In the past, Vembu performed the Hyper-V backup using the Rsync algorithm and during every incremental backup, a complete dump of VM’s would take place and the changes would then be transferred to the server.
This however caused performance and time delays in taking backups. We gathered that this was not the best way for our customers to perform the Hyper-V backup using our product for a wide range of deployment scenarios. So, we have come up with an efficient solution for Hyper-V backups by optimizing the CBT (Changed Blocks Tracking) process.
What is new in Hyper-V backup?
Short answer: Proprietary Changed Blocks Tracking driver
Our Vembu Hyper-V CBT driver is an extended mini-filter driver, which is designed to monitor and track I/O’s to the Virtual Hard Disks residing in the physical host machine. The driver can be subjected to work under high degree of concurrency and each virtual hard disk is tracked through a very light weight process. During the backup, our Vembu BDR application coordinates with the driver in a synchronous manner to make sure that the backups we take are highly consistent.
How does Vembu BDR ensure consistency?
The most important part in tracking is to make sure that neither key I/O operations are missed nor any unneeded I/O operations being tracked. To make this work the tracking process must start at a point when the disk comes to a consistent state.
During a backup schedule the VSS service pushes the disk to a saved state and redirects the writes to the native disk until the snapshot process ends . Vembu BDR makes sure that all the I/Os are committed to the disk after the write quiesce event is captured . This ensures high degree of consistency .Apart from this our agent supports a feature image integrity verification that ensures that data you backup is always consistent.
Vembu BDR Hyper-V Image Backup:
The entire disk content of a Hyper-V virtual machine is represented by a single file called VHD (Virtual Hard Disk). Every disk format has its own way of storing and retrieving data . Basically a disk format should comprise of data blocks and its meta data. If we can formulate a way to read the data blocks and store them along with their hierarchy, we can construct a disk of the format of our choice .This is exactly what Vembu BDR does and thus we are in a position to support restore formats of our choice .
We use Vembu Hive File system to store the backed up data. The unit of storage in a Vembu hive file system is called a chunk. The size of a data block in a chunk is 512k (which is the smallest for a VHD). Thus the size of incrementals are bound to be as less as possible. Efficient storage is guaranteed as our Vembu Hive File system supports features like synthetic merge ,vacuum and deduplication.
