We need to begin backing up our servers and network here at the office. Any recommendations on resident software like windows backup or 3rd party software like carbonite? Would prefer something that is automatic, but is also easy to restore from in case something happens (and it always does).
Virtualize your servers, host them on Amazon (or in house), create snapshots of the entire VM. Now you have an automated "moment in time" of the entire system, data and all. historical archiving is only limited to your storage space. NOTE: Snapshots won't work with Domain Controllers. It breaks AD. Since virtualizing our entire classified infrastructure, we've completely moved away from traditional, file-level backup schemes. Data is more secure and disaster recovery takes minutes instead of hours/days.
In English? Not that I need to know, I've just heard my friend the IT guy say a lot of those words, and it'd be nice to know what he's talking about.
Virtualizing = Taking your entire computer (operating system, applications, data, etc) and packaging it up into a single, movable, manageable unit and running it in an environment that can host multiple "machines". Think about it this way - imagine a single physical machine (server, desktop, whatever) running many instances of Windows or Linux, or whatever, each with its own apps and data... When you do that, you have the ability to "copy" (or snapshot) an entire "virtual machine" at a point in time (everday at 3:00am, for instance). To recover a traditional system, you have to reinstall your Operating System, then some of the apps (including your backup app), then you have to restore the data and configurations. Takes longer than just copying a "virtual machine". There's many different and complex ways of setting it up, what I explained above is its most basic. Depending on your solution, you could host multiple "virtual machines" across many physical machines, for instance.
We've got a few VMs but I'm still wary of virtualizing our entire infrastructure. We don't snapshot as the space requirement seems huge from what I understand and although storage is cheap it's still expensive enough to keep me from wanting to update 13+ SCSI 10k disks. Canolli, are the initial snapshots the only big ones is every snapshot a full cone of the VM? If subsequent ones are incremental I may look into switching to that route now. We're using a backup service for incrementals now that takes the data off site. Our DC is a real box but it hosts 3 or 4 VMs. We have another 4 pizza box servers as well for various roles -- exchange, pbx, ftp / data mgmt, accounting etc. I think we're going to take exchange offsite when it's time to upgrade the software or hardware.
A question, what do you want to backup? Data or the entire machine? Incrementals of data are good if you have a need to go see what something was at a certain point in time without restoring a whole machine. Full machine backups are disaster recovery / hardware failure protection. At least that's how I see it in my environment. If someone breaks something to the point of no return we can restore the last known good instance of whatever they broke and keep everything else current.
Hi Jeff, It really depends on your budget, your applications, and how you want to manage it. How many machines are you looking to backup? How much data? What kind of data are you backing up - files, applications, databases? Do you require centralized backup management? (manage all backup and recovery from one place) How fast do you need to recover?
Virtualization (physical to virtual migration in this case) means you take a physical computer system and extract the software and hardware information into a collection of files that "plays" on a physical host system running hypervisor software, which is the framework to run (play) the virtual machine. Once a system is virtualized, that is turning it into a VM (virtual machine), the entire computer system is now just a collection of files that can be backed up using a variety of methods. Note that virtualizing allows you to run multiple virtual machines on a single physical host that when native (non-virtualized) only ran one server OS. Snapshots are a "moment in time" of the state of the VM, both machine state and software state that can be rolled back in the event of a problem with the system or software. This differs from the traditional Microsoft restore points in that a snapshot is a read-only version of the entire system, not just the system settings. There is so much more to it then this but this is a thousand foot view.
It can be very simple, or it can be very complex. You can run VMWare workstation on a desktop, or you can run something like ESX on a farm of servers, with the ability to dynamically relocate a VM from one server to another while its running.
1) We have about 5-6 servers and it'd be nice to backup the workstations too - around 20 of those. 2) Our biggest server has around 100gb of data 3) We need to backup everything. System settings, all types of files, applications, email server settings, active directory - basically a complete mirror of the servers would be nice. 4) It would be nice to keep it organized in a central location, but it's not absolutely necessary. Redundancy off site either at my house or at a 3rd party would be an added benefit. 5) Speed depends on the server that goes down. We just lost our primary file server, and while it was a pain in the ass for everyone, it wasn't mission critical. If we lost our primary application server, we'd need to be back up yesterday (though we do have a complete spare server for this server).
Is it just one server? If so, get a tape unit, or a big honkin' hard drive (or set of hard drives) and use Windows backup. Use the built in scheduler and go... Your sticking points are AD and email. Generally (and I don't know what you're using for email), these require special backup procedures because they're databases with constant access. Usually, 3rd party software uses a plugin to extract/copy data from the database while its running. Otherwise, you have to shut down the database to copy the entire file.
Virtualization on our primary app servers sound like an interesting option. Not sure if it would work with our primary fileserver which hosts AD & email. We'd like to get a carbon copy of this fileserver so it stores all system settings.
Jed, I single handedly built an entirely new infrastructure of 1200+ virtual machines along side production systems in just under 13 months. This includes multiple AD domain controllers, Exchange 2010 server, Lync 2010 server, multiple SQL servers, auditing servers, MATLAB licensing servers, KMS server, multiple Team Foundation servers with software build servers. All this in addition to fully virtualizing our Windows and Linux high-performance computing grids. I'm not even counting the 250+ VMware View virtual desktops that the department connects to through $300 zero clients instead of $1500+ desktops for general purpose computing. Virtualization of entire infrastructures is well into prime-time. We've achieved greater results through virtualization then we ever could have through native builds. We also saved over $600k in hardware costs. I'm not including HVAC and administrative costs either. Those numbers are quantifiable but I haven't calculated them.
Full backups of workstations can make this a spendy endeavor. You're going to find quick that you're backing up MP3s, photos, and shit people download from Facebook. That said, you can do it. But, you're going to need storage and time (one thing I didn't cover is that your solution must be able to complete the backup in the allocated time - full backups of 20 workstations plus 100GB of data, AD, and Email to a single tape drive ain't gonna happen in a couple of hours).
Jeff, When you go virtual, you break out each servers function into it's own VM. So one VM for AD, one for Exchange, one for application A, one for application X. This sandboxes each application from the others in the event of a system failure so you don't lose more then one resource. In the native world, you would place several server and application instances on a sigle box because of the cost associated with buying hardware. Now on todays hardware, you can run many virtual machines on a single system or a cluster of two or more systems for redundancy. For most server virtualization, you can realize a 10:1 ratio on a modern x86 64-bit Intel server. Cost savings are realized very quickly even though there is an initial up-front investment for storage and servers.
I remember when you started that project. That's really bad ass. I'd have to sit down with you and diagram all that kind of stuff out. I'm more of a network guy by default than by desire or role. With the VMs what kind of hardware do you spec to support the systems and what kind of VMWare do you use? We don't have a need for anything more than we have now, but I'd like to better understand the capabilities and capacities of virtualization. If we could buy one quad processor machine with 64g RAM and another storage device I'd load that puppy with VMs, but we're tight with money and what we have works for our needs. Can you virtualize DCs and Exchange hosts? As far as the HPC grids, I don't even know the basics of that, but it sounds like a really fun playground for geeks.
