Achieving High Performance Data Analysis with ACL

 2013-08-01

Achieving High Performance Data Analysis with ACL

By Keith Cerny

For my first blog post at ACL, I want to focus on an area that is close to my heart; product performance. When I first joined ACL, I was impressed to learn about the sheer volume of data that is regularly analyzed by ACL customers using our software. Everyday there are thousands of people across the globe analyzing millions of transactional records with ACL Analytics. Our customers have been doing “Big Data” analysis with us long before it was a buzzword.

Just like a race car, fast performance comes from having a well-tuned engine. The analytic engine that drives ACL is called “Ironhide” and it is an incredible piece of technology. This is what powers both ACL Analytics 10 and Analytics Exchange 4 and our engineers have been focused on improving its performance on an ongoing basis. As an example, we improved analytic speed in our recent ACL Analytics version 10 release by an average of 30% and we are working hard to do even better and improve it for future releases. We were thrilled to learn that after the release, some customers experienced up to 60% improvement in some of their analytics compared to ACL Analytics version 9.3, and in other areas like importing data from text files, we achieved up to 70% gains in performance.

These performance improvements are great but we are not satisfied yet. ACL customers can be confident that we’ll continue to make the analytic engine faster and more robust. We thoroughly test performance with every build of the software and keep a close eye on the feature changes we incorporate to ensure our high speed analytics continue to function as expected. Future releases will continue to be faster, but there are things that you can do now to get even better performance with Analytics 10 and Analytics Exchange 4 and these fall into two categories: system hardware and optimal scripting. Kevin Legere has a great series of posts for optimizing ACL script performance, so I’ll focus on the hardware side of things.

Certain equipment and operating systems can help you get some significant gains in analytic performance. For example, if you use a 64bit version of Windows or Windows Server with a Solid State Drive (SSD), you will see a notable improvement in analytic performance. This can be as much as an additional 30-50%, depending on the speed of your disk and the type of analysis you are conducting. ACL Analytics greatly benefits from flash memory and a 64bit OS, so leveraging modern SSD drives like an OCZ Vertex 4 or similar with Windows 64bit will make a noticeable difference. Having a faster CPU and memory also make a positive impact, but OS and hard disk speed will net you the greatest initial performance gains.

One common pitfall that some customers fall into is having their data on a network share away from the actual machine conducting the analysis. Depending on the quality of the network drives and latency, this can have an adverse impact on analysis speed as data moves across the network. We recommend having the project data moved local to the machine that will be conducting your analysis.

ACL Analytics works great on many types of hardware but if you want the optimal setup, you should pair a Windows 64bit OS with a modern and fast SSD drive and keep your project data local. In future posts, I will elaborate further on how we are improving performance in ACL Analytics, leveraging technologies like multi-core processors, in-memory data, advanced sorting algorithms and enhanced 64bit support.

Thursday, August 1, 2013 In: Hot Topics Comments (None)

Contact

Pricing

Demo