Companies have a love-hate relationship with VMware. Most companies know that most of the resources on their existing Linux and Windows servers are underutilized by the applications running on them. So the idea that they can consolidate multiple physical servers (typically 10 – 30 servers) down to just one appeals to them on multiple fronts. Lower energy costs, lower administrative costs and less data center floor space are just some of the benefits that they will likely experience.
The problems start when companies begin to manage applications running on individual virtual machines (VMs). No longer do applications have access to an entire server’s resources such as CPU, memory and bandwidth when they running. Rather the VMware hypervisor must allocate specific server hardware resources to each guest operating system and the applications running on them. To appropriately manage these environments, server administrators therefore need a higher level of aptitude so that when they virtualize 10 – 20 servers on the same physical hardware they do not create application performance bottlenecks.
Companies need to decide what types of data need to be protected. All backup methods do not protect all data and file types. Companies need to consider if they want to protect VMs, files, ESX and VMDKs and if bare metal restore (BMR) is a requirement. Companies have the option of using different backup methods to ensure total protection or to opt for an integrated data management system.
Also problematic in VMware environments is when each VM’s backup process kicks off in the middle of the night. Now each backup software agent on each VM is consuming server resources (CPU, memory and network bandwidth) plus it is accessing storage resources assigned to that server. At this point, the VMware hypervisor kicks in and limits how much available server resources each VM can have according to preset policies. Since multiple backup jobs may kick off at the same time or kick off while other backup jobs are still running, the problem is exacerbated
The impact of restricting server resources on each VM on the physical server is difficult to predict. However problematic in VMware environments is when each VM’s backup process kicks off in the middle of the night. Now each backup software agent on each VM is consuming server resources (CPU, memory and network bandwidth) plus it is accessing storage resources assigned to that server In a worst case scenario, it can even impact mission critical production applications on the VMware server because the application can not access needed server and storage resources because backups are kicking off on other VMs on that particular server.
What companies are coming to realize is that introducing VMware into their environment needs to change their entire paradigm of how they manage servers – from the applications running on them to the data they protect. In the case of data protection, the change is even more extreme. Enterprise companies can not and should not expect their existing version of backup to work well in this new virtual world as it was designed to work from a totally different premise.
This new data protection paradigm is what Asigra’s Televaulting is designed to address. Rather than placing backup software agents on each VM, Asigra’s agent-less architecture remotely accesses the data on each VM and backs up VMs, files, ESX and VMDKs using a daily incremental backup or via CDP (continuous data protection) while providing for bare metal restore if needed. How it does this with minimal impact to each VM and the underlying physical server is the focus of the next blog entry.