So how much do you run your cluster memory usage? Currently we are at 50% usage but still adding VM's. My boss wants to run this to 100% and doesn't understand why I keep telling him this is a bad idea. I am trying to keep it at 2/3 to 3/4 usage of capacity. Am I right or wrong on this?
I am doing a bad job of explaining this to my boss who doesn't seem to believe me so do you guys have any resources that can help me explain this?
Right now we are running application servers and no core functions but we are looking to migrate our DHCP servers. We have 1.1 TB of memory capacity running at 51% usage. CPU usage is not a problem running at about 12%. Storage is also fine. We are running 6x Dell 740XC if that helps.
Question
Cluster Memory Usage
This topic has been closed for replies.
+10As a general rule of thumb, you never want to run over your HA limit on RAM and especially storage. You cannot over commit RAM anyways on AHV (For Now), but you can on ESXi. Basically you obey the N+1 mindset. If you have node awareness, you'll never want to provision more than your HA can handle of losing a single node. If block awareness, you'll never want to over provision over a full blocks worth of resources on memory and especially storage and so on and so forth. On any storage platform, once you exceed 80% storage utilization, it generally starts impacting IO capabilities. Some platforms "swear" you can go up to 90%, but you should make it a goal to never reach that high.
I would spin it as "To ensure we maintain optimal performance, as well as ensuring we have resources available in the event of hardware failure, this is why we can't provision to complete capacity." He's likely getting pressure to use the resources paid for, but I would document out why what he's wanting is a bad idea and work with support of the reference documentation for best practices. You'll likely find a lot in the portal, and if all else fails online too.
I would spin it as "To ensure we maintain optimal performance, as well as ensuring we have resources available in the event of hardware failure, this is why we can't provision to complete capacity." He's likely getting pressure to use the resources paid for, but I would document out why what he's wanting is a bad idea and work with support of the reference documentation for best practices. You'll likely find a lot in the portal, and if all else fails online too.
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.