2010 Prediction: Clive Cook, RNA networks
Memory Virtualization Heats Up in 2010
In 2009, we saw a number of vendors like Cisco (UCS) and HP (Matrix) focus on the problem of memory limitations in the data center. In 2010, we believe that memory virtualization will be as widely accepted in the data center as server virtualization, desktop virtualization and storage virtualization. Therefore, attention to the age-old problem of memory limitations will only heat up and memory virtualization will be one of the top technologies to watch in the coming year. Three specific trends are fueling this growth:
Memory is 100X Faster Than Storage
Storage is often looked at as a way to address I/O bottlenecks and application performance limitations, but this approach does not address the core issue, limited memory capacity. Storage acceleration can only marginally improve application performance as it connects too far down the stack and is not ‘application-aware’. In contrast, memory operates at the application layer and is able to more easily and quickly interact with active data - providing dramatic increases in utilization, performance and speed. Regardless of the type of storage, (spinning or SSD), or the acceleration options (NFS, storage caching), memory is up to 100 X faster than storage.
Data Proliferation is Putting a Strain on the Data Center
Data volumes continue to outpace available memory on a server. The rate of proliferation and amount of data in the data center that must be quickly analyzed and interpreted is growing exponentially. Methods to make data available to distributed servers are expensive both in terms of bandwidth, IT oversight and face scaling issues. Caching 2.0, a new caching model that reduces replication, supports a broad range of data types and scales is now available. Alternatives such as adding more database nodes, expensive storage or faster processors can’t keep up with the increase in data volume. To keep data centers running effectively and efficiently, IT must look for long term answers beyond these alternatives that will dynamically scale as the volume of data in the enterprise continues to increase. By enabling TB’s of active data to be kept in memory while leveraging existing infrastructure, IT can keep ahead of the coming data wave.
Cloud Computing's Missing Link: Don't Forget Memory!
The success of a cloud provider hinges on its ability to ensure SLAs, and keep costs down by harnessing as much power from the infrastructure as possible. Tools to provision servers, storage, and I/O bandwidth are common in the cloud. However, memory is a core resource that previously was not separable or sharable apart from the physical server it’s sitting in. Memory virtualization breaks that fixed relationship between memory and server giving cloud providers far more flexibility and scale. Pools of networked memory are fundamental requirements for clusters, and will be a key component of the cloud infrastructure.
Conclusion
Addressing today's IT performance challenges, virtualized memory enables new business computing scenarios by eliminating application bottlenecks associated with memory and data sharing. Memory virtualization delivers optimized data center utilization, performance and reliability with minimum risk and immediate business results.
