Server Virtualization is Turning the Data Protection Model Upside Down
Organizations tend to give insufficient thought to the protection of an application before it is deployed. Too often it is only after the application is developed or purchased and put into production that the organization takes the time to consider the protection of the application's data and, even then, it is usually not a major problem to implement. As organizations look to virtualize more of their physical machines that are hosting these applications, waiting until the application is in production before deciding how to best protect it creates new sets of problems.
Over the last decade or so, most organizations have adopted the approach of assigning every application its own server. This was done to expedite the deployment of applications as well as keep administrative costs low since organizations found one administrator could cost-effectively manage tens or even hundreds of application servers and problems could be dealt with after the fact.
Adopting this approach resulted in applications only using a small percentage of each host server's resources (CPU, memory, network bandwidth and storage). This left ample resources on the server for it to run other applications like performance monitoring and backup software. In the case of backup jobs, since they ran in the evening and/or during off-hours, organizations just assumed that there would be sufficient processing power and network bandwidth available and that backup problems would not surface.
According to Mike Sparkes, Quantum's Product Marketing Manager, the problem that starts to surface is that as companies consolidate and virtualize servers each physical machine will host multiple virtual machines, each with its own application. As a result, the resources of the physical machine are now shared among the numerous virtual machines and the applications that it hosts. Sparkes says, "Virtualization can make better use of server and storage resources but since backup runs on every virtual machine, virtualization can be disruptive to existing data protection processes."
Organizations will tend to analyze the mix of existing and/or new application servers before they consolidate and virtualize them on a single physical machine. However they can forget to consider the load that the backup jobs that run on each of these virtual machines introduces on the underlying physical machine.
So as each virtual machine's backup job executes, typically during the same time window, it competes for that physical machine's resources. This can overload the server's CPU, congest that server's network connections and result in failed backup jobs. Worse, while the backup jobs are running, it can impact applications executing on other virtual machines running on that same physical server.
Of course, organizations can monitor each backup job on each virtual machine and schedule them so the jobs occur sequentially. This minimizes the impact on the underlying physical server by giving each backup job access to more of the physical server's resources. But, again, the problem with this approach is that it requires administrators to schedule and constantly monitor each backup job on each virtual machine, verify the backup job completes within its designated backup window and verify all of the backup jobs complete before the next day's production activities begin again. The only way to effectively do this is do it in real-time which is not a viable or practical option in most environments.
Server virtualization is turning the traditional data protection model upside down as it can no longer remain an afterthought in organizations. Rather organizations now need to put as much thought into planning data protection in a virtualized environment as they do planning for the virtualization of the applications.
Organizations are now changing how they bring application servers into their environments by first looking to virtualize them instead of dedicating a server to host each application. But as they do so, organizations need to make a similar change in approach in how they approach the protection of data in their new virtual environments.
Data protection can no longer remain an afterthought in virtualized environments. Installing a backup agent on each virtual machine and then protecting data like they have in the past is no longer a reasonable or even a viable solution in many circumstances. Rather organizations need to make planning for the protection of data as much a part of the planning for virtualizing an environment as planning to virtualize the application servers. To do so, they need to look at new ways to automate data protection in their virtual environment by keeping solutions simple and cost-effective to deploy, configure and manage.
The good news is that the resulting consolidation of resources opens the door to consolidated backup solutions with new approaches to reducing space, power, cooling and cost. In an upcoming blog entry, I'll examine some ways that organizations can deliver on those objectives as well as some of the pitfalls that they can encounter as they do so.
Over the last decade or so, most organizations have adopted the approach of assigning every application its own server. This was done to expedite the deployment of applications as well as keep administrative costs low since organizations found one administrator could cost-effectively manage tens or even hundreds of application servers and problems could be dealt with after the fact.
Adopting this approach resulted in applications only using a small percentage of each host server's resources (CPU, memory, network bandwidth and storage). This left ample resources on the server for it to run other applications like performance monitoring and backup software. In the case of backup jobs, since they ran in the evening and/or during off-hours, organizations just assumed that there would be sufficient processing power and network bandwidth available and that backup problems would not surface.
According to Mike Sparkes, Quantum's Product Marketing Manager, the problem that starts to surface is that as companies consolidate and virtualize servers each physical machine will host multiple virtual machines, each with its own application. As a result, the resources of the physical machine are now shared among the numerous virtual machines and the applications that it hosts. Sparkes says, "Virtualization can make better use of server and storage resources but since backup runs on every virtual machine, virtualization can be disruptive to existing data protection processes."
Organizations will tend to analyze the mix of existing and/or new application servers before they consolidate and virtualize them on a single physical machine. However they can forget to consider the load that the backup jobs that run on each of these virtual machines introduces on the underlying physical machine.
So as each virtual machine's backup job executes, typically during the same time window, it competes for that physical machine's resources. This can overload the server's CPU, congest that server's network connections and result in failed backup jobs. Worse, while the backup jobs are running, it can impact applications executing on other virtual machines running on that same physical server.
Of course, organizations can monitor each backup job on each virtual machine and schedule them so the jobs occur sequentially. This minimizes the impact on the underlying physical server by giving each backup job access to more of the physical server's resources. But, again, the problem with this approach is that it requires administrators to schedule and constantly monitor each backup job on each virtual machine, verify the backup job completes within its designated backup window and verify all of the backup jobs complete before the next day's production activities begin again. The only way to effectively do this is do it in real-time which is not a viable or practical option in most environments.
Server virtualization is turning the traditional data protection model upside down as it can no longer remain an afterthought in organizations. Rather organizations now need to put as much thought into planning data protection in a virtualized environment as they do planning for the virtualization of the applications.
Organizations are now changing how they bring application servers into their environments by first looking to virtualize them instead of dedicating a server to host each application. But as they do so, organizations need to make a similar change in approach in how they approach the protection of data in their new virtual environments.
Data protection can no longer remain an afterthought in virtualized environments. Installing a backup agent on each virtual machine and then protecting data like they have in the past is no longer a reasonable or even a viable solution in many circumstances. Rather organizations need to make planning for the protection of data as much a part of the planning for virtualizing an environment as planning to virtualize the application servers. To do so, they need to look at new ways to automate data protection in their virtual environment by keeping solutions simple and cost-effective to deploy, configure and manage.
The good news is that the resulting consolidation of resources opens the door to consolidated backup solutions with new approaches to reducing space, power, cooling and cost. In an upcoming blog entry, I'll examine some ways that organizations can deliver on those objectives as well as some of the pitfalls that they can encounter as they do so.
Leave a comment