By Lingxiang Cheng, Senior Bardess Consultant
In my last “QlikView Enterprise Deployment” article, I discussed VMware, the benefits of separating Publisher from QlikView Server, the choice of QlikView web server vs IIS, and security. However, there is more to consider in making your system solid.
The QlikView deployment has been so successful, people cannot live without it anymore. All of a sudden IT’s are victims of their own success. People are demanding high availability and high performance of the QlikView system, which of course will make the system more complicated. QlikView Servers need to be put on more than one machine, and a load balancer needs to be put in front of these machines. QlikView Servers also need to be clustered. By doing this, if one server is down the load balancer will stop sending users to this server. Users also enjoy a better response time because they are spread across multiple QlikView Servers.
The same concept also applies to Publisher. I have seen customers trying to get many dashboards refreshed (e.g., month-end or quarter-end) in a short timeframe. It might be a good idea to also install Publisher on multiple machines and cluster them. If one Publisher goes down, QMC will wait for it to be back to normal before sending data refreshing tasks to it again. In the meantime, QMC keeps sending tasks to the good Publishers and nothing gets interrupted.
A lot of QlikView Servers objects (sessions, shared objects, bookmarks, licensing information, etc.) are stored in a set of system files. When QlikView Servers are clustered, both servers read and write from the same set of system files, which can cause contentions. Some people would use one of the QlikView server’s hard disk as the storage for these system files. This is normally fine except that when that server is down, the good servers cannot work either because the storage location is not there anymore. The best configuration we have seen is to use a SAN owned by a third windows server as the storage, and that windows server can be a VM.
Every component of your QlikView system is properly configured and it looks like it’s ready to go. But wait, how can it be sure the system can support the desired user load and data volume? The only way to find out is to run load tests. QlikTech offers a great tool that is based on JMeter and it is relatively easy to setup. The tool can automatically generate JMeter scripts based on specified testing scenarios, and it can simulate any number of concurrent users. The testing results can be loaded into QlikView for analyzing. Of course, if the company has a strong load testing team that uses other tools like LoadRunner, you can also use your own tool to perform a load test.
These are the most common things people consider in an enterprise deployment. I hope you enjoyed reading it as much as I enjoyed writing it.
If you missed QlikView Enterprise Deployment (Part 1) you can read it here.
About the Author
Lingxiang Cheng is a Senior Consultant for Bardess Group Ltd. He has been working business intelligence and data warehousing for over fifteen years. Lingxiang contributes his expertise to Bardess for solution architecture, development, and training. He has been working as business intelligence consultant and architect for many Fortune 500 companies in retail, pharmaceutical, financial services, insurance, and life sciences.