Ask Slashdot: Open Source Back-Up Tool For Business?

New submitter xerkot writes: I am looking for a tool to make backups of PCs in a big company. We want to replace the one that we are using at this moment for this new one. The tool will be used to do backups of PCs (mainly Windows, and a few Linux), and we want to manage these backups centrally from a console, being able to automatize the backup process. The servers of the company are backed up with another tool, so they are out of scope. In the company we are being encouraged more and more to use open source software, so I would like to ask you, what are best open source tools to do backups of PCs? Are they mature enough for a big company?

Comments

What exactly are you backing up?

[i.r.id10t]

What exactly are you backing up? Entire disk images? Or just user files?

If disk images, then something like clonezilla, perhaps set up to boot from a TFTP server. Boot the machine via WOL, kick off the TFTP, automatically dump the image out to a server using the machine name or MAC address or something as a unique identifier

For user files only (ie, My Documents or whatever) can you set up network based home directories ? And then just back up the server they live on.

Re:What exactly are you backing up?

[Feneric]

Agree, in my experience Clonezilla does as well in this arena as the commercial offerings. Be wary, however. All of these products can behave a little funny sometimes in certain environments. Whatever you choose, test it out all the way through a backup and recovery before relying on it.

Re:What exactly are you backing up?

[slinches]

For user files only (ie, My Documents or whatever) can you set up network based home directories ? And then just back up the server they live on.

If you do this, you better have an insanely reliable network with at least gigabit speeds. Where I work we have all of our *nix boxes set up with the user home directories on NFS and when the network goes out, we have a lot of employees who can't do their jobs. Even a few hours of outage can be extremely costly. On top of that, no one who works with large files uses their home directory for anything due to the limited space and slow (100Mbps) network speeds. Instead everything is stored locally and backups are up to the end user (read "non-existent").

Re:What exactly are you backing up?

[Spazmania]

we want to manage these backups centrally from a console

You don't want to do that, you only think you do. That needlessly creates a security problem by intentionally installing a remote control vector. What you want is a central notification system so that if a machine expected to back up does not, you find out about it and can send on the IT staff.

What i.r.id10t alludes to is also correct: figuring out how you can restore a machine to operation is step one. That determines what form the backups can be stored in which, in turn, determines what backup tools you can use. Linux machines can be restored by booting from a live CD, writing individual files back to disk and then running "grub install." Can Windows do that, or will you need to grab a full disk image of the C: drive every time you do a backup?

Re:What exactly are you backing up?

[Mark of the North]

Yes, but once you have a Windows network to the point where anyone can sit down at any workstation, with all user data being stored on a file-server, you no longer have to back-up the user workstations. I managed several networks of between 50 and 200 workstations, each with a single Samba-based DC and file-server that was backed up off-site via rsnapshot. The workstations were imaged via FOG. Once we had a working image for a workstation (which could take a couple of days to prepare from scratch), the bottleneck in deploying workstations was physically putting them in place and hooking them up. We could image forty machines in about 25 minutes. Best of all, we could reimage machines remotely. This meant I could set up forty workstations on a Friday afternoon, register them with FOG and deploy an initial bare-bones image, then head back to my office to fine-tune a new image before redeploying it (from my office). Active Directory took over from there. If a workstation started acting up, we could trigger reimaging remotely, and 7 minutes later the user had a fresh workstation. In that environment, rsnapshot worked fine backing up about 8TB of user data from about ten Samba servers over a 20Mb/s WAN. Nightly back-ups rarely took more than an hour. Storing many snapshots (7 nightly, 4 weekly, etc.) didn't have much of an impact on the required backup storage. It was also nice that rsnapshot could run a script on a server before and after backing up. So one could dump a database to a file before backing up.

Re:What exactly are you backing up?

[freeze128]

Learn about "Offline Files". It will alleviate the problem you describe.

Re:What exactly are you backing up?

[cthulhu11]

Or, just set up CrashPlan and be done with it, and not make the mistake of having only on-site backups.

Never mix business and ideology

[NotDrWho]

Just look for the best tool for the job, and don't worry about whether it's open source or not.

Re:Never mix business and ideology

[martiniturbide]

Open Source is not only an ideology, it is a tool to reduce the risk of dependencies on a single vendor. I found that his question is very valid if he wants to reduce vendor risk.

Re:Never mix business and ideology

[Anonymous Coward]

Because there's no such thing as abandoned OSS? If anything, you're more likely to see development stop and become orphan software on some OSS freeware. Also, things like operating system upgrades can cause OSS to suddenly stop functioning or start behaving erratically, often with little or no warning until the time comes when you need it. Often it is months before new versions are finally released to solve these problems. Backup software is not the area your company should be trying to save a few dollars.

Re:Never mix business and ideology

[Anonymous Coward]

Because there's no such thing as abandoned OSS? If anything, you're more likely to see development stop and become orphan software on some OSS freeware. Also, things like operating system upgrades can cause OSS to suddenly stop functioning or start behaving erratically, often with little or no warning until the time comes when you need it. Often it is months before new versions are finally released to solve these problems. Backup software is not the area your company should be trying to save a few dollars.

As long as the source code is still available someone else can take over the project. I hate F/LOSS projects that shut-down and the source code is removed from the World Wide Web; at least move it to GitHub and post a link to the repository keeping only a redirector page active at the original domain with information about how to contact the now former maintainer to take over the domain and development.

Re:Never mix business and ideology

[Anonymous Coward]

You can pay any vendor to support your open source solution. It's illegal for a third party to fix the bugs in your proprietary software.

Re:Never mix business and ideology

[mwvdlee]

There is as much abandoned OSS as abandoned closed commercial software.
The entire point of OSS licenses is that all software will be abandoned at some point.
OSS guarentees that you have a fighting chance to keep it working yourself or atleast migrate.

Re:Never mix business and ideology

[Anonymous Coward]

Because there's no such thing as abandoned Anonymous Cowards? If anything, you're more likely to see development stop and become orphan software on some Anonymous Coward freeware. Also, things like operating system upgrades can cause Anonymous Cowards to suddenly stop functioning or start behaving erratically, often with little or no warning until the time comes when you need it. Often it is months before new versions are finally released to solve these problems.

Your points can be made about everything. Backup software is not the area your company should be trying to depend on a single vendor or a proprietary protocol.

Re:Never mix business and ideology

[gnupun]

There is as much abandoned OSS as abandoned closed commercial software.

Really? "as much"? Wasn't there a /. story two days ago about an OSS dev not interested in supporting his own app unless he got paid?

There's got to be waay more OSS abandonware than closed source because the proprietary people get paid for their work and therefore work hard to make sure the money keeps flowing in. Whereas the OSS guy quits after he's satisfied scratching his itch and there's nothing new left to explore/clone.

