Largest Game Developer
Render Farm Optimization in a Gaming Industry
PRODUCTIVITY IN RENDER FARMS
CGI companies deploy CPUs in clusters or grids, going up to several 100s, in order to provide the computing power required to render computer generated imagery (CGI). Even with sophisticated computing, producing realistic images is a time consuming and compute intensive activity. Performance is nearly always a concern in render farm settings. There are other concerns driven by industry trends - data sets becoming larger, jobs becoming more complex, more collaboration in the projects than before, delivery deadlines becoming tighter and of course, budget spending becoming smaller. Small wonder then, that there has been a demand to make the render farms more efficient and productive.
THE COMPANY'S PROBLEM
The Company was a premier developer and publisher of entertainment software that has quickly established itself as one of the most popular and well-respected makers of computer games. Like most gaming houses, they had a very demanding IT environment. In the company, animators worked on different frames and scenes and hundreds of jobs were deployed across the different servers in the grid cluster. Jobs typically serviced the data, from NAS mounts. With several hundred jobs writing and reading data from the network storage, soon multiple nodes accessed the same set of shares repeatedly. Unable to deal with the I/O stress, the back-end NAS became a bottleneck and applications started slowing down.
Although the company had a mature IT environment, they were missing vital intelligence regarding the use of data by applications. Not all applications were as data intensive as the others. Some applications were more critical and some less so. There were some files that were frequently used by most applications and some weren’t. The company needed such a file level understanding of the dynamics of active data, for them to be able to optimize data management around their workflows, so that they got the speeds and the consistent performance they required.
THE DATAGRES PERFACCEL SOLUTION
PerfAccel gave administrators in the company, for the first time, a micro and macro level reporting of all I/O in the grid. Administrators could identify frequently accessed scene files. This visibility as well as rich analytics enabled optimal placement of data across the grid.
Administrators could now fine tune their caching rules, so that the frequently accessed files could be held inside the cache and made easily available to the application. This not only reduced I/O latency but also helped them to streamline their workflows.
VALUE TO THE COMPANY
PerfAccel yielded the following gains for the company:
- Reduce existing NAS workload by up to 80%
- Increase server I/O performance
- Reduce the need for additional hardware
PerfAccel brought download on the back-end NAS, improving the TCO and extending life of expensive hardware.
DATAGRES’ PerfAccel provided administrators with a single pane which combined analytics and insight through performance dashboards, as well as a simple command line, for running commands i.e. creation of cache/source, deletion of cache/source, adjusting sizes and so on.
Its flexible interface let users configure their own policies of persistent cache, pre-fetching, predictive cache, real-time cache size configuration and auto-caching hundreds of NFS mount points.
PerfAccel Commands were easy to use and an administrator could learn them in a few minutes. The company's system administrators were able to learn the commands easily.