3

My company wants to institute a backup plan for all of the clients on our network, which is about 200. We back up our servers and SQL databases regularly, but its been our policy to not backup individuals.

What is most critical for people is their Documents and PST files in Outlook. PST files can be very large, and most people's are ~1-1.5 GB around here. So with PST files alone that is 200-300 GB of data needing to be transferred daily to a sever for backup. Or compressing first, then transferring, but many of the machines are VERY old and such a task would grind their computer to a halt.

Isn't this the reason networks use things like VMware -- to reduce network traffic and streamline backups? Or is this only to reduce hardware costs? Would this much network traffic everyday drastically slow down our network? Enough to the point we'd have to mandate it to be done at night only? Or could we stagger then through out the day?

Really appreciate any input, thank you.

3
  • Maybe there is some confusion, let me clear it up. We do not have a Microsoft exchange server. This is an old newspaper company that isn't doing so well financially, and no one really knows what they're doing. I have been interning 2 months there now. What we have is iPlanet running on two microsoft server 2003 machines, and mostly outlook 2000 or 2003 on clients boxes. I think I will push for us to use IMAP as that is the best mail solution I think.
    – mtkoan
    Apr 28, 2010 at 18:42
  • Simply moving to IMAP isn't going to get you everything you want. If your users are using Outlook for more than just mail, you'll need to find a way to backup things like their contacts, calendar, tasks, etc. Outlook also isn't a particularly great IMAP client either.
    – afrazier
    Apr 28, 2010 at 20:16
  • You also need to consider that you're asking your servers to store a lot more data and do a lot more work. Make sure that they're not only up to the task but that there's a disaster recovery plan. One desktop HD crash means one person is offline until their PC is repaired or replaced. Your IMAP server crashing means nobody can do anything until it's up and running again. "Having a backup" might not be enough if it's going to be a week before the server can be repaired.
    – afrazier
    Apr 28, 2010 at 20:19

7 Answers 7

2

PST files should probably be considered caches of data on already on exchange/imap servers. Generally, caches should be ignored, like temporary files. If using exchange, you can backup everything at once, using something like veritas on the exchange server. If using an imap server, it should be simpler again.

Users' Windows Documents (most of their profiles actually) can be stored centrally too, and again, can be backed up from a central location. You could do this using a fileshare on the server, or using a NAS/SAN box.

You also have the option of replicating windows active directory/exchange servers (which gives you failover/scaling), and then backing up the replicated box to avoid slowing down the main server.

Oh, and stay well away from solutions like VMWare: especially when they're solutions in search of problems, or like in this case, the wrong solutions to problems. VMWare will sell you this whole song and dance about how using their stuff can consolidate servers and save you money. What they don't tell you is that VMWare means big performance hits, and to even it out, you need to invest in really good disks subsystems (like a SAN solution) etc.

Xen is better, but I wouldn't use virtual machines except in a few specific use cases:

  • building virtual dedicated servers, and having clients rent them at about 1/4 of the price of the actual physical hardware. For internal stuff, that just needs to be functionally/organisationally isolated rather than secure, something like chroot or lxc is much better.

  • trying things out on a small scale before buying real hardware

  • testing cross-platform development work.

  • running multiple OS's on my own desktop machine for app compatibility or infrequent user support

Edit: Given what you've added about being a struggling company etc., I would suggest hiring in a unix or windows specialist, and having them set you up a mailserver with tried and PROVEN backups, along with a simple way to monitor the backups that you won't start ignoring when 99% of the messages all say "Last night's backup completed successfully."

For the files, buy two good NAS boxes that support remote backups, move everyone's profiles to one, and backups going to the other. Or, rent an off-site backup service. Remember that backups of just yesterday is no good. You need a backup of yesterday, a backup of the day before, the week before, the month before, six months before, and preferably each year back as far as tax/industry records require.

Bear in mind that off-site backups are always recommended --- a company's data represents all its work, so even if the building burns down, it's the DATA backups that really matter. And use that argument if you need to fight for funding on this.

2

We deal with our similar situation with SecondCopy (there are many other products that will do the same thing, e.g. rsync). Although we use Exchange, many of our users have local PSTs with archived messages and/or have local documents that for whatever reason make sense to leave local.

We have tasks set up in SecondCopy to copy any files stored locally to the user's private share on a server. From there, it's backed up by our regular backup system.