Re:Never mix business and ideology

[i.r.id10t]

No, that link was about a company throwing a fit and demanding that their bug be solved *immediately*

Re:Never mix business and ideology

[Anonymous Coward]

You embarrassed yourself with your wingnut sig.

Re:Never mix business and ideology

[macs4all]

You can pay any vendor to support your open source solution. It's illegal for a third party to fix the bugs in your proprietary software.

I would believe it is only illegal if the vendor then turns around and tries to sell the fixed code to another.

Oh, and if the company that wrote it is out of business, who then has the standing to sue? The government might (probably does), under Copyright laws; but won't. So who?

Re:Never mix business and ideology

[Anonymous Coward]

I see you've been living under a rock and never heard of these things called patent trolls, or SCO, or Oracle.

Re:Never mix business and ideology

[rahvin112]

FOSS didn't start until the mid 80's and Linux (which gave FOSS real legs) wasn't a thing until the early 90's. There was so much abandoned commercial software at the point that FOSS became popular that FOSS will never be able to catch up. There are literally entire companies and classes of software that are just gone and lost to humanity entirely because copyright prevented the retention of the code.

Have you ever even looked at an abandonware site?

Re:Never mix business and ideology

[macs4all]

I see you've been living under a rock and never heard of these things called patent trolls, or SCO, or Oracle.

And how, pray tell, are any of those entities going to find out about your private little contract with a vendor to fix your particular instance of whatever code?

Re:Never mix business and ideology

[RabidReindeer]

I see you've been living under a rock and never heard of these things called patent trolls, or SCO, or Oracle.

And how, pray tell, are any of those entities going to find out about your private little contract with a vendor to fix your particular instance of whatever code?

Software audits. Oracle and Microsoft are famous for going on-site and sifting through computers looking for non-compliance. Non-compliance doesn't only mean using unauthorized copies of software, it also means unauthorized use of authorized software. And if that isn't enough, the DMCA can be used on you for reverse-engineering a product in order to modify or augment it. It also tends to "void the warranty" on your legitimate uses.

Open Source isn't necessarily the Solution to Everything, but most OSS has a license that explicitly permits customization for private use, at least. And no one's going to come around auditing it.

Re:Never mix business and ideology

[RabidReindeer]

There is as much abandoned OSS as abandoned closed commercial software.

Really? "as much"? Wasn't there a /. story two days ago about an OSS dev not interested in supporting his own app unless he got paid?

There's got to be waay more OSS abandonware than closed source because the proprietary people get paid for their work and therefore work hard to make sure the money keeps flowing in. Whereas the OSS guy quits after he's satisfied scratching his itch and there's nothing new left to explore/clone.

I doubt it. I have shelves full of commercial "abandonware". Mostly just in case I need to fire up an antique system for the sole purpose of dealing with some dust-ridden data. At least with OSS, you can usually get the source and do something with it. If a commercial product is truly abandoned and the vendor is defunct, you've got to reverse-engineer it and hope patent/copyright laws aren't going to get you. If it's something worse - like an OS/2 version of a Norton/Symantec product, you can basically give it up, unless you have enough cash to make them listen. And for some products/vendors, that would have to be a lot of cash indeed. Think about Windows XP and various world governments, for example.

Re:Never mix business and ideology

[macs4all]

Non-compliance doesn't only mean using unauthorized copies of software, it also means unauthorized use of authorized software.

Depends on what is considered "Unauthorized use".

the DMCA can be used on you for reverse-engineering a product in order to modify or augment it.

Show me ONE case of that ACTUALLY happening. NO Jury would Convict, if the "reverse-engineering" was simply for the private use of an end-user.

I agree that this MIGHT be technically illegal; but in a practical sense, it would NEVER fly in Court; nor would you find a Persecutor with enough time on their hands to mess with it.

The Feds will typically only get involved if the "damages" exceed $100k. Most software that isn't Government-Contracted or Oracle-based fails to meet that standard.

Re:Never mix business and ideology

[Anonymous Coward]

Copyrights don't revert to the government when a company goes out of business. You really shouldn't be participating in a discussion of software licensing.

Re:Never mix business and ideology

[Anonymous Coward]

There's got to be waay more OSS abandonware than closed source because the proprietary people get paid for their work and therefore work hard to make sure the money keeps flowing in.

That does not work if the money flow is not large (or stable) enough. If you are using and paying for a closed source system but other clients decide to stop using it, there is a significant risk of the software being abandoned even though you pay for it.

If you use a widely used commercial software, this risk is small. If you depend on a small player's software, you'd better require a source code escrow. If you use open source software, you already have that.

Re:Never mix business and ideology

[RabidReindeer]

IANAL, but I'm sure that someone could enumerate specific cases where things like that have been prosecuted. Consider what the RIAA and MPAA have done to people.

You're assuming a sane and reasonable legal system, and a lot of us doubt that such a thing exists. If there's a law, someone will abuse it.

amanda

[Anonymous Coward]

you can use amanda in case you want to backup files. amanda is production grade and has clients for windows and linux and
possibly unix alike. -- mallah

Re:Amanda

[Anonymous Coward]

broken link

Re:amanda

[rklrkl]

Amanda is great if you're backing up only Linux clents, but the Windows Amanda client is a total abomination. Not been updated in over 2 years (don't believe the version numbering - check the timestamp of setup.exe), pointlessly uses a MySQL DB on the client side (ridiculous!) which the Linux client *or* server doesn't use, regularly times out, regularly crashes, produces byzantine error codes without any description/documentation of them, much slower than the equivalent Linux client and produces ZIP64 backups that are pretty well impossible to extract from on Linux (which is annoying, because the Amanda server side is Linux-only).

We have a Ultrium tape drives with multi-slot autoloaders/barcodes, so there are very few Open Source backup solutions that can handle this (including tape-spanning if needed) out-of-the-box. Bacula was another potential solution, but it's horrific to set up (makes the dreadful Oracle DB install process look like a breeze) - just reading the Bacula install docs brings me out in a cold sweat :-) I guess business-level backups just aren't sexy enough to warrant a decent Open Source solution...

Don't cut corners

[cloud.pt]

I once was in a crossroads of choosing between stuff like Clonezilla and Bacula, for small business purposes. Bottom line is they add a lot of complexity for low to no flexibility. I ended up building my own tar/move/ script with cron triggers at after ours downtime, then I would simply move them around network locations for avoiding single points of failure messing up the backups. Adding your own exceptions for the backup is a plus. At the last point, I had something reliable, fast, and that would require the simple overhead of re-installing Debian before the actual restore, then an update-grub and a change in fstab for the new disk replacing the broken one's UUID (because you don't really do that many restores so it's a fair trade-off, while you do save time exponentially by not backing up the entire OS). A good starting point is http://www.aboutdebian.com/tar...

