Memory Cache

MVX Memory Cloud
Storage Offload with MVX Memory Cache

Sharing files is one of the oldest and most widely adopted uses of a data center network. Network Attached Storage (NAS) services are at the heart of many applications, from web servers to collaboration systems to document management systems. Fast access to the NAS is critical for these applications. Under heavy load, a NAS server can become overloaded, or its network interface can become backlogged, slowing down applications (even ones that are not the source of the overload). In response, the operating systems in most individual servers will employ a cache in RAM to retain the most recently used NAS files and thereby reduce its own demands on the NAS server -- the larger the cache the better.

RNA MVX has a Memory Cache function that takes this strategy to another level. The RNA Memory Cloud is a large pool of RAM that is shared among any number of servers in the data center. The Memory Cache is a file cache that is usually far larger than any single-server cache, and is shared among many servers using the NAS file system. The result is that a much higher proportion of NAS requests are satisfied at RAM speed, and far fewer requests burden the NAS server itself.

Strategic caching across your data center

As a single central pool of RAM, the MVX Memory Cache provides a virtualized resource that can be deployed to precisely address the performance issues in a data center.

MVX Chart: Reads per Second

MVX Chart: CPU Utilization

Multiple caches can be set up, each sized to fit a specific need:

  • Any number of NAS servers can be cached.
  • Caching can be focused on a specific file directory structure, rather than an entire NAS server.
  • A cache can be very large, even terabytes in size, so that all file requests for a strategic application can be consistently served at RAM speeds. The end-user effect can be dramatic, especially for horizontally scaled web servers.
  • Distributed algorithms, like Hadoop’s map/reduce, may call for the same files to be retrieved at the same time for large numbers of independent servers. MVX can fulfill all the requests without leaving the compute tier -- all requests can be RAM-to-RAM transfers from the cache without waiting for the NAS.

Applied to the critical parts of demanding applications, the effects can be dramatic. Improvements of up to 10x in compute times are typical. Large reductions in excess of 90% in NAS server load can be achieved in many applications. With shorter execution times, CPU resources are more fully utilized and are released sooner for other needs.

Growing with your existing resources

By offloading NAS traffic to MVX, existing NAS servers can get a new lease on life. If the NAS is approaching its performance limit today, MVX may remove enough load to keep up with traffic growth for additional months and years. When the alternative would be the purchase of expensive NAS buffer memory, using the larger MVX memory cloud will yield much better results, while using less costly (and often underutilized) server RAM.

Other alternatives are also enabled by MVX. Slower and less expensive NAS resources may become more usable, especially for read-oriented applications. File-based applications that may scale poorly on a conventional NAS can grow further with a large MVX cache.

Best of all, these improvements are available without risky and slow changes to application programs or to hardware.

With Memory Cache functions from RNA MVX, the data center gains fundamental new capabilities, and new choices for growth, that are impossible without it.