Bad performance copying files

So we’re having an issue with our Nutanix cluster and wanted to ask if anyone else has seen this or can duplicate it on their cluster.

We have a 3 node cluster on Dell XC430 servers. 2 480SSD and 2 4 TB SATA drives / node. They’re networked using 10GB SFP+ cables to 4 Dell switches that are stacked. Running ESXi 6.5. 1 big container. Jumbo frames are not being used.

Issue: 1 VM with 2 drives: C: and 😨 When we try to copy a large file from C to D or vice versa, we’re getting anywhere between 0 – 40-50 MB/s performance. In addition, the disk latency goes WAY up to around 200-300.
This does NOT seem normal to me. Can anyone explain this or help please??

Nutanix and Dell engineers seem perplexed. No one can tell us if this is “normal” but I can get better performance copying to a USB 3.0 device than I can from drive to drive on this!

3 replies

Badge +1
I had something similer, not exactly same...
It was nothing to do with Nutanix, it was ESXi 6.5 thin provisioned disk.
After I migrated to thick eager zero and paravirtual went ligtening...

Badge +1
Have you seen info at this link?
Badge +1
i have similar situation, but is SQL DB create full backup. some general question below concept/knowledge that i confuse.

1. How much SSD & HDD ratio is consider best performance?

2. if I have small SSD size but huge HDD size, I won’t be able to get best performance for transfer/backup?

3. Nutanix node need to depend of the SSD as entry point of the overall throughput?

4. if one of the SSD faulty in CVM Node, Will the node performance slow or it will failover to another node?

5. storage node which don’t have SSD storage space, will be data Input and Output impact with low SSD space CVM Node?