Alignment issues have been around since the first PC’s. In an environment with a single server locally attached to a few disks alignment was seldom a noticeable issue. In a shared SAN with multiple servers accessing the same disk array it’s become more of an issue. In a virtual environment with potentially hundreds of virtual server machines accessing the same disk device it’s a storage I/O nightmare, especially if the hypervisor’s file system that’s going to house all of the virtual machine file systems is also misaligned.


Re-aligning all of the virtualization file system and all the VM file systems can dramatically improve performance. Yet most data centers that are in production with server virtualization have chosen to live with the performance loss instead of fixing it. The reason is that it’s typically a very time consuming, manual process, one that can also destroy data.


First, the misaligned VMs and their data volumes need to be identified. Without a tool this requires a manual VM-by-VM inspection. Each VM needs to be logged into and that VM's disk partitioning tool must be run, typically a command line experience. Once the affected VMs are identified, time must be scheduled to realign them. Again, without a tool in place, this means backing up the partition to be aligned, downing the VM and creating a new properly aligned partition with the correct starting cylinder. Then drive letters need to be reassigned from within the OS, correctly, so that applications will see the same storage environment as they did before. Finally, data must then be restored from the backup of the old partition. A snapshot will not work in this scenario since the partition is destroyed, and with it, all the relevant information to rebuild from a snapshot.


While performance may double, or more, the risk and time involved are too great for most organizations to undertake this complex process. Instead, as stated earlier, they suffer through it. Some vendors have begun to develop scripts for their arrays to fix the alignment issue in a more automated way. However, they typically address only one VM at a time and the VM must be put into a ‘ready for partitioning’ state prior to the script executing. This means that the VM must be powered off, the script must be executed and then the VM must be manually restarted; still time consuming, but an improvement over the prior, totally manual process.


Companies like Vizioncore are taking the alignment issue a step further with the vOptimizer product. Products like this allow the entire process to be automated and the re-partitioning done ’in-place’, meaning that no data needs to be manually recovered. It is still advisable to make sure a solid backup of the environment is run before any re-partitioning is done. After that, the environment can be scanned to identify the VMs with partitions that are not in alignment to create a work list. While the virtual machines do need to be powered down during the re-partitioning process, products like this can be preloaded with a list of virtual machines to realign and then work through this list automatically VM by VM. They will stop the VM, realign the partition and then restart the VM. This means that an IT administrator does not need to manually do those steps, which were often done at 4:00AM or on a weekend. Instead, the list is created, the software processes through that list and the IT administrator can check the progress as the VMs are realigned.


Proper alignment is critical for maximum storage I/O efficiency, something that’s especially true in virtual environments. In the future hypervisors and operating systems may automatically align volumes correctly, but for now it must be done by an IT professional. The good news is that solutions now exist, like those from Vizioncore, to make this process far less painful than in the past. While it does require a maintenance window it can be done automatically and without the risk of human error, making the investment in time minimal and the payoff great.

Vizioncore is a client of Storage Switzerland

George Crump, Senior Analyst