Abstract:

The development of IaaS enables cloud service to fast deploy largescale virtual machines (VMs). However, implementing version control in VM clusters is in low efficiency. Current version control solutions have some problems like network transmission overload and slow operation speed. We present a novel version control approach for VM clusters, called FlatVC. FlatVC creates VM version incrementally on compute nodes to avoid transmitting version data to persistent storage, and downloads data blocks on demand during the VM restoring process, thus reducing network transmission overhead and speeding up version control process. By using cache tree structure to share the data chunks transmitted over the network, FlatVC reduces the data transmission pressure of root nodes. Besides, we make I/O optimization to avoid performance degradation caused by the version chain, which is constructed by incremental VM versions. Experimental results show that FlatVC can efficiently enforce version control for VM clusters, and speed up both version creating and restoring process.

Key words: cloud computing;IaaS;VM clusters;version control