Re:Don't cut corners

[cloud.pt]

Looking at your question again, this might not be your best bet due to scale and usage of Windows. Still a good choice for the Linux, 5-machine tops sysadmin.

Re:Don't cut corners

[Anonymous Coward]

mainly windows and a few linux

Re:Don't cut corners

[cloud.pt]

triple post, but just noting my compressed backups for a 100 mailbox, jboss instance with heavy app deployment and traffic, and fully configured ISP-config would range between 2 to 4GB. I had a 30 day rotation set in place, then every 6 months I would run some cleanup tools to go back to 2GB'ish archives.

Re:Don't cut corners

[Anonymous Coward]

I once was in a crossroads of choosing between stuff like Clonezilla and Bacula, for small business purposes. Bottom line is they add a lot of complexity for low to no flexibility.

Talk about not having a clue!

Bacula is highly, higly customizable: for example, I wrote a program in shell which connected Bacula with Oracle RMAN, and now Bacula obs kick off RMAN database backups every day, then syphon the RMAN backups off to tape. And Bacula controls the tape library and the barcode reader through other external programs. Such a thing would not be possible were Bacula not designed to be so flexible as to interface with everything and anything, including emulating virtual tapes as entire drives, or even just files...

Talk about not having a clue.

Re:Don't cut corners

[Anonymous Coward]

In my experience, there are five different backup programs:

Ones that actually should be worth being called "backup programs", with features like deduplication, multiplatform capability, bare metal restores, encryption, ability to use different media, copy a backup to different media if needed, expiration, and so on. NetBackup, Backup Exec, TSM, and ArcServe fall here. To a lesser extent, Retrospect.

Then we get the piecemeal tools:

Bacula is OK... but after doing tests and finding it generated corrupted backups quietly on the NFS server, it was found worthless.
Obnam is promising with deduplication... but after 2-3 months of running, the deduplicated backups were not restorable with dedup fails.

So, I have to cobble something together. OS images, I use CloneZilla. File backups get tarred off (hard to break tar), deduplication is used as a second means.

Sad thing, since there is little to no interest in backups, for all but the enterprise, we are still in the Stone Age here.

Re:Don't cut corners

[Anonymous Coward]

Take a look at tools like borgbackup instead of using tar/cron/move/rsync. The main advantages are:

- Deduplication of varied size blocks (more backup snapshots in less space)

- Encryption

- Very efficient over the wire and during the file scan

- Efficient storage in the repository (low file count)

Case in point, I used to backup about 100GB of Maildir IMAP mailboxes (a few million files). Tools like rdiff-backup would take hours to backup the file system snapshot and would create tens of millions of inodes on the backup target file system. The rdiff-backup solution would also get slower over time. Throwing borgbackup at it, the backups now run in under 15-20 minutes, and it only creates a few thousand inodes on the backup target file system.

That means I can keep a lot more snapshots around, with the same backup performance.

Re:Don't cut corners

[cloud.pt]

I believe the problems of a company needing backups are usually too specific for there to be a swiss army knife for the backup topic. That's why I mentioned the "on-the-rocks" solution always being the better one. That's why many companies end up releasing their own versions of so many things, backup tools included, and then only a portion end up gaining traction.

Amanda

[Anonymous Coward]

http://amanda.zmanda.com/

checkout restic

[Anonymous Coward]

free open source runs on windows freebsd linux and supports sft and s3 aws

Re:checkout restic

[Anomalyst]

https://github.com/restic/rest...

Ther is no such thing as a free sandwhich.

[mt2mb4me]

Backula I think would do it, but you have to pay for it. If you are looking for free, you will have to learn how to script your own agents. For a small business (25-50 PC+) Windows Server Essentials is Boss! It boots a tftp server for recovery and nicely manages all the machines it backs-up. Conversely, You should centralize user data, so if their PC goes down, they don't lose any work. Then just keep a windows PE custom install of the PC's when they are build fresh and you can just do an image restore on failure.

Bacula

[Trevin]

I use Bacula for my home computer; it feels powerful enough for a small office, and is very versatile.

It has three main components: a client daemon that you install on the computers you want to back up, a storage daemon that you install on the computer that will write the backup files and/or tapes, and a director daemon which controls the backups. The director and storage daemons only run on unix-like operating systems (BSD, Linux, Solaris) but the client daemon has also been built for MS-Windows.

http://blog.bacula.org/

Re:Bacula

[mridoni]

I regularly use Bacula to back up a network of about 20 workstations and a couple of servers, plus a remote one (connected through a VPN). It's fast and reliable, if I really have to nitpick there are some quirks, though they depend mostly on lack of disk space for the backups (this is a non-profit with a tight budget) and lack of time on my side for fine-tuning the configuration. I'm really satisfied, since 2008 Bacula has saved our ass a number of times with minimal effort and maintenance.

Re:Bacula

[Anonymous Coward]

And there is a fork: Bareos.

Re:Bacula

[Anonymous Coward]

I used it to back up some critical DNS servers. Well, it silently failed, and generated corrupt, unusable backups on the NFS server. Thankfully, I had another script back up the config files that just launched, did a tar, and kept those up to date.

Zmanda the company, Amanda the open source backup

[Anonymous Coward]

http://www.zmanda.com/company.html

"Zmanda: The Leader in Open Source Backup
Proven, Cost-Effective Open Source Backup and Recovery

Zmanda is the world’s leading provider of open source backup and recovery software. Our open source development and distribution model enables us to deliver the highest quality backup software such as Amanda Enterprise and Zmanda Recovery Manager for MySQL at a fraction of the cost of software from proprietary vendors. Our simple-to-use yet feature-rich backup software is complemented by top-notch services and support expected by enterprise customers."

http://www.amanda.org
"Amanda Network Backup
Open Source. Open Formats. Open APIs.
What is Amanda?

AMANDA, the Advanced Maryland Automatic Network Disk Archiver, is a backup solution that allows the IT administrator to set up a single master backup server to back up multiple hosts over network to tape drives/changers or disks or optical media. Amanda uses native utilities and formats (e.g. dump and/or GNU tar) and can back up a large number of servers and workstations running multiple versions of Linux or Unix. Amanda uses a native Windows client to back up Microsoft Windows desktops and servers.

The most recent stable release is version 3.3.7p1, released on February 15, 2015. "

Re:Automatize this!

[Anonymous Coward]

I think you mean outsourced? This is a perfect fit for India. We should not be paying American IT wages for a backup monkey.

