7,200 rpm | 80-90 | .11-.12 |
1000 GB | 8.9 ms (read),9 ms (write) | 7,200 rpm | 80-90 | .08-.09 |
Realities
If capacity doubled and performance doubled, the access density would remain unchanged. Unfortunately this is not happening. Higher-capacity disks will normally contain more and more files with the increased probability that they are concurrently accessed. Larger disks become less beneficial for applications that have a large number of concurrent applications or users. Increased contention for the single disk actuator results in elongated response times and high queue depth. As disk drive capacities continue to increase much faster than drive performance, expect the disk drive industry to deliver two categories of disk consisting of smaller-capacity, high-performance drives and larger-capacity, lower-performing drives. The advent of imbedded flash memory in disk drives may offer some relief but the cost/GB will increase.
In reality, the access density has steadily declined as the capacity has increased substantially. Larger caches and actuator-level buffers help improve overall subsystem performance. Access density is becoming a significant factor in managing storage subsystem performance and the tradeoffs of using higher-capacity disks must be carefully evaluated as lowering the cost-per-megabyte most often means lowering the performance. In general, very high-capacity disks are not well suited for applications that have a large number of concurrent users as too much contention for the actuator results. Future I/O-intensive applications will require higher access densities than are indicated by the current development roadmaps.
Action Item: Adding higher capacity disks generally mean lowering your acquisition cost per gigabyte while increasing your total operational costs per gigabyte. Users should balance capacity, performance CAPEX and OPEX and not just jump to the lowest cost devices.