
Few people would argue, that when it comes to IT, performance and power remain integral to IT supporting business. People want more and more of it, especially in this era of big data, where to get the required insight from that data requires immense computing power.
But there is a perception that performance costs money. That’s why it is surprising to learn that many end-users are not getting anywhere near the maximum performance from their current infrastructure, and that with a few minor tweaks and adjustments, could get up to 60% more performance with no need for a costly upgrade or migration. This applies both to users of bare metal and to those with a layer of virtualisation.
How can this be? Well, my company recently conducted a series of benchmarking studies that sought to assess areas of potential performance improvement in infrastructure. As an IaaS we operate tens of thousands of servers and constantly test their performance and had realised that performance was not as it could be. So we used the Linpack, SysBench and TPC-DS standard hardware benchmarking tools and the findings make interesting reading for anyone involved with infrastructure performance. Users can get up to 60% in any infrastructure, by following these straight forward steps:
1. Doubling the amount of CPUs does not double performance and removing one can actually improve database performance by up to 15%
One might think that adding additional CPUs would improve the performance of a database. Yet in the benchmarking – on different databases, as well as in SysBench and TPC-DS tests – showed that single processor bare metal instances can actually generate better performance than dual processor machines, assuming everything else is equal. So adding an additional processor can in fact, reduce a database’s performance. Both processors will in fact use each other’s memory, which means things take longer. Used individually, eah CPU will only access its own memory
2. Choice of OS can be critical to application performance
The Linpack benchmarks revealed that CentOS 6.4 can yield up to 20% better performance, when paired against Ubuntu 12 LTS. It is a widely-held view that OS choice has only minimal impact on performance. But by checking the differences between kernel versions and settings and configuring the OS to match those of the better performing system, a user can obtain a major performance boost, without any hardware upgrade.
3. By improving memory frequency, users can increase performance by more than 20%
RAM speed is on the whole, considered to be sufficient, meaning that memory frequency gets overlooked by many. This is an oversight. In Linpack benchmarks, replacing 1333 MHz DIMMs with 1866 MHz DIMMs from the same provider meant an increased overall computing performance (total no. of GigaFLOPs) of 20%
NoSQL databases and other applications that are heavily impacted by memory access time, are particularly improved by this metric. The physical placement of memory in slots will also impact performance. Memory chips placed incorrectly will not make use of their top frequency and may only even function at half their intended performance. Some vendors provide online configurators for optimum memory placement inside the bare metal server. A simple change in the memory distribution across machines, could result in a sizeable increase in frequency and performance.
4. CPU intensive application performance is impacted by hyper-threading
The idea behind Intel’s Hyper-Threading technology is simple – to accelerate CPU performance. In many instances it does exactly that. But not all. With CPU intensive applications, deactivating Hyper-Threading can actually yield 5-10% better performance, and in bare metal environments, deactivating Intel Virtualization Technology (Intel VT-x) offers an additional performance boost of up to 5%. These are small improvements in isolation, but together they are truly significant.
Additional performance is always valuable to any organisation. Most do not have the budget to easily upgrade their infrastructure so they must instead turn to other routes. Applying these measures, based on our recent benchmarking could help any IT director obtain up to 60% more performance from their existing infrastructure – an improvement not to sniffed at.
Disclaimer: This article was written by a guest contributor in his/her personal capacity. The opinions expressed in this article are the author’s own and do not necessarily reflect those of CloudWedge.com.