Before I start this post I will start with a bit of a caveat, I don’t want this post to turn into a slagging match between backup vendors. The contents of this post is simply what I have found from my tests and experience. If you believe your backup product could easily match this functionality or better it please email me on barry(a)virtualisedreality.com with the information and I will happily test and blog my findings.
I deal quite extensively with small and medium businesses and education from companies with 30-50 customers up to customers with over a thousand users. These customers usually have a lot in common when it comes to backups, the first is that they can’t afford to spend huge budgets on backups after virtualising their environments, second is what ever solution is delivered for backups must be reliable and easy to use and finally it must be able to complete within the backup window.
For the larger SMB’s and enterprises this can often come in the form of offsite SAN replication and archival and other such non budget friendly solutions. For most SMB’s this normally means we are having to reuse existing hardware or cheap NAS / Server storage for backup staging and archival.
Although these customer may only have smaller budgets and numbers of employees it often doesn’t equate to the size of their data, with normally several TB of VM’s that need backing up on a regular basis. I have in fact at times found these smaller customers to have larger data usage due to less data management being in place compared to bigger customers. With working with these kind of limitations you can often have problems writing the backup to disk quick enough to fit within the backup window, especially where full backups are involved.
The Traditional Solution
The traditional solution in this situation would be to run a full backup at the beginning of the week and follow this up with incremental backups for the rest of the week. This solution works but introduces a number of problems, the first being if your weekly full fails for what ever reason, you are then left chasing trying to get a full completed during the working week. The second is with needing to hold a full backup online each week, it makes keeping several weeks of backups online storage hungry and costly. Obviously you could archive to tape daily or weekly but this adds additional complexity and slows down the ability to be able to quickly restore VM’s and data.
For me this is where Veeam steps in and meets the requirements, first of all Veeam Backup and Replication’s user interface is so easy to use and makes the job of configuring, monitoring and restoring your backups a breeze. One of the main points for me that makes Veeam the solution to have in these requirements is their synthetic full method.
Taken from the Veeam Backup and Replication User Guide http://www.veeam.com/veeam_backup_4_1_user_guide_pg.pdf
Synthetic backup presupposes that a full backup is performed only once. All subsequent backups are incremental: only data changed relatively the most recent version is backed up. In contrast to repeatedly performing full backups, this approach ensures a faster and less space-intensive backup.
When a full backup is performed, the resulting backup .vbk file is written to the target host. At each incremental backup, Veeam Backup & Replication 4.1 rebuilds this full backup to the most recent state of a VM and uses historical data to calculate a reverse increment.
Additionally Veeam Backup and Replication supports de-duplication and compression of backed up VM’s, instant granular recovery of VM files and Guest files. They give you the ability to take a standard full backup when required, for example if company policy insists a full is taken weekly / monthly etc. Finally they also give you the ability to be able to replicate backups to another virtual environment using the same system for no more money! This is allowing us to create a VMware SRM like environment for small customers on a tight budget! (That’s a blog post for another day!)
So when you are in a situation where you need to be backing up to low performing storage and / or if you are tight on storage space but still wish to keep as many backups online as possible Veeam fits the bill.
The more I read this post the more it sounds like an advert for Veeam, but I can assure you this comes from personal experience and no kind of connection with Veeam.
For me the only point where Veeam falls down is when you start archiving your jobs to tape. Veeam creates a single VBK file for the synthetic full backups, this makes it easy to create an archival job to write your backups onto tape, where you start to have a bit of difficulty is when you need to restore an individual VM or File from that tape archive. You will need to restore the whole VBK from tape first before recovering your VM or Files with Veeam. Obviously there are ways around this such as create individual jobs for each VM, but the main issue with this would be loss of de-duplication between VM’s and the amount of jobs you are then left to manage. Luckily because the amount of recovery points we are able to keep online with Veeam this situation very rarely arises. There is a conversation about tape archiving that can be found on the Veeam forum here >> http://www.veeam.com/forums/viewtopic.php?f=2&t=1441&start=0&hilit=tape+support. Veeam are also making some changes in V5 that will make backups more archive friendly.
Often we will need to re-use old hardware or cheaper server based storage for the backups. It can be very easy to start running into performance issues when this is the case but sometimes they can be easily resolved.
Please check out my blog post from last year about Virtualised Backup Performance >> https://virtualisedreality.com/2009/05/21/virtualised-backup-performance/
Recently I ran into some issues using a HP DL180 G6, filled with 750Gb Midline SAS disks in a RAID5. We were getting dreadful performance whilst backing up with Veeam, with a backup of 3Tb’s of virtual machines with 100Gb compressed and de-duplicated changes taking over 30hours to complete. After some investigation it was discovered the servers P400 RAID controller although containing 256Mb cache ships with no battery as standard so the cache is disabled. You are able to override this configuration (not recommended) but as standard the cache is configured for 100% reads. So after upgrading the RAID controller to BBWC, tweaking the cache configuration for 75% writes and upgrading the firmware on all the disks (I highly recommend this on the midline SAS disks) we are now able to complete the same backup in well under 3 hours!
So just goes to show there is no reason cheaper hardware with the right kind of backup solutions can’t do the job, but ensure you know the limitations of your hardware and ensure everything is fully up to date before starting. Too often the backup vendors end up being the point of blame for poor performing backups and I have been guilty of this, but usually you will find a bottleneck else where that will be causing the issues.
All of the major backup vendors offer a free trial period so give them a go and come to your own conclusions.