Hey everyone, I'm looking for some insights on an RDS setup that's been causing some headaches. We recently set up a single RDS server for 9 users on a Lenovo host. This RDS VM is running with 18 vCPUs and 128 GB of RAM, and it's a straightforward session host – I assumed we wouldn't need an RDS farm, but I might be mistaken.
The users primarily operate Sage 50 (both Canadian and US versions), Chrome (for news and browsing), Microsoft 365 apps, and Adobe Acrobat. The RDS is accessed locally.
We've also implemented FSLogix profile containers stored on a separate file server VM, which is on the same physical host.
However, users have been expressing concerns that the environment feels sluggish, and Sage crashes multiple times throughout the day. Overall, the performance just doesn't seem satisfactory.
In terms of the hardware, the host specs include 2 Intel Xeon 6507P CPUs (8 cores each; 16 threads total), 256 GB of RAM, and the host OS is on RAID1 with NVMe drives (480 GB). The VMs, however, are running on RAID5 using Seagate 10K SAS mechanical drives.
My manager suspects that the FSLogix containers could be causing the slowdowns since the profiles are being fetched from the file server rather than being local, but I don't think this is the issue. Personally, I suspect the RAID5 mechanical drives might be the bottleneck, especially given that Sage 50 is quite intensive on hard disk usage.
I'd love to hear your thoughts on this issue!
3 Answers
It seems like the spinning disks are likely the main problem here. They're just not fast enough for the workload you're describing, especially when it comes to a disk-intensive application like Sage 50. Consider upgrading to SSDs for better performance.
Running performance monitoring could help identify the bottleneck. It's possible that while the RAM and CPU usage seems low, the latency on those spinning drives is causing issues. Make sure to check that as well!
You might want to look at the vCPU allocation. Your VM has 18 vCPUs while your host only has 16 cores available. That’s going to create challenges for your hypervisor when scheduling tasks. You should probably reduce the vCPUs to something more balanced. 8 vCPUs for your VM with 128 GB of RAM would be a better fit for your hardware.
I agree; that’s a good idea. I did some checking but didn’t evaluate the drive latency yet.