Storage Performance and Cloud Enablement

Adding Users Without Rendering Delays

Posted by Jim Thompson on Fri, Feb 01, 2013 @ 09:39 AM

Adding Users Without Rendering Delays

Image Engine boosts performance in a tight storage footprint and positions network storage operations for big-time summer movies. 

Zero Dark ThirtyIn Friday's interview on Bloomberg West, Bloomberg host Cory Johnson talked about how Avere’s NAS optimization helped Image Engine produce dust storms without a speck of actual dust. With the FXT Edge filer’s ability to handle the transactions at a 50 to 1 offload of the core storage, renders are done more quickly and costs are driven down.

To keep up, both staff and rendering capacity needed to grow, but physical data center space was nearly maxed out. Image Engine needed something to enhance the performance of their existing storage without increasing the footprint. An expansion of their render farm and the addition of one-hundred more artists utilizing demanding applications such as Maya, Nuke, Houdini, 3delight, and Arnold. 

It's a challenging feat to increase the performance of a stable platform that has proven to be a reliable servant for the past few years without going through the pains of a “rip and replace” exercise. Enter Avere. We enabled Image Engine to increase performance AND use the largest capacity drives their storage vendor could provide. This freed valuable data center space that could now be used to further expand their render farm or allow for even more much-needed storage.

Initially, Avere installed a couple of our 3200's but quickly found that Image Engine’s data set was larger than estimated, so the nodes were swapped-out for our larger capacity 3500's which have 9TB of SAS and 144GB of DRAM with 2GB of NVRAM. Once the edge filers were able to hold their entire data set, Avere was able to keep the Image Engine’s latency in the 400-microsecond range.

Image Engine hovers around 4 billion operations a day and, as their engineer, I have seen them get to 11 billion several times. Operations consistently run near 200,000 op/sec. This is with latency of a millisecond or less and the load on the NAS system at around 25,000 ops. Their previous infrastructure would have fallen flat right there, but with the Avere scale-out NAS optimization/abstraction layer they can continue to separately grow their performance and their capacity.

With a full production schedule far into the future Image Engine is now positioned to meet their growth in the visual effects industry. The additional performance boosts and ability to add users without fear of the effects on the storage infrastructure may be a blockbuster of its own. 

 

Looking to achieve better performance in limited physical space? Here’s a white paper to get you started.

Topics: NAS Optimization, Data Center Management, WAN Latency, Rendering & Transcoding