System optimising

[yswery]

Hi All

We have a proxmox 5.4 node (running in a smal cluster) with the following specs:
* 256GB Ram
* 2 * Xeon CPU E7- 2870
* Multiple SSDs in raid
* 1Gbit single port ethernet uplink

We have around 220 LXC containers doing various things and around 40 KVMs boxes (and plans to continue to fill out the hardware)

We push around a constant of 200Mbit/s traffic throughout all the CT/VM combined on this node.

My question is: Is there anything I can do to further optimise the proxmox node?

I was wanting to potentially play around with the sysctl.conf but unsure what parts should in theory be altered, does anything have useful recommendations?

 

[Vladimir Bulgaru]

If i were you, i'd look into what bottlenecks are most likely:

1. will skip the obvious - RAM, CPU and uplink - if you feel that you're getting close to 80-90%, work on improving those
2. if you have multiple IOPS or your storage RW speed is getting close to RAID throughput, look into getting PCIe SSDs and moving IOPS hungry CT/VM to those drives
3. if you are not using ZFS yet, look into using ZFS to ensure data integrity (+added value of compressing the data and minimising data size)
4. start getting rid of single points of failure. In case you are using a HW controller - it's a good start.

Again, quite difficult to offer suggestions not knowing the specifics. Each setup is unique and in some cases IOPS may matter more, in other cases the core frequency, etc.