This optional feature added to the BlueArc cluster offerings introduces a way to help customers optimize workflow related data access and processing capabilities. Dynamic Read Caching is a high performance option designed to scale NFS read-based workload profiles for client users and applications that need shared accessibility and scalable performance into a common shared data set.

The read caching approach dynamically and transparently distributes and caches where needed one or more designated data sets across individual Titan and Mercury Servers within a cluster. The option has intelligent policies that automatically distribute read-only data sets based on historical client and application access patterns. This feature allows clustered Titan and Mercury Servers to automatically distribute customer selected read-only data sets while retaining the level of coherency needed to create a dramatic impact on application speed and productivity.

• Dynamically caches read data across a cluster
• Accelerates read intensive application performance
• Allows read-only performance to scale by up to 400%
• Scalable and parallel concurrent access for read-only data
• Fully coherent and automated caching policies
• Distributed cached copies allow linear scalability
• No additional hardware required

Customers with read intensive workload profiles and the ability to stage data in an optimized workflow process can leverage read caching as a way to scale performance when and how they need it. In the entertainment industry this translates to a common library of files that is centrally accessed and expands performance on-demand as simultaneous clients and workstations scale in both number and performance for rendering operations. In the internet services and retail markets, where distributed applications require shared access to a common read-only database or data resources, Dynamic Read Caching helps you scale when both application use and customer workload increase unexpectedly. In Life Sciences markets, critical research data like large libraries of thousands of small genomic sequences can be scaled to match the I/O needs of ever expanding compute clusters and compute grids in the high performance and technical computing space. Wherever read-only storage systems are hitting hard limitations with performance, scalable and sustainable client/server access; Dynamic Read Caching can help to achieve new levels of optimization and speed time to results.

Dynamic Read Caching was designed to eliminate performance bottlenecks and network connectivity limitations associated with read-only data sets that are typically constrained by the underlying limitations of the physical platform architecture and/or file system in which they are being stored. This new software feature allows critical data sets to be copied, cached and stored across multiple nodes within a Titan or Mercury cluster. Policy driven and fully automated, the read caching functionality dynamically and transparently monitors file access patterns and caches only those file necessary to satisfy individual client and application request received by each individual server in the cluster. A new synchronous and coherent copy of a common read-only data set can be added with each new node added to the cluster to increase performance when needed the most. This allows throughput in both IOPS and Bandwidth to be increased, improved latency of access and most significantly a continuous scaling of clients and application servers that all need shared access to a common set of data.

Dynamic Read Caching is an optional option. No additional hardware is required and anywhere from 2 up to 4 nodes can participate in the read caching functionality. Currently Dynamic Read Caching only supports NFS v2 and NFS v3 network file system protocols. Customers can control automated or manual distribution of designated file systems for read-only caching. Up to 500,000 files can be designated within a read caching file system.