can you please describe the nature of the file you were copying?
A VM, migrate from one ESX host to another. ALL 1GB NIC, ALL same segment, ALL same ESX build
which metric was reporting 15MB/s?
Network utilization on the source and destination host. The calculation is flat out wrong.
what was vFoglight saying for the same period, same metric?
I have long suspected vCenter was inaccurate, because the math never works out. If a file takes 9 minutes to copy, and its 36GB, how can that be 15 MB/s speed? Obviously not even close to being accurate. Same with the disk usage as well on Vizioncore backup, I can do a backup in a certain amount of time, but if I watch the ESX where the backups are taking place and the ESX where the VM resides they show a different value than what is actually being calculated. Besides which 2 different appliances were reporting the same thing (not to mention the value that I can do with a simple calculator) are ALL showing vCenter is wrong. Maybe it's right for SOME things, but if it's not right for EVERY performance meter, then that's suspect for EVERYTHING, since it's not reliable.
was vFoglight monitoring concurrently, or a separate iteration of the test?
Same time.
if available, what was esxtop saying for that metric?
That's the other thing, on vCenter watching one metric for NIC let's say. And disk transfer is the other metric. How can a Network transfer show 40MB/s yet disk show something that doesn't make any sense.. I am aware of caching, but to be like 20 or 30MB/s off? Cache can do SOME of that but not that much. Also esxtop (which is more accurate, and reliable) is showing differently values that were more reasonable than vCenter.
This is a NEW vcenter Install, as recently as maybe 3 or 4 months, and I have noticed with previous vCenter installs as well, but I never really investigated until recently when I was trying to isolated another issue with vCenter and then I started getting this information.
On a side note, I disabled performance stats, and now vCenter runs better (maybe it's my imagination) but I would rather have the performance of vCenter and let a 3rd party do the metrics, even if all things are equal, letting a 3rd party tool gather metrics is better than vCenter (especially when there is an impact on performance) makes more sense.
And another side, side note... I just realized that vCenter status hardly ever (and there are multiple posts on this) is error free. There is always LDAP, rollup stats, or heal status error messages, that bugs me that even on a new install and reinstalls and after following instructions how to fix it, there are problems around the performance stats, so that to me shows there are glaring flaws, and this is just one incident.
.png "jhanekom"){kind=avatar}
