Gartner predicts that by 2014, 60 percent of server workloads will be running on virtual machines and that almost half of enterprises will be using a hybrid cloud model by 2017. With cloud adoption becoming a mainstream, businesses have already migrated the ‘easy’ workloads to these environments. What are left are the complex business-critical applications such as high-performance databases or enterprise resource planning. So, how can virtualizing and bringing these applications into a cloud computing architecture make them not only more cost efficient and scalable, but ensure they are also faster and more reliable?
As data volumes grow, the key to both raw performance and delivering higher VM density, to meet the needs of these systems, is to eliminate the bottlenecks between storage and compute. When multiple VMs are waiting for I/O requests to be serviced, as a result of these bottlenecks, overall efficiencies can plunge beyond that which is tolerable for high performance systems.
Deploying higher bandwidth storage networks helps, as does reducing processor interrupts by offloading storage I/O requests to dedicated hardware. Yet neither eliminates the fundamental bottlenecks presented by the conventional storage controllers in the chain.
This year saw the launch of ULLtraDIMM™, a new form of solid state storage, which offers system architects new choices for their infrastructure. ULLtraDIMM™ connects flash storage directly to the system’s memory bus using the DIMM form factor, and delivers ultra-low latency access to storage devices at a cost that compares very favorably to using large amounts of RAM to preload data from conventional storage arrays and reduce swapping.
This move to the memory bus offers ultra-low latency, enabling virtual servers to read and write data faster, and with a determined latency. ULLtraDIMM™ can be scaled linearly for both capacity and performance, which is unique in a storage solution, and thereby delivers a great advantage in a flexible, scalable cloud environment. Ultimately, this means that more virtual machines can be serviced per physical server; cutting data center footprint, power and cooling costs.
With this new option freeing system designers from the constraints imposed by existing issues, such as DRAM pricing large memory systems out of consideration, we can look forward to new evolutions of cloud technology that provide greater VM density, scalable performance, more transactions per second and greater performance predictability. All of which add up to more assured performance of demanding applications in the cloud.
SanDisk® at CloudExpo Asia
SanDisk® is a sponsor of Cloud Expo Asia in Singapore, November 13th & 14th, 2013. The event set to be the leading event in Asia for senior IT decision makers in client side and public sector organizations, as well as the cloud industry, and is co-located with Cloud Expo Asia is Data Center World Singapore and the CloudAsia Conference organized by the IDA.
Visit us in Booth #J09
Join our session:
Enabling high-level performance with lower-cost solid-state storage
Virtualisation, Infrastructure, Platform & Applications Theatre
Date: Nov. 13, 2013
Time: 12:40 – 1:05 p.m.
Speaker: Esther Spanjer