We currently have a relatively virtual large file server, hosting around 20TB of unstructured file data (Windows 2012 R2). The disks are physical RDM, but providing that I convert them to virtual RDM I'd like to know how the CBT & Veeam would improve my backups. Currently it backs up over 1Gb NIC to tape ( ) and this of course (with millions of tiny files) takes days!
If we adopted Veeam, and took advantage of VMware CBT, how much of an improvement could I realistically expect to gain in backup times compared to tapes. Tapes obviously have to troll through the entire data set even if only 10GB of the 20TB has changed since the last backup. Would Veeam still have to process the 20TB of data, or would it actually be able to connect-in and immediately backup the 10GB changed blocks only? And would this result in a backup time of minutes/hours as opposed to days?
If we adopted Veeam, and took advantage of VMware CBT, how much of an improvement could I realistically expect to gain in backup times compared to tapes
Hard to say, but improvements are definitely expected, as only changed blocks would be transferred thanks be to CBT.
I'll have to look at this. We use something called DoubleTake GeoCluster which sits atop of MSCS, and whilst the two servers don't actually share the same physical/virtual disks I'd have to validate whether vmdk's are possible.
BTW - what advantage is there then of vmdk over vRDM?
Me personally, I don't think there is any significant advantage/difference between vRDM and VMDK, regular virtual disks should work fine in all setups.
As the amount of changes, then given that this is a file server, you should cut your backup window dramatically. Do you have volume deduplication enabled for this file server?
No, we haven't enabled any de-duplication on these volumes - we don't have a squeeze on capacity, and I probably have (admittedly unqualified) concerns about 1) deduping larger volumes and more importantly 2) how dedupe'd data might affect backup/restore speeds.
The reason why I have asked this question is that enabling dedupe will generate more dirty blocks that have to be backed up. Since you don't use it, VMWare CBT should narrow down your backup window.
) and this of course (with millions of tiny files) takes days!