My advice..

[Anonymous Coward]

in a big company

Promise me something... that your big company gives back to open source.

Yes? Awesome!

Now here is your free advice.

OwnCloud, and back up that server

[cerberusss]

I worked at a scientific institute, and they simply installed OwnCloud everywhere. It's got a client for most platforms, syncs to a server, and allows you to back up the server in the usual fashion.

It worked so well, that when I started doing consulting (at the client site), I got my own VPS with Debian, and installed OwnCloud server on that. Then installed the client on my private laptop and the laptop that I got from the client. Works beautifully, because communication is over HTTPS. Company firewalls don't block that. I tried other things like BitTorrent Sync, but these use special ports.

Re:OwnCloud, and back up that server

[cloud.pt]

This seems like the "kill a fly with a canon" solution. But it does look like it could work if money and infrastructure is no object, with full-fledged system images as a plus.

Bacula

[Anonymous Coward]

The tool you are looking for is Bacula, and it's enterprise grade.

You'll need a GNU/Linux or UNIX server; Solaris (or OpenSolaris variant, such as SmartOS or OmniOS) is recommended, as tape library support in Solaris is first class, especially for libraries with barcodes and robots.

Configure it with SQLite support, which is the simplest and fastest, and off you go.

The client side supports Windows. The storage and director nodes must run on some form of UNIX or UNIX-like OS. storage and director node can be combined on a single system, but they do not have to. The software really is enterprise grade, which is amazing for free, open source software.

BackupPC

[Anonymous Coward]

They use this around here, seem to like it. Seems apt to your situation.

http://backuppc.sourceforge.net/

Re:BackupPC

[Anonymous Coward]

I also use BackupPC. For file backups, it works reasonably well. You have to get your exception list right and it does have its quirks, but it's saved me many times. Doing a full bare metal restore is difficult and time consuming, though - this is more for people who only care about their actual files, not quick recovery.

I use standard rsync (over ssh) for Linux, and rsyncd for Windows, with DeltaCopy as the rsyncd service on the clients. This has an additional problem, the Windows backups aren't encrypted. Not good if you have sensitive information.

You need a reasonably powerful backup server in terms of disk performance. I recommend a raid 10 of enterprise class 7200rpm disks at a minimum. The file deduplication saves a lot of space but it chews up those IOPS. If you intend to do archives, you need even more. I currently have about 13TB of backups in 1.25TB of files, but YMMV depending on your data.

It has email notifications and a web console, where you can see when the last backup was, and details about the backups and hosts.

If you have a decent configuration management tool in place, you can automate the setup too. I use Salt, and I have the Linux client side fully automated, haven't quite gotten the server config automated yet, and haven't bothered to try the Windows side, but I'm sure it's pretty easily done. My environment is pretty small so that hasn't been a priority.

Re:BackupPC

[Sun]

This has an additional problem, the Windows backups aren't encrypted. Not good if you have sensitive information.

<plug>Throw rsyncrypto into the mix</plug>

This has the downside of being a preprocessing step (i.e. - you need local storage for the encrypted form of the files), but solves the encryption problem better than your suggestion (which encrypts the transit, but not the actual backup).

Shachar

Backups

[Anonymous Coward]

If just backing up file shares I like backuppc. However if you are virtualized and need a hypervisor level backup solution I like the Unitrends Free.

Awesome backup solution.

[Anonymous Coward]

I have been using Fog and OwnCloud.
Fog (Free open ghost) for desktop backup and reimaging and OwnCloud for data backup.
http://sourceforge.net/projects/freeghost/

Free Opensource Ghost

[waspleg]

FOG can back up disk images and do inventory and scheduled re-images and keep images for things like AV scan and whatnot.

We mostly just use it to wipe machines back to clean slates ourselves but it's supposed to have quite a rich feature set. It does also have some bugs though.

I set one up at my school that has re-imaged 1000+ student netbooks 7 at a time, the school district also has all of our images centrally located and available throughout the district for all models of laptops/desktops that we use and manages the images there. It uses iPXE/Partclone/TFTP and LAMP. The original .32 version has plugins that can do things like identify a machine based on its DMI info and pick a set of images based on that (Capone).

I've set them up on CentOS and Ubuntu. There are always quirks though, I suggest strong perusal of the wiki before deciding, and toy with it before putting in to production.

Amanda and S3

[mi]

Use AMANDA to do the back-ups. Use Amazon's S3 to actually store the dumps compressed and encrypted at the source — AMANDA has had the S3 back-end for a while. No, you do not need "Amanda Enterprise".

Having set just such a thing up at my last job, I'd be happy to help you out for a regular consulting fee. Should not take more than a week or two even on a large organization.

syncthing

[Anonymous Coward]

syncthing, is like dropbox but decentralised, might be worth a look. I use this for my own stuff. Someone out there is using it to sync 20000 GiB of data. It also lets you use ignore patterns on devices.

rsync --link-dest $PREVIOUS

[Anonymous Coward]

Rync's --link-dest option lets you keep incremental backups that look almost exactly like full archivals but without the extra size footprint. Advantages:

* Rock solid, proven tech, been around forever, wlll probably outlast almost any other less standard solution.

* Fully and easily scriptable - no awkward GUI or web service to deal with

* Produces a standard filesystem image, usable with any other tool that deals with files, rather than some tool-specific backup image.

* You have full control over the policies - e.g, how many previous backed up versions to keep.

* Back up to anything that has a filesystem supporting hardlinks. Local, over the net, USB drives, anything you can name.

Oldie but goodie, Backuppc

[Anonymous Coward]

I have used this for home and work to replace tape backups (been awhile but still good software) web based and has multiple ways of syncing files (tar, rsync, smb)

Plus allows users to do restores themselves if needed

http://backuppc.sourceforge.net/

Re: Oldie but goodie, Backuppc

[Anonymous Coward]

+1. Using it in environment with 1100 windows clients via rsync, works great. You can define what folders and what file types you want to backup, define schedule ...

Don't back up end user PC's at all.

[enjar]

Set up home drives on file servers and back those up. Teach users that those are the only locations that are backed up. Set up the PCs to use that as the default home location. You can do this on Windows and Linux just fine. Invest in the server -- redundant power supplies, RAID arrays, failover, etc. You could even look at various open source NAS devices, or whatever works for your environment.

Why?

Backing up user PCs doesn't scale well and becomes a thankless task for some poor employee who has to keep up with broken backup clients. It's far easier to scale when you only have to keep up with the file servers. You have some number of clients saving to each server, but that's that number of backup clients you don't need to deal with. This frees up IT staff for other, more useful tasks.

