Browse Prior Art Database

Method and System for fast and scalable Virtual Machine (VM) Provisioning

IP.com Disclosure Number: IPCOM000216427D
Publication Date: 2012-Apr-05
Document File: 3 page(s) / 22K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method and system is disclosed for provisioning virtual machines (VMs) by sharing chunks of one or more VM images among one or more physical hosts in a topology-aware manner.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Method and System for fast and scalable Virtual Machine (VM) Provisioning

Disclosed is a method and system for provisioning virtual machines (VMs) by sharing chunks of one or more VM images among one or more physical hosts in a topology-aware manner. Sharing chunks of one or more VM images may be faster and cost effective as compared to transferring an entire VM image over a network. Further, a topology-aware design ensures that chunks of one or more VM images are shared among closer physical hosts (hosts that are at the same level in the hierarchical topology of data centers) to utilize a higher bandwidth between them.

In an exemplary implementation, the method and system disclosed may include the steps of dividing each VM image into one or more chunks, distributing required chunks of VM images to one or more physical host, and updating changes in chunk information resident on a physical host.

Initially, each virtual image file is divided into one or more fixed-sized or variable-sized chunks. Subsequently, each chunk is identified by hashing its content using a hash function such as, for example, SHA-1. This allows easier identification of identical chunks from multiple VM image files. In an exemplary instance, hash collision detection techniques may be employed to avoid cases where chunks with different content are mapped to the same hash value. This design feature exploits the similarity among distinct VM images and enables sharing of image chunks across distinct VM image files.

In a particular implementation, a local cache mechanism is adopted in each physical host. This host-level cache resides below a hypervisor software module and therefore its content is exposed to all VM instances running on the host. The local cache may provide storage of one or more of, but not limited to, VM image chunks for alive VM instances, and chunks that previously resided in the host and are likely to be requested in the future. A number of cache policies may be utilized such as, for example, live-cache (where VM chunks are only available for those running VM instances, and chunk data is garbage collected from the cache as soon as no running VM instance is

1


Page 02 of 3

referencing it), and fixed-cache (where a fixed size of disk storage is assigned as cache and FIFO policy is adopted to mange VM chunks).

In an instance, trackers (index nodes) are designated within each hierarchy in a physical domain. Lowest level trackers are selected for each rack in the hierarchy while higher-level track...