RNA Networks intros new memory virtualization application

Print Friendly, PDF & Email

Yesterday from ISC memory virtualization startup RNA Networks announced the release of RNAcache. I met with this company when I was out in Portland earlier this month, and they have a really interesting story. From the release

RNA networks…announced it has extended its Memory Virtualization Platform (MVP) product family with the release of RNAcache. MVP transforms server memory into a shared network resource.  RNAcache allows servers to leverage RNA’s unlimited pool of memory by loading entire working datasets into a single shared, virtual memory pool for simultaneous access and processing. RNAcache significantly lowers the cost of supporting business critical, memory-intensive applications without making further investments in costly data center equipment, enabling companies to wring as much revenue potential out of existing servers as possible.

“We’re seeing a flurry of announcements that attempt to address the critical need for access to more memory. However, many of these approaches require the adoption of proprietary or expensive hardware, or changes to software,” said John Barr, Research Director for Financial Markets at leading analyst firm The 451 Group. “But RNAcache takes the concept of memory virtualization to a new level by making memory a shared resource through software. This adds value to both the IT organization and the business – by delivering great performance increases from unmodified applications.”

The idea here is that you install software on each node of a cluster that has memory you’d like to virtualize for sharing with the rest of the cluster or participate in using memory on other processors. You tell the software how much goes to the local CPU and how much is available for other processors in the cluster, and that’s it. RNAcache handles the rest. It does create a single global address space, but doesn’t take the approach of creating a fully virtualized environment as ScaleMP’s solution does for example.

This approach makes sense if you have a memory-bound, unevenly distributed workload such that you are short memory on some nodes but have spare memory on others. A few examples of early customer testing from the press release

Predictive Analytics and Modeling: improved predictive modeling capabilities by 20X, delivering results to analysts in a fraction of the time.

High Volume, Fast Internet Applications: improved query speed 100X, increased simultaneous user capacity 5X and expanded customer revenue opportunities each business day.


  1. […] company’s RNAcache system allows data sets to access virtual, pooled memory shared across multiple servers. Not only does […]