It also allows you to replace end-user PCs with a simple re-image rather than trying to recover or fix anything. End-user calls and says their PC is going whacko, you pull a spare off the shelf and lay down a fresh install. Show up, take the malfunctioning equipment away and diagnose it on your time, while they get back to work. Since all the files are on the server they can just get back to it rather than waiting on you to try and fix whatever might be going wrong.

Re:Don't back up end user PC's at all.

[swb]

I agree with this but I can't tell you the number of times, even where reasonably well implemented, it resulted in A User of Some Organizational Clout losing some data they wasn't saved where it belonged. Mostly this was ultimately consequence free, but getting to that state often required some kind of extensive recovery.

It's also iffy for mobile users, or at least was kind of iffy prior to the proliferation of cloud sync options.

I kind of wish you could do some kind of automated desktop backup of the user profile -- it'd make the process of swapping in a replacement desktop. Like scripting Windows 7 EZ transfer or something, so that on the replacement PC you could have the user regain whatever BS customizations they had.

Re:Don't back up end user PC's at all.

[enjar]

Fortunately we have buy-in from the top on this strategy, and they also eat their own dogfood. So people bitching about it can complain all the way to the CEO and he will say "too damn bad, that's how we do things, and you were told about it on multiple occasions"

For mobile users there are far better options available now than there ever used to be. For Windows mobile clients we implemented DirectAccess and some other stuff that automatically takes care of a lot of the syncing stuff. We also keep all email on the server only to avoid dealing with local Outlook mail files.

The user profile thing still is a pain. Roaming Profiles can help somewhat but it's not 100%.

And of course I can't go into full detail about everything we do ... but when I started in this job we were doing workstation backups ... and eventually when presented with the estimate for scaling that out as the company grew, executive management bought into the central server model. The costs of staff, software and equipment to back up thousands of workstations just didn't make sense at all, even if the software is "free", the staff and disk space isn't.

Re:Don't back up end user PC's at all.

[Anonymous Coward]

I cant recall the details, but at a previous place I worked at we enabled offline files for the users home share, and redirected their local my docs etc into it. Windows then keeps a local encrypted cache of their files, and resyncs changes back to the server if the system has been offline.

Enable roaming profiles, then the "user state" is also preserved.

The above two can all be done with GPO, it require research and a few tweaks to make it all work seamlessly.

If a PC crapps out, you simply reimage a spare, log the user in, and boom, all their stuff is there just like before.

Real life example - Accounting Firm, its "tax season" accountants are busy and stressed out to the max because their clients leave everything to the last minute, and they up until 3am every morning getting your tax retruns done. One of the head accounts comes by around mid day on a Friday to say his PC wont boot up. I go take a look, hard drive has toasted it self. I find a spare drive, put it in, reimage the system from the network (we had WDS/MDT setup), this takes maybe an hour tops, user logs back in, all their settings, their desktop, their files, the whole shebang like nothing had happened, reinstalled a couple apps they like (thats not in the standard company image), they just work because roaming profiles are enabled, and the settings are still in %appdata%/roaming. 3PM, job done, I go home at 4pm like usual. Now if this had of been a manual process, I would of been there all weekend putting the OS, apps and settings back-together and helping them get all this stuff configured again. No 3rd party tools or apps needed, just what MS already provides you with Windows 7 and Server 2008 Standard, its all there.

Then backup your servers as per usual.

CrashPlan

[Hohlraum]

Can be used entirely for free and encrypted backups. The CrashPlan makes money with their cloud backup features. You can back machines up to a single machine or multiple machines on your network that have CrashPlan installed one them. You can run in a scheduled mode or do realtime (every 15 minutes?) backups.

Re:CrashPlan

[Hohlraum]

Sorry, not opensource but free. :(

Re: CrashPlan

[Anonymous Coward]

I second this. I would pay for the cloud option as well. Another one that is decent is Mozy, which cost money too. They both back up to the cloud, which is nice since it's easier to manage when you have a mobile workforce. Also, not having to purchase and maintain the hardware yourself and worry about data replication / offsite storage is a big plus.

Personally, I tend to go with services the company pays for anyhow--especially something critical like backups. So that when there is an issue we can just do a WebEx with the support team from the company that we pay and they fix it.

why the hell do you care what's on individual PCs?

[Anonymous Coward]

Anything important is stored on a server, PCs are cloned/imaged/whatever and interchangeable/disposable.

First off, store most data on servers

[davidwr]

As much as is feasible, store files on the servers you have already.

I realize this may not be feasible if your "daytime bandwidth" or latency makes it impossible, but do it if you can.

I'll leave it up to others who know more than I do to answer your original question about open-source, centrally-managed, business-grade (read: vendor-supported and hack-resistant) solutions.

Oh, one more thing: this is a business. Unless you are going to dedicate a programming team to bug-fixing this and a security team to regularly audit it, spend the money on buying software from a reputable vendor who will stay on top of security bugs. Don't make the mistake of thinking "open source means fee as in beer" - if you do, you and your company will pay for it big time with the first preventable security breach.

Did you even do any basic research?

[jawtheshark]

Anyway, I use BackupPC to backup user files from Windows machines on the company network. Works just fine. I tell my users it's "no guarantee", as they should store on the network shares any way.

Re:Did you even do any basic research?

[Anonymous Coward]

We use this to do file backups for all of our infrastructure. Its simple, and works really well. Also check out http://www.michaelstowe.com/backuppc/ for getting rsyncd working on windows.

Rsync Script + Cron job

[celest]

Rsync is simple, lightweight, has been around forever, and gives you incredible power. Assuming by "manage centrally from a console" you mean that you have remote admin access to all the computers in the scope, it's as simple as a cron job running your Rsync script. You can trivially make several versions for different use cases (Linux vs. PC) and only have to configure the setup once in the cron job. After that, you only need to touch it if you make changes.

Rsync can push deltas to any remote server you have access to via a wide range of protocols. The rest of your IT team will appreciate that you're only sending deltas and not sending full copies every execution and hogging bandwidth.

Here's a link to get you started: https://wiki.archlinux.org/ind...

Good luck!

Re:Rsync Script + Cron job

[Miamicanes]

Is there any free rsync SERVER capable of running under 64-bit Windows 7 Pro and working with any known Android rsync client over wifi (to an ext2-formatted hard drive mounted using ext2fsd)?

Re:Rsync Script + Cron job

[Anonymous Coward]

I don't know what you mean by "rsync server", but you can run a cygwin version of rsync and openssh. You want to initiate your backups from the computer storing the backups, not the computer being backed up. That way, malware on the backed up computer can't connect to your backup storage and ruin it.
http://www.trueblade.com/knowl...
http://backup.programtech.eu/n...

