Printer Friendly Version

Flash Solid State Disks: A New Breed of Enterprise-Class Storage
By Jun Alejo, BiTMICRO Networks

Page: 1 | 2 | 3 | 4 | 5 | NEXT

CPU processing speeds have evolved tremendously over the past decades, with giants like Intel, AMD, and Transmeta playing major roles as pioneers of the CPU revolution. These innovators have persistently pushed through major advances in silicon technology to extend or even expand Moore's Law. Although hard disk drive technology also underwent major improvements over the years, significant progress has only been achieved in the area of storage capacity. Figure 1 below illustrates diverging paths between CPU speeds and storage transfer speeds through the years.



Disk storage has evolved from a mere few MBytes in large form factors to several GBytes in much smaller form factors (see Figure 2). As mentioned earlier, performance improvements have not kept up with its CPU counterparts.

Although rotational speeds for HDDs have constantly increased over the years, and access times have been reduced to approximately 3.2 milliseconds, the CPU-storage performance gap is significantly undermining the potential of computing and storage networks. Instead of boosting system performance, HDDs are in fact hindering the CPU's performance, preventing the entire system to unleash its full potential. Looking at I/Os Business enterprises are very interested in disk performance as it relates to their enterprise-level performance requirements. The most-often quoted metric measuring disk performance is that of the rotational speed of the HDD which, for most major disk vendors, defines the difference between high-end and low-end disk lines. Today, rotational speeds are typically 5,400rpm, 7,200rpm, 10,000rpm and 15,000rpm. However, in a vast majority of enterprise applications where non-sequential data access is required, access time becomes a much more important, yet usually ignored, metric for determining performance. Access time is the combination of three important disk metrics. The first is controller overhead, which is the time taken by the controller embedded on the disk to process the operation. The second is seek time, which is the time taken for the disk head to move to the correct disk track. Finally, there is rotational latency, or the time taken for the disk to rotate far enough so the head reaches the data required. Let us now take a look at how many I/O transfers a storage device can accomplish in a given time frame for different block transfer sizes. Charts 3 and 4 show internal benchmarks that plot the performance of BiTMICRO's Ultra320 SSD vs. a Seagate ST336753LC Ultra320 80-pin, 15,000rpm HDD in terms of random I/O for varying block transfer sizes. These results lend credence to the astounding advantage that SSDs claim over mechanical disks for random I/O in particular, as delays due to rotational and seek latencies are virtually eliminated. However, as the block size of the I/O transfers increases, the need for HDDs to seek to different sectors is                                                                                                        diminished.

Page: 1 | 2 | 3 | 4 | 5 | NEXT

[ Top ]