Protecting Data in Virtualized Environments
IT budgets are shrinking, datacenter space is becoming limited, power requirements are skyrocketing, and server and storage needs are ever increasing. Today’s IT managers from the small SMB to the large enterprise are struggling to meet their company’s needs. Server consolidation is fast becoming one of the primary tools being used to address this new paradigm. This is driving the deployment of virtualization across enterprises of all sizes. More and more companies are turning to solutions like VMWare, Microsoft Hyper-V, Citrix Xen, and Parallels. However, to fully realize the benefits of virtualization, organizations must also consider their information recovery management strategy. Traditional backup and recovery strategies are not adequate to deliver the granular recovery demanded by the business. More important, the cost associated with traditional or agent based technologies essentially negates many of the cost advantages of virtualization.
As well, an often forgotten part of the equation in implementing virtual server environments is reining in the ensuing costs and complexity that virtualization introduces. Nowhere do these management costs and complexity become more evident than when a company goes to protect its guest VMs.
A specific challenge emerges when protecting the data of guest VMs. Each physical virtualization server may host ten or more guest VMs that individually host different operating systems and applications. Data protection software now needs to account for the following variables when protecting guest VMs on the same physical virtualization server:
Discovering when new VMs are created: The ease in creating guest VMs can result in lapses in data protection since companies can forget to configure data protection software to protect new guest VMs as they are created.
Agent installation: Each guest VM acts and functions like a normal server. To protect each guest VM, companies may need to install and configure an agent on each guest VM just like when they were protecting individual servers.
Scheduling backup jobs: As servers are consolidated and virtualized on the same physical ESX server, scheduling backup jobs across the different guest VMs becomes more challenging. Schedule too many backup jobs at the same time and the jobs compete for the server’s limited resources. Schedule backup jobs too far apart and they may not complete before the next day’s production activities begin.
Different backup products: Guest VMs on the same physical machine may use different backup software due to different operating systems and/or applications on each one. This adds to the complexity of managing and scheduling backup jobs across guest VMs on the same physical server.
From an information recovery perspective, the trend towards virtualization has a major impact on information recovery management:
- Applications run isolated and utilize the hardware more efficiently. Before virtualization, any backup activity could likely afford to take away some hardware resources from the underutilized application server (CPU power, RAM); with virtualization this may add up to too much. A small 5% resource utilization for a backup agent on a physical machine may go unnoticed, however in a Virtualization Server running 10 O/S instances, this will add up to 50% resource utilization (on top of the fact that the virtualization server is not under-utilized anymore, as the physical machine was).
- Any hardware connection will likely be abstracted away. Data protection software cannot rely anymore on the physical location of data remaining fixed, as storage resources may be migrated at any time by IT administrators to improve processing efficiency.
- The introduction of console-less (COS-less) virtualization servers prevent backup applications from running on the virtualization host operating system.
- Most applications are expected to be available as a “virtual appliance,” isolated from other applications. This applies to backup applications as well, where an ability to deploy a backup appliance on both physical and virtualized servers, as well as centralized deployment without any agents, is required.
In part 2 of this article, I will show how Asigra addresses the needs and concerns raised above through its agentless remote backup and recovery architecture.