URBackup

[Anonymous Coward]

Maybe URBackup is the right tool.

Re: URBackup

[Anonymous Coward]

Second this.
http://www.urbackup.org

BackupPC

[Anonymous Coward]

BackupPC .

Problem solved. Next?

Re:motivation?

[Anonymous Coward]

Is management doing this because they want the benefits of open software (ability to audit and change the code, etc)? Or because they don't want to spend money?
Because if they're trying to avoid spending money on *backups* of all things, managements priorities are fucked.

$0.02

There Fixed it for you.

Veeam Endpoint Backup Free even if not opensource

[Anonymous Coward]

Use Veeam Endpoint Backup Free. Problem Solved!

THIS!! ^^

[Anonymous Coward]

Most end-user machines are full of crap.

At one company, lots of people decided to use their work PCs as storage for their personal, home, backups. We expected about 2G of files per PC, but found people had 50G-300G of stuff they wanted backed up. That changed the project economics completely.

Politically, it was decided that during the next system refresh, very small SDDs were deployed with tiny Data partitions for users. People where pissed, but the faster CPUs made up for it. Many started using SDHC memory as expansion for their personal crap, which was fine.

Data areas for documents was provided inside a new DMS - Alfresco-based. In quick training, we made it clear that all storage was shared at the group level, the few shopping remaining lists were removed and no cat videos or music was ever placed onto the DMS.

Keep business data on servers.

BareOS fork of Bacula

[Anonymous Coward]

BareOS Works well in our Enterprise environment. It's our secondary back up system to Netbackup. I've hooked a StorTek SL100 (100 Tape Jukebox) and works well.

Re: motivation?

[Anonymous Coward]

Yes, you don't want to use free software for anything important.

Re:Yes, there is!

[darkain]

Seconding this. In the companies I work with, this is the solution we've put into place. Windows PCs use mapped network drives for personal folders and shared folders to a server. The server runs ZFS as its file system. Simple cron scripts on the server itself automate the process of creating snapshots and doing send/receive with other servers both inside and outside the building. These additional machines also store a certain number of snapshots, so we can recover previous versions of files easily. ZFS + Samba 4 also appears to Windows as "Volume Shadow Copy", so snapshots also become browsable directly by Windows clients for the more technical users who know how to work with it.

The great thing about this strategy is that 1) you're not installing a single thing on client machines, just an initial config for mapped network drives. 2) next to nothing to install on the server machine, just ZFS and Samba (which both come with FreeNAS if you want a very simple drop-in solution)

What about ZFS?

[macs4all]

What about using ZFS and essentially foregoing traditional backups?

I know that is heresy to traditional data-management methodology; but with ZFS' resilvering and anti-bit-rot self-healing capabilities, it would seem that, other than a fire or tornado hitting the server closet, or outright theft, that ZFS totally answers the need for traditional backup. And if you combine that with incremental backup to an offsite data-store (also ZFS?), then how wonderful would that be?

I admit my interest in ZFS FAR outstrips my knowledge of same; but it seems that the OpenZFS community pretty much has their act together.So, are there any backup tools that can backup a ZFS Pool to another ZFS Pool, or a network-full of Mixed-Platform Servers to a ZFS Pool?

Re:What about ZFS?

[Anonymous Coward]

ZFS != backups, just like RAID != backups. It provides no protection against ransomware or someone doing a "rm-rf /".

Seriously, 99% of the stuff in this post is pure rubbish. Yes, one can throw a copy of files to another drive, but serious backups really require a dedicated machine that pulls the data from a machine. It also should either offer bare metal image restore, or at the minimum, the ability to push an OS + template, + a CM tool, then throw the data back on.

There are not many items that do this. Windows Server Fundamentals/WHS comes to mind. EMC Avamars and NetBackup appliances also are useful... but $50,000 is not in most users' price ranges. There used to be Retrospect, but it has a proprietary format that doesn't really handle damage well.

You basically have two choices: An enterprise level backup utility, or cobble together a solution and pray it works. If budget prevents a proven solution, I'll go with the amateur night stuff... but I'll make sure to have at least two different ways of backing up machines, so when (not if) one fails, I still am OK... and I also run test restores on a monthly basis just to check for hidden corruption of stuff.

Re:What about ZFS?

[macs4all]

ZFS != backups, just like RAID != backups. It provides no protection against ransomware or someone doing a "rm-rf /".

[snip]

You basically have two choices: An enterprise level backup utility, or cobble together a solution and pray it works. If budget prevents a proven solution, I'll go with the amateur night stuff... but I'll make sure to have at least two different ways of backing up machines, so when (not if) one fails, I still am OK... and I also run test restores on a monthly basis just to check for hidden corruption of stuff.

I wholeheartedly agree with your statement that ZFS != backups (and even more so with RAID, in my bitter experience!)

However, I worry (probably too much) about bit-rot, and the problem of overwriting a good Backup with a bit-rotted Original. I was hoping that ZFS would at least help prevent the bit-rot from spoiling the Original.

I also agree with you that it's NOT the "Backup" that's critical, it's the RESTORE. That's why I won't mess with ANY Backup software that creates Proprietary backup "packages". IIRC, that's what turned me off to Retrospect, way back when it was a Mac-Only product. You're right: They ARE fragile, and NOTHING is as disheartening as a Backup that won't Restore!

In fact, that's one of the things that is off-putting about ZFS: If something goes wrong, where are the tools to try and pull something/anything out of the data-soup?

URBackup

[Anonymous Coward]

I've successfully used URBackup (urbackup.org) in a heterogenous Enterprise environment for a while now. No issues, simple and effective.

You and what army?

[cloud.pt]

While you do present very fair arguments, you failed, like me, to address important issues of the OP. Not to mention your inconsistency on such examples.

For instance, you argument for customization with an example based on the trigger of a proprietary recovery system. So you can trigger a remote process with Bacula. I have a one-liner shell script command for that, nice job. Especially with a proprietary system which the OP specifically excluded in the title (Open. Source.). And from the clues I got, Bacula has issues with Windows systems, or you're gonna be needing at least 2 different sets of binaries to support both MSW+NIX. You need something pre-main-OS boot for a completely dual solution (as an under user, I think of the likes of Acronis or Ghost, and stuff too complex for me even to start name dropping for enterprise-level).

