Organizing Windows for Performance

Pages: 1 2 3 4 5 6 7

Purpose

At this point, some may be wondering what my reason is for doing these tests. After all, everyone knows that you always put your fastest drive as the primary master and install your other hard drive on the secondary channel. This obviously results in the best performing configuration.

Well, it isn’t always that simple, in my opinion. The one constant in the PC market is that cost is the main consideration in the majority of purchasing or upgrading decisions. Those who sell components and/or provide upgrading services must always keep this in mind, because the one who can offer the same or similar component/service as the competition at a lower price will win the customer.

So, the question in my mind as I performed these tests was “Can I get similar performance by putting the fastest drive as the secondary?”. If the answer to this is yes, it provides vendors and service technicians an alternative for those who just don’t want to spend much money. Consider the cost savings of having a new hard drive installed if it does not also require a backup and restore of the existing drive onto a new one – or even worse, a complete reinstall of the OS and applications!

Keeping in mind the limitations of these benchmarks, as outlined on the previous page, and the usage profile of the customer, might it be possible to offer an upgrade for a customer that is not only cheaper than the competition (i.e., includes only the cost of a new hard drive and installation), but provides performance that is almost as good? This is one of the questions I was seeking an answer for.

Another question was whether moving the swap file and/or data files onto a separate drive would provide the same or similar performance improvement as a RAID 0 setup. It is my impression that the benefits seen with RAID 0 in these benchmarks is essentially due to the reduction of I/O latency by taking advantage of the cache on two drives rather than due to greater bandwidth. This theory could be proven by setting up identical tests with the same drive as both a primary slave and a secondary master. If bandwidth were the issue, then the secondary master should perform better, otherwise we should see essentially identical performance between the two setups. In addition, if latency is the issue, then a dual-drive setup configured properly should result in a performance improvement roughly equivalent to a RAID 0 setup.

Though I did not do any RAID 0 testing myself, John Howland’s “IDE Raid – Is There a Benefit?” provides some reference numbers. According to the Winstone and Content Creation scores, RAID 0 provides about a 3% to 4% improvement over a single IDE drive in actual operation. Again, this does not measure boot time, nor does it fully measure application load times. However, as John indicated in another article, though the OS itself may boot faster with RAID 0, the additional time required to go through the RAID BIOS POST routine tends to offset any gains there. Users of RAID 0 have claimed significantly reduced application load times, but since this is a subjective evaluation I would prefer to see some hard numbers for it – which are not easy to come by, unfortunately. Therefore, I can only go by the numbers provided from the benchmarks, with the caveats given previously.


Pages: « Prev   1 2 3 4 5 6 7   Next »

Be the first to discuss this article!