Abstract / Description of output
Flexiscale, is a next generation data centre solution. Based on virtualisation technology, the services are intended to be location and hardware transparent. It currently employs Virtual Iron as the virtualisation management platform, and Xen as the virtualisation platform.
In this talk, we will describe the current architecture and some of the problems that we encountered with Phase I of the project, which is already in production use. We will then discuss the second phase of the project which is currently under research and development; this is intended to allow data centres to automatically cater for usage fluctuation with a guaranteed QoS level for underlying services.
Using various algorithms for load prediction, hotspot detection, and migration, we hope to guarantees the maximum utilisation of the data centre hardware without breaking the QoS levels -- the current prototype of phase II is able to add and remove virtual servers for a specific service bound by QoS levels. Within an 18 minute period of fluctuating usage, the prototype system is able to scale up the service on to 6 new virtual servers and load balance them. When the usage goes down the service is scaled down by removing the extra virtual servers and only keeping only the number of virtual servers to satisfy the QoS level.
In this talk, we will describe the current architecture and some of the problems that we encountered with Phase I of the project, which is already in production use. We will then discuss the second phase of the project which is currently under research and development; this is intended to allow data centres to automatically cater for usage fluctuation with a guaranteed QoS level for underlying services.
Using various algorithms for load prediction, hotspot detection, and migration, we hope to guarantees the maximum utilisation of the data centre hardware without breaking the QoS levels -- the current prototype of phase II is able to add and remove virtual servers for a specific service bound by QoS levels. Within an 18 minute period of fluctuating usage, the prototype system is able to scale up the service on to 6 new virtual servers and load balance them. When the usage goes down the service is scaled down by removing the extra virtual servers and only keeping only the number of virtual servers to satisfy the QoS level.
| Original language | English |
|---|---|
| Publication status | Published - 2008 |