Numero dos: tapes? Your argument for having a clue for backups is name dropping TAPES? In case you didn't notice, snapshot'ing, with rotation doesn't really need tape support. You're not going for the petabytes, and I doubt you're going for the 10+years, even for corporate. A snapshot will have only the diff of the previous one. With that said, Why da fck do I need tapes for, or barcodes, or bells and whistles?. You could have mentioned diff support in Bacula, that would have been big, if it wasn't so deep in its manual (mostly what took me away from Bacula/Clonezilla. Now to be fair, my example also does full backups over time, no diffs and integrity checks, but with exceptions - my method happens to produce really, REALLY minimal, compressed tars with 100% parity and 99% reliability to what was on that backup day if needs be (you just update Debian until that point of time, single difference in the system will be the hardware/UUID).

You know what else was designed to interface with everything and anything (including emulating virtual tapes as entire drives, or even just files)? The Unix CLI

And just to make a fking point, this guy came to /. asking around for a backup solution. Put yourself in that perspective before you go demanding the manpower of setting a datacenter-wide B&R process. Because when you start mentioning the big bucks Oracle (seriously, when there's something which's presentation page has a Training/Certification section, like that Oracle crap does, it's big bucks), that pretty much translates that you're gonna need an IT section just for managing that B&R process. This guys came to slashdot, Ima' repeat that, to SLASHDOT asking for FOSS. No matter how much he keeps mentioning "we" or "the company", he is 100% either a manager of a small team (1-3 ppl) who is trying to focus on not-so-scalable stuff. He doesn't need tapes or Oracle's antics, unless he's already on that wagon (which he's not, he was ordered to use FOSS... and HE CAME TO SLASHDOT, not Oracle support forums). And he if wants FOSS, if he does have that big a company, there's nothing more FOSS than going the extra mile and making your own fucking shell script-based super-duper-complex process, since the company is pressuring wide FOSS usage, then upping it to github with a copyleft/apache license afterwards. In the end, that's what the big guys who want to go FOSS are doing now, unless you didn't have a clue.. If that company is going FOSS for financial reasons, they are doing it wrong.

tarsnap

[Anonymous Coward]

http://www.tarsnap.com

It is actually the only backup service I trust. The content is backed up on the net but it is fully encrypted.
They charge for the service (cheap!) but the main tool is opensource.

Filezilla to offsite for the win

[Anonymous Coward]

That's all. One remote system (or better yet, many), and a client on your workstation to sync with accounts on n remote systems.

Caveat: Filezilla is partnered with Sourceforge. If you don't know why I'm warning about this, where have you been - under a rock? The short version: Sourceforge have been found sideloading crapware on project downloads. I avoid them. I suggest others do the same. In case you're wondering, I have an older version (that's still perfectly functional) in my own library.

One of many citations: http://www.howtogeek.com/218764/warning-don%E2%80%99t-download-software-from-sourceforge-if-you-can-help-it/

Filezilla: https://filezilla-project.org/

~ihtoit

UrBackup

[Anonymous Coward]

UrBackup is what you're looking for. Centralized management, file backup or image based, works over the internet if need be. It's not perfect, but it is free and open.

http://www.urbackup.org/

rsync?

[bytesex]

rsync and a few scripts. Perl scripts.

Re:rsync?

[plazarch]

I would extend this answer a little bit. At a small business with a mix of Windows and Linux boxes I use Backuppc. It uses Rsync, and does de-dup and compression on the server side.

Re:rsync?

[rongten]

Backuppc is exceptional for linux, but for windows and the infamous pst files, better to have a client/server architecture of the BURP software.

Same server, running backuppc and burp, and you save to tape the whole backup FS every 6 months.

You can use bacula for the last step if you so wish, we use straight copy to lto tapes.

CrashPlan

[Anonymous Coward]

I use crashplan for my backups. It's not open source, but its TOS allows for this use case.

butterbackup

[Anonymous Coward]

https://github.com/mikaelfrykh...

Uses rsync and stores everthing as btrfs snapshots on a central server.

Re:butterbackup

[Anonymous Coward]

Is this anything like a butterface?

Sounds like a job for Amanda

[Anonymous Coward]

Check out Amanda.org. It was developed at the University of Maryland for a scenario that sounds like yours.

A non-open-source solution...

[Voyager529]

Look, I know that "open source solution" is in the title. The low hanging fruit is already camping out in the thread - Bacula, Clonezilla, and script/cron/rsync are the major solutions there.

If the business is okay with "free, even for commercial use", Veeam Endpoint Backup is excellent. It will either back up to a Samba share or a Veeam B&R if you have one in the environment somewhere. It's legit freeware, and works very well.

Even if not for this particular case, it works well for laptops. It's the only free backup application i've seen that will back up to a USB drive, such that connecting the drive triggers a backup, rather than relying on a schedule. This is great for laptop users.

Re:A non-open-source solution...

[Anonymous Coward]

You saw the post below, right?

Non-open-source is a nightmare for any required long-term use software like backup software.

That poor bastard below is now going to have to keep his old-vendor-abandonware solution around forever (or until it's backups are expired) along with an operating system (that will get old and abandoned) that will run it while he starts using a new one. And if he makes the same mistake and gets vendor-abandoneware fucked again, he will have 3 solutions that he has to keep around. Rinse-repeat.

We were using Arkeia, but WD abandoned it

[Anonymous Coward]

We were using Arkeia, a commercial product made by the company of the same name. It was not opensource, but worked in mixed environments and was fast, easy and reliable. Unfortunately, the company was purchased by Western Digital, who finally discontinued the product. We asked WD to release it as free software but got no answer.

We are now in the process to move to a nightmare of rsync backups, difficult to audit and to maintain. BackupPC works to some extent, but open files are a problem, it has no support for application hot backups (mysql, an others) and so on.

Re:We were using Arkeia, but WD abandoned it

[Anonymous Coward]

But you see the problem with !open_source though, right? Surely you are not going to make that mistake again are you?

It Depends...

[mhkohne]

As usual.
My company uses Bacula for all our server backups, and it works pretty well, once you beat the configuration into doing what you want.
Some things about Bacula that I've noticed:
1) It's scheduling is more than little rigid. I'm not using it on desktop PCs for that reason (the PC pretty much needs to be there when Bacula wants it to be, or you miss that backup cycle. As near as I can tell, anyway).
2) Trying to configure the retention times for Bacula is NOT for the faint of heart. Get someone to help you. It's goofier than it should be.
3) Bacula thinks of all backup media as tapes. You can make it use disk (which is REALLY convenient for frequently accessed backups), but it still treats it like a bunch of tapes.
4) If I understand correctly, the Windows backup client software isn't free anymore.

For my PC backups both at home and at the office, I'm using Burp (http://burp.grke.org/) (I'm using the 1.4.40 stable version).
Burp is REALLY easy to configure, and when a backup is missed due to the PC being off at that moment, it just figures it out when the thing comes back on line.
It's capable of continuing an interrupted backup.
One possible downside, depending on your setup, is that Burp DOES NOT DO TAPE.
It does backups to disk. That's IT. If you need tapes, you need to go elsewhere.