Note that for PSTs, you have to have Outlook closed when you're copying them or you risk having the copy by unusable. Also, even if a PST only holds old, archived messages, every time you run Outlook, the file is updated, so it has to be copied again even if nothing has actually changed. For gigabyte+ PSTs, that's a lot of redundant data to be copying around.

1

Why the heck?

  • You should not use PST files for ages now. If you have an exchange server, ust OST files. If you HAVE to use PST files, put them on a server to start with ;) Granted, it is only Outlook 2010 that finally allows more than 1 excvhange postbox per user. Still, in 99% of the cases the use of PST files is a case of an admin who should be fired because he just does it "because he always did it", and has ignored the significant changes to Outlook that changed the storage model.

  • As other pointed out - user files should never be on the workstations to start with. Roaming profiles + Folder REdirection (Documents redirected o a file server) make sure this is the case.

  • What is left is stuff like developers work folders (as in: it sucks compiling on a network share). THese, though, are irrelevant ;) You only loose the stuff since last check into the source archive.

So, in general - there is normally NO need to backup a workstation.

3
  • Most of our data is stored centrally. It is a newspaper company, so ads, stories, images, page layout, etc., is all stored in a database. For 200+ clients, all PST stored on the server, what kind of specs should said server have? Also, we have no server with enough space locally to store that much data, only several network drives.
    – mtkoan
    Apr 28, 2010 at 18:50
  • If you're going to do server side PSTs with a Windows server, you'll have to have a 64-bit OS. With 32-bit Windows, it's possible to deplete the non-paged pool and cause all kinds of problems with the server. I don't have any links handy, but you can Google for more information.
    – afrazier
    Apr 28, 2010 at 23:19
1

If you absolutely must continue with your users having local PSTs and documents that must be backed up, any backup solution you pick really, really should do 2 things:

  1. Only backup the changed parts of files. Slinging hundreds of GB of PST files across your network that have only had a few MBs of changes is sheer insanity.
  2. Have some way of reading files that have been locked with exclusive access in case someone leaves Outlook running all night. Ideally, this would use the Volume Shadow Copy services built into Windows, but a proprietary implementation is better than none. You can say that this won't happen, but it will. Murphy will ensure not only that it happens, but that it happens to a C-level executive who abso-positivi-lutely requires restoration of a Word document or email from last night's backup. :-)

No matter what you decide to implement, your initial backup is going to take a long time.

2
  • afaik, incremental backups with PST files are not possible. They are one huge file; so changing one email flags the entire PST file as changed.
    – mtkoan
    Apr 28, 2010 at 18:44
  • 1
    That's exactly what delta encoding is for. Determine which parts of the file have changed and only transmit those pieces. xdelta and rsync are programs to do this on a low level. Any good online/cloud backup software also does this to minimize the amount of data that needs uploaded, including CrashPlan, Mozy, Carbonite, BackBlaze, etc.
    – afrazier
    Apr 28, 2010 at 20:14
0

figure out what you need to restore (after all that's why you are backing up). Then you can whittle down the problem. E.G. if they say we need to back up for those pst files then the need for PST files should go away. I would have the users move their mail back up into the exchange server for backup where it belongs. You also need to worry about things like access databases, and sql server express databases.

If they want to restore user docs then point them to folder redirection so that data gets backed up on a server. If they are worried about restore times use the WAIK to build an image.

You can also add these technologies in conjunction with desktop backups in order to reduce backup time if you end up doing desktop backups.

0

I recently saw a presentation of avamar from emc that can do differential backups and also only store data blocks once (to simplify if an email was sent from somebody and received by somebody in your company it would be stored only once).

3
  • Recent versions of ZFS have this, and I think other SANs may as well. Exchange's mail store also does Single Instance Storage.
    – afrazier
    Apr 28, 2010 at 20:20
  • rsnapshot had done this for ages, using only common tools on any POSIX filesystem
    – Javier
    Apr 28, 2010 at 20:43
  • Backuppc is another one which has this feature, basically anything which wraps around rsync can do this.
    – Ali Chehab
    Apr 29, 2010 at 0:14
0

Check out Druva for laptop clients if the PST files are NOT cache of existing data on the server. The laptop clients use at client deduplication to reduce transfer times. It is PST aware too so will only sync an email inside of a PST once back to the server.

Clients that are local can use roaming profiles/folder redirection, however if the user is not local, or is far away, these do not work. We have segments of our lan that are 60ms from the server and the users cry about the speed.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .