RDS, VMs and Dells, Oh My!
-
Looking at a bear in the face and not sure what to do. We have 3 older Dell R610 servers that need to be replaced. We also need to do quite a few updates/upgrades, I think. I'm trying to make it as cost effective as possible without hampering future growth and/or making us do things all over again in a few months or a year. The main thing that is driving this is, we have ran into sluggishness with our current setup. We are running 3 RDS servers with a session broker doing the load balancing. 12 branches with 60+ users total are accessing them on a daily basis. Here are what I think we may need to upgrade at the moment:
- Dell R610 -> R430 or ?
- vSphere 4 -> 5.5
- Workstation 8 -> Workstation 11
- Equalogic firmware
- Win2k8 server -> Win2012 or Win 10
My initial thought was to just buy new servers and replace the old ones, but that would leave us with an old version of ESXi. Thanks to @scottalanmiller that idea was quickly removed from the options. Any helpful tips or suggestion on this? We are also looking for a consultant that maybe able to help us, but I wanted to see if any of you may have a experience in this area and could point me in a good direction.
I'm sure this isn't enough info, but please ask for more clarification or details.
Thanks
-
Have you identified the source of the sluggishness? Is it IO, CPU, memory?
-
@scottalanmiller said:
Have you identified the source of the sluggishness? Is it IO, CPU, memory?
The CPU is getting hammered at least that is one area that I am for sure of now. We have ruled out the IO and Memory is/should be sufficient.
-
Another reason for thinking about upgrading our ESXi license and moving to Win2012 is that we can't increase the amount of CPUs used on the VMs with our current setup. We are limited.
-
With ESXi 6, you have quite a leap to make. And Windows 2012 R2 is a decent leap too. Just in software, that will help some.
-
If you move to newer hardware, have you considered going to fewer hosts? Once you are taking the time to invest in new, you have a chance to rethink the design. Why three hosts instead of two or one?
-
@jevans said:
@scottalanmiller said:
Have you identified the source of the sluggishness? Is it IO, CPU, memory?
The CPU is getting hammered at least that is one area that I am for sure of now. We have ruled out the IO and Memory is/should be sufficient.
That's something to beat the users on, not upgrade and spend tons of money on hardware.
60+ users on a terminal server is nothing. I have had Citrix farms with 200 users and 15 different servers running against it on much older hardware than you got. You need to take a look at your workload and determine what is causing them to suck down the cycles.
-
@scottalanmiller said:
If you move to newer hardware, have you considered going to fewer hosts? Once you are taking the time to invest in new, you have a chance to rethink the design. Why three hosts instead of two or one?
We also run many other VMs on these three servers. About 8 VMs per host. We have a mix of Windows and Debian servers. The majority of our servers are VMs. I'm not sure two would suffice. Also, we want to be able to have the ability to move everything off of one host in the event that we need to replace it or repair it and still have enough resources for the VMs.
-
@jevans said:
@scottalanmiller said:
If you move to newer hardware, have you considered going to fewer hosts? Once you are taking the time to invest in new, you have a chance to rethink the design. Why three hosts instead of two or one?
We also run many other VMs on these three servers. About 8 VMs per host. We have a mix of Windows and Debian servers. The majority of our servers are VMs. I'm not sure two would suffice. Also, we want to be able to have the ability to move everything off of one host in the event that we need to replace it or repair it and still have enough resources for the VMs.
Two will definitely suffice. The three that you have can easily be compressed down. Those are not big servers. SMBs are often shocked by how little they actually need. I'd bet you a pretty fancy dinner I could get you a single server that would do a better job too
-
@PSX_Defector said:
@jevans said:
@scottalanmiller said:
Have you identified the source of the sluggishness? Is it IO, CPU, memory?
The CPU is getting hammered at least that is one area that I am for sure of now. We have ruled out the IO and Memory is/should be sufficient.
That's something to beat the users on, not upgrade and spend tons of money on hardware.
60+ users on a terminal server is nothing. I have had Citrix farms with 200 users and 15 different servers running against it on much older hardware than you got. You need to take a look at your workload and determine what is causing them to suck down the cycles.
There are only a few programs and services that our users actually use or have access to use on the terminal servers. AIX(uses SSH), Office 365(2013) Intranet and limited access to intranet. That is about it.
-
@PSX_Defector said:
@jevans said:
@scottalanmiller said:
Have you identified the source of the sluggishness? Is it IO, CPU, memory?
The CPU is getting hammered at least that is one area that I am for sure of now. We have ruled out the IO and Memory is/should be sufficient.
That's something to beat the users on, not upgrade and spend tons of money on hardware.
60+ users on a terminal server is nothing. I have had Citrix farms with 200 users and 15 different servers running against it on much older hardware than you got. You need to take a look at your workload and determine what is causing them to suck down the cycles.
My Sys. Admin and I think it could be an issue with Office 365, mainly Outlook, and having most, if not all, users running that at the same time on the server.
-
Well, we actually found out what was doing most of the CPU hammering on one of our terminal servers. Adobe reader was running in the background. User could't see that it was running. We found it by using a big comb.
https://www.youtube.com/watch?v=g4OBUupicWg -
Don't forget to force logoffs after an amount of time. Usually Acrobat only goes nuts when its been running for a long long time.
-
@PSX_Defector said:
Don't forget to force logoffs after an amount of time. Usually Acrobat only goes nuts when its been running for a long long time.
We have it set up to force logoffs, but I'm not so sure it is working like it should(obviously
). We have to manual kick users off, quite frequently. Truthfully, I think the terminal servers and the session broker were set up incorrectly. I'm sure a few other things were missed as well. This was set up before I was hired. I really want to tear it down and build it out, correctly.