I'm really fond of the easy-to-configure nature of Burp, but of course my needs match it's limited capabilities very well.

Both are in active development.
If you try to use Bacula, see if you can find someone experienced to help you. I don't know that I'd have ever gotten that thing working correctly on my own the first time.

Use shadowspawn.exe to avoid windows locked files

[Anonymous Coward]

You "should" be able to use shadowspawn.exe along with most other backup systems (rsync, robocopy, whatever) to be able to get pst files and such with the VSS snapshot that it makes.

For example, on my home PC, I run this command each night to mirror my "C:\" to the "E:\". :: shadowspawn.exe C:\ Q: robocopy Q:\ E:\ /NS /NC /NFL /NDL /NP /LOG:H:\nightly_backup_log.txt /R:3 /W:1 /MIR /B /E /XJ /XD "Temporary Internet Files" /XF pagefile.sys hiberfil.sys thumbs.db

Here's a longer tutorial for copying over the network:
http://xpertnotes.net/blog/2014/01/20/using-shadowspawn-and-robocopy-to-copy-in-use-files-over-a-low-bandwidth-connection/

My main goal was that I have about 1TB of data, all on my C:\, and if that single hard drive fails, then I want a copy of it on a spare 1TB drive (E:\).
The "Q:\" above is a VSS Snapshot of C:\, which gets mirrored to my spare drive "E:\"
H:\ is just another drive that I have, it keeps the log.

I ran it overnight on the first pass.
After that, it only takes only about 10minutes to mirror my C:\ drive to E:\ each night.

Re:Use shadowspawn.exe to avoid windows locked fil

[Anonymous Coward]

I used to use rsync with cygwin to run a backup the my dyndns server.
I do already have cygwin for ssh tunneling, but didn't feel like using it this time around.

Re:Automatize this!

[tehcyder]

I think it is about time they 'automatized' your job since it doesn't sound like you know what the fuck you are doing....

Why is this down-modded? It's exactly what I was going to say. If I ran a "big company" I wouldn't be expecting my IT people to be asking for advice on a public internet forum.

And "automatized" really isn't a word.

Powershell

[musicon]

I'll just throw this out there: I was tasked with the same requirement of backing up people's desktops 6-7 years ago and the solution that I went with was some home-grown Powershell scripts and using the built-in VSS service on the workstations.

I grab the bare necessary files to rebuild a workstation and then dump the backups to the user's home directory on the server (which is then automatically backed up). Take a look at the scripts I wrote

Re:Automatize this!

[SadOldTechie]

Down-modded probably because it makes too many assumptions. The question may not even be asked by "IT people". Abusive too, so fair game for -1 IMHO

Bacula

[edwyr]

I used Bacula to back up a company of ~25 users. It doesn't have all the bells and whistles of TSM. Once you figure out how it works, it's nice (and it's free). I ran the Bacula server on GNU/Linux (Redhat I think). There are clients for GNU/Linux, Windows, and Mac. Backups can also be encrypted.

UrBackup can do both files backup and image backup

[cupnoodleboy]

Highly recommend UrBackup. It can do both files backup and system image backup. The backup server can run in either Linux or Windows. Clients softwares for PC to be backup are also available for both Linux and Windows. For Windows clients, system image backup is made using Volume Shadow Copy function of Windows. Not sure if system image backup is available for Linux client or not. Overall it is a very powerful, yet easy to use, backup system, avaialble at http://www.urbackup.org/

How big is "big"?

[mlheur]

The first thing I have to say to everyone who asks me to design a backup solution is "what's your recovery solution? what are your recovery needs?". Then design your backups around that. Don't back up anything that won't be restored. You have to protect against both disaster recovery (loss of total system) and operational recovery (file deletion, corruption, historical trails). DR for a PC is usually from a stock image; OR for a PC can be managed through much better methods than PC backup.

The only thing I do at work all day every day is backup. I'm certified on one of the major commercial backup applications, two purpose built backup appliances from different vendors, and have formal training in a few other commercial applications. We have a team of 6 that manage backup for the company's data internally, another team of 5 that manage backup for our IT customers, and a couple of part-time backup admins for rogue corporate business units. We have an offshore team of 12 that support backup & storage for both internal & external 24x7 (max 3 on shift at any time). Internally our biggest backup server runs 10,000jobs/day for 1300 hosts; across all internal servers it's something like 25,000 jobs on 4,000 host. I spent almost $1M in capital this year just to refresh EOSL backup infra, and have asked for $3.1M next year to get into the 21st century for all our backup storage. My colleagues have spent at least as much in growth this year. Across the enterprise we protect 12PB of front end data. Globally we are considered a small customer to our backup software/hardware vendors. We don't do endpoint protection because it's not worth the effort.

If you really are going ahead with what you've described (I suggest you don't), then my biggest worry would be software & config deployment during version upgrades, OS re-installs and infrastructure changes. If you add another backup server are you going to have touch every PC? If you upgrade your server to a version that drops support for a given client version and the user doesn't bother upgrading can you push it down? Can you make sure it gets pushed the next time he logs in?

If you have any remote users, use a tool that does client side deduplication and incremental forever with synthetic full backups being hydrated on the server.
Know what your requirements are for portable media, and make sure the tool you use includes/excludes it as per corp policy. Will rsync of / pick up /cdrom? Set the users expectations appropriately and communicate with them what they can expect of you. Make sure everyone knows where their division of responsibility starts and ends, and make sure there's training material available that aligns with those divisions.

That said, if you really are a big company, then forget the endpoints, put your user's data on LAN shares, SharePoint, Exchange, etc and protect their data using server backup. Encrypt the endpoints in case some PHB decides to not use the LAN shares and stores corp data locally, but if he didn't put it on the LAN as per corporate policy, then let him sweat it out for losing his data when the laptop gets stolen/dropped/dies/etc. I have no sympathy for anyone that doesn't follow documented policies & procedures. They all get the stock corporate image, which gives them a personal and a team share. I don't even cheat like some of my colleagues, my laptop is not backed up to our backup servers and I've had it replaced/reimaged every 2 years or so. I keep my historical PST files in two locations. Two hours of setting up my windows preferences or importing app settings from dumps I put on my home drive and I'm back in business. Data doesn't belong on PCs - I wish they were all dumb terminals and any machine you log in to mounts your home drive, runs apps off shares or in remote sessions, and your profile follows you around. Or follow the mainframe strategy and RDP into your desktop / use VDI.

Anything less than 20 endpoints I'd consider small, go ahead and use si