TheBallsMan
asked on
Need to reimage/clone machines every 4 hours: Looking for speed, efficiency
I am CEO of a business of about 1500 employees. Right now, we have a single image for all of our 1700 machines. This "hard drive image" consists of linux partitions and windows partitions (excluding swap). We reimage the machines every week (new patches, new changes to our build).
Things have changed. We need to now reimage the machines 3 times a day (Don't ask why). Imaging takes time and is stressful on our imaging server. It takes a while to do them by how its getting done now (G4L Linux over network). Can someone provide a faster and more efficient way of imaging hundreds of computers. We've looked at G4L, Ghost, RIS for Windows (Remember, we need linux and windows). We were thinking of using rsync if that was an option. What do you guys think?
Things have changed. We need to now reimage the machines 3 times a day (Don't ask why). Imaging takes time and is stressful on our imaging server. It takes a while to do them by how its getting done now (G4L Linux over network). Can someone provide a faster and more efficient way of imaging hundreds of computers. We've looked at G4L, Ghost, RIS for Windows (Remember, we need linux and windows). We were thinking of using rsync if that was an option. What do you guys think?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
VMware Virtual Desktop Infrastrcuture is worth looking into. You run all your desktops as virtual machines within a VMware server farm and access them via either an RDP session on a locked-down box or a thin client.
The advantage to you is that the next release (I believe) includes thin provisioning - that is; running multiple desktops off one image file SAN. For 1700 desktops, you'd obviously need multiple images, but it's a lot easier to image, say, 170 desktops than 1700.
The advantage to you is that the next release (I believe) includes thin provisioning - that is; running multiple desktops off one image file SAN. For 1700 desktops, you'd obviously need multiple images, but it's a lot easier to image, say, 170 desktops than 1700.
Along the lines of the last post ... single image would be the way to go.
Given your situation, have you looked into Citrix?
if you set up a Citrix farm then you could hae the users running off one set of servers while you update another and then cut overbody over to the updated set while you then go back and update the previous farm.
Another idea, along the lines of the last post would be VMWare's Virtual Workstation.
What about leaving the main OS intact and just pushing out a new VMWare image of the desktop?
Why re-image the machines? Can't you just patch them?
If that is not an alternative, what about a combination of the two ... instread of a full re-image, something like using MOM, would a roll-back of the desktop and then afull desktop package install pushed to the desktio be an option?
If a complete re-image is the only option, then optimise and throw hardware at it.
Gig networking and as many imaging servers as you can get. If you can go one per subnet then Ghost Enterprise might be the way to go, you can multicast the image and update all the machines at one time (well, mostly, a few are bound to fail from my experience).
Some ideas.
Given your situation, have you looked into Citrix?
if you set up a Citrix farm then you could hae the users running off one set of servers while you update another and then cut overbody over to the updated set while you then go back and update the previous farm.
Another idea, along the lines of the last post would be VMWare's Virtual Workstation.
What about leaving the main OS intact and just pushing out a new VMWare image of the desktop?
Why re-image the machines? Can't you just patch them?
If that is not an alternative, what about a combination of the two ... instread of a full re-image, something like using MOM, would a roll-back of the desktop and then afull desktop package install pushed to the desktio be an option?
If a complete re-image is the only option, then optimise and throw hardware at it.
Gig networking and as many imaging servers as you can get. If you can go one per subnet then Ghost Enterprise might be the way to go, you can multicast the image and update all the machines at one time (well, mostly, a few are bound to fail from my experience).
Some ideas.
This doesn't have anything to do with the question, but can I ask how big your tech staff is? And why they are not looking in to this? Or are you just checking on them?
The company I work for only has about 500 employees and about 500 desktops and our CEO wouldn't even know what a disk image is, yet alone go out and start asking questions about how to improve the process.
But then again, if your tech staff has one image server for 1700 machines, that may be a clue as to why you are asking the question. :)
The company I work for only has about 500 employees and about 500 desktops and our CEO wouldn't even know what a disk image is, yet alone go out and start asking questions about how to improve the process.
But then again, if your tech staff has one image server for 1700 machines, that may be a clue as to why you are asking the question. :)
Gigabit is a definite plus for multiple pushes.
You may need to mix and match based on the type of computer, etc.
I hope this helps !