Data Locality

Published: April 2017
Fill in the form to continue


Description:

The IT infrastructure performance depends on I/O speed, so the throughput balance of all the components is very important. Thus, the overall performance of the system is determined by its slowest part. This problem is called “bottleneck”, meaning that even one slow component will bring down an ultra-fast system composed of high-end hardware. At the same time, one high-performing component will not make a difference in an otherwise mediocre system.

The problem is that interconnect fabrics are much slower than high-speed data storage like flash or DRAM, making them the “bottleneck” of a system since no matter how fast the I/O is, the data still must go through the slower wire. This neglects all the benefits of fast storage like ultra-performing all-flash setups and greatly reduces cost-efficiency of the IT infrastructure.

The solution is to process as much of the I/O as possible locally to avoid the slow data transfer through interconnect fabric and achieve the highest performance the setup can provide. This approach is called “data locality” and it is utilized in StarWind appliances for maximum performance by keeping the compute and storage resources for every virtual machine on the same physical node. This way the VM data will rarely have to go through the slow wire and the performance will increase.

Additionally, “data locality” delivers lower latency, since there is no network stack processing overhead, as there is much less data to be transferred.

Related content:

To download the software products, please, make your choice below. An installer link and a license key will be sent to the e-mail address you’ve specified. If you consider StarWind Virtual SAN but are uncertain of the version, please check the following document Free vs. Paid. The recent build of Release Notes. A totally unrestricted NFR (Not For Resale) version of StarWind Virtual SAN is available for certain use cases. Learn more details here.

Request a Callback