"Cloud Computing" is a new computing paradigm where a "cloud" of computers collaborate with each other. Many companies incding Amazon, Rackspace, and Salesforce, provide various kinds of clouds today.
Challenging problems in the cloud era are as follows:
- Different amount of resouces or energy consumpution in each machine → How the orchestrate many machines?
- Different kinds of users settle together within one cloud or one machine → How to guarantee the performance or the security.
- Different amount of the load, cost of the energy and the different throughput of the network in each time → How to manage unpredictable changes?
Our group approaching these problems from the layer of fundamental software technologies of cloud. They include operating systems, virtualization, or middlewares, which are mandatory for buidling clouds. Approaching the above three problems from fundamental software layer can benefit both cloud provides and users making little burden to them.
Fast Live VM Migration for VM Placement Optimization (Soramichi Akiyama)
VM placement optimization, which means how to place virtual machines (VMs) in a cloud, is a challenging issue. The optimization can achieve lower energy consumpution in a cloud or performance gurantee of the VMs on a cloud.
Live migration technique is mandatory to change the placement of VMs. Live migration allows VMs to move aound physical hosts without interrupting the execution of the VM. However, live migration takes much time and has great impact on VM performance.
This leads exising research on VM placement to either:
- Stating simulation results that ignore the migration costs, or
- Conservative optimizations that avoid migration costs by limiting the number of migrations to use.
The former case prevents the research from being applied into real world, and the latter case cannot achieve sufficient optimization.
Our research focuses on reduing time cost of live migration. We both reduce the total migration time of datacenter-local migrations within a LAN and cross-datacenter migrations using a WAN. The future directions is to research more efficient VM placement optimization that uses our fast migration techniques aggressively.
- Shigetoshi Yokoyama (Specially Appointed Professor)
- Soramichi Akiyama, Takahiro Hirofuchi, Ryousei Takano, Shinichi Honiden, "Fast Wide Area Live Migration with a Low Overhead Through Page Cache Teleportation", International Symposium on Cluster, Cloud and Grid Computing (CCGrid), pp. 78 - 82, May 2013.
- Soramichi Akiyama, Takahiro Hirofuchi, Ryousei Takano, Shinichi Honiden, "MiyakoDori: A Memory Reusing Mechanism for Dynamic VM Consolidation", International Conference on Cloud Computing (IEEE CLOUD), pp. 606 - 613, June 2012.
- Shigetoshi Yokoyama, Nobukazu Yoshioka, "An Academic Community Cloud Architecture for Science Applications", IEEE/IPSJ International Symposium on Applications and the Internet (SAINT), pp. 108-112, July 2012.