Memory Virtualization Primer
| Memory Virtualization Primer |
|
This Primer provides background information about memory virtualization and memory pools. To learn about RNA memory virtualization products click RNAcache or RNAmessenger. RNA’s Memory Virtualization PlatformTM (MVP) delivers a new capability for clustered networks by creating a large shared pool of memory that is accessible from any node on the network fabric. This capability fits easily into the current network infrastructure to allow more efficient use of existing resources while taking the network infrastructure to a new level of performance. This section will look at the structure of clustered computing today and how memory virtualization can change the way data center architecture deals with increasing demands for memory and improved latency. BackgroundIn corporate data centers and high performance computing clusters, network resources are used in a variety of ways to optimize for overall application performance and efficiency. Even though many new products are focused on delivering incremental capabilities, the fundamental setup of the compute infrastructure has remained unchanged for a long time. A set of systems is linked together to perform the tasks required for the business or scientific entity. Each system includes a Central Processing Unit (CPU) or limited number of CPUs connected on a system board to DRAM memory. Data storage is provided by hard drives or solid state drives (SSDs) that are either inside the system or shared through a Storage Area Network (SAN) or Network-Attached Storage (NAS) connected via the network.The method of configuring and sharing these resources to match application performance needs has been the basis for many IT industry debates over the years. For older applications that began on mainframes, running on a single system with many processors and large amounts of memory was the only alternative. The proprietary hardware associated with these applications is often extremely expensive, though. Most applications are now designed to run on only one or two processors, which requires a different, “scale-out” infrastructure using a larger quantity of smaller computers. Often, these systems include large amounts of memory that can be up to 3X more expensive than standard memory on a per-GB basis. A number of techniques have arisen to make the best use of available hardware resources. Hard drives have been pooled into SAN for block data and NAS for files. High-speed networks enable applications to more quickly share data across servers to link parts of applications that are run on multiple servers. But the one resource that has remained unchanged has been memory. What is Memory Virtualization?Virtualization is used to describe two distinctly different methods to improve resource deployment efficiency: RNA’s memory virtualization uniquely allows applications to take advantage of this memory pool without requiring any changes to existing applications. With this capability integrated into the network, applications can easily take advantage of a potentially very large amount of memory to improve performance and increase memory usage efficiency. For critical applications with exponentially increasing data set sizes, Memory Virtualization is the only way to keep data delivery out of the critical path for performance. |
Memory Virtualization
RNA enables memory virtualization in the data center. Capital Markets
RNA solutions provide the fastest securities trading and analysis available for trading firms. 
Support
Telephone, online and email support for RNA customers and partners.