One of the key jobs the Operations team in general, and myself in particular, undertake is the managing of the data protection systems that we use, to well, protect data. Essentially this entails having copies of data in locations other than the live ones in case of catastrophic problems. It also means that data can often be recovered in the case of less world ending scenarios, such as finger trouble with the delete button.
We use a number of systems to protect data, the simplest of these being the backup tape nightly of data, with a full once a week and just the changes backed up on other nights. This approach is fine for 90% of the systems, but some, such as email, are more important and require a different approach.
For around 18 months we’ve been using a Microsoft product, Data Protection Manager (DPM), to protect this data as well as some central file servers and MSSQL databases. The theory behind DPM is that a replica of the data to be protected is built on the DPM server, changes made to the live data are then applied to this replica on a frequent basis, meaning that the window for restorations is much smaller. Instead of potentially losing a days worth of data, as little as 15 minutes is at risk with DPM.
Unfortunately the product didn’t perform as seemlessly as we hoped, with numerous bugs requiring fixes by Microsoft and a lot of manhours being spent maintaining the system. Restores were also less than painless, meaning it wasn’t adequately performing the function it was built for.
We were happy with the way DPM was meant to work, and so looked for a product that did something similar, but in a more robust way. This is where Netvault FASTRecover comes in.
Netvault FASTRecover (NVFR) works in the same way as DPM but with even more granularity. It is possible to select a time frame down to 30 seconds to recover from (depending on how frequently you are replicating data of course). It also does restores in a clever way, essentially restoring a stub first and then restoring the rest of the data in the background. They claim the initial data restore can be done in 30 seconds, one to test I think.
This fast recovery capability is key for systems such as email. It means if a mail database on our Exchange system becomes corrupt we can restore a good point in time very quickly allowing people to start working again. The background restore prioritises requested items so emails that people want to read will come back quicker. We’ve yet to fully test this ourselves, but the demonstrations are impressive.
So far today I’ve installed Redhat Enterprise on the server (itself a rather impressive beast, being an HP DL380G6 with dual quad core processors, 38GB of memory and all the NICs in the world!) and installed the actual NVFR package. Configuration was easy and I’m currently running tests on a VM that’s pretending to be a file server. Next stage is to see how well it protects Exchange!
Report post
Ian Carter Data Protection