Does vCops take into account vMotion affect on DT?

[MattG]

If I vMotion a VM from HostA to HostB,  does vCops take this into account and adjust down the DT range for HostA based on the DT of the VM and adjust up the DT on HostB since it now has a greater known load with the additional VM?

Thanks,

-Matt

[admin]

Hi Matt,

Suppose one of our dev team is best to answer this, but I will take a stab at it.

I think the answer is Yes...but over time. DT is calculated only at 1am every morning so if you have added a lot of more VM's onto that host now it will most proberly violate a DT threshold as this is not expected.

But if you have a regular schedule where u put a host into MA Mode (Like on every Friday evening at 9pm) DT will learn this as normal behavior that the other hosts will be more utilized that time.

DT is around learning what is normal in your environment.

[kitcolbert · Accepted solution]

The answer to your very specific question is "no".  vC Ops does not sense a VMotion and then adjust DTs based on that.  The reason is that the VMotion may be a good thing, or it could be a bad thing.  vC Ops should not adjust DTs assuming it's a good thing, right?  (I.e. the VMotion may push metrics outside their DTs, meaning Anomalies will increase, which may generate an alert.  THis is the desired behavior when the system is behaving differently from normal.)

The *effects* of VMotions are taken into account in DT calculations though.  So if VMs are constantly moving around on the hosts, then the upper and lower bounds for the DT will be fairly wide (meaning that it's possible there's a lot of VMs on the host, or fewer VMs on the host, depending on how many VMotions have occurred).  If the VMotions happen in some predictable way every day, then the DT bands will be more narrow to account for that.

In general, the analytics are just looking at behavior.  When VMs VMotion onto a host, the host's aggregate utilization will rise, meaning most metrics (like CPU and memory usage) will also rise.  Analytics will "see" this and take it into account.  Thus, as I said, if VMotions are frequent and continuous, CPU and memory usage will vary widely across the day and so the DT bounds will have to be wide to account for this.

Thus VMotion isn't special from an analytics point of view.  It's just another behavior of the system that is captured and analyzed.

[MattG]

Kit/Hugo,

Thanks so much for your timely responses.

-MattG

[vkaranam]

Hello guys,

The Custom dashboard works on DT. Will the Standard Dashboard Uses the DT for reporting the scores of the Health, Risk and efficiency.

Please help me

Thanks

Venkat

[admin]

Hi,

The vSphere UI Health does take into account DT for Health.

Risk and Efficiency I would think yes...but will confirm.

[vkaranam]

Thanks Hugo for the info

I will wait for your confirmation. Can you suggest me any good document on DT

Thanks

Venkat

[admin]

Pleasure,

What specific are you looking for on DT ?

[vkaranam]

Hey Hugo

I don't know anything about DT. I want to find how DT is calculated, how DT varies with respect to badges , and what r the features that DT don't support like vmotion etc

thanks

vk

[kitcolbert]

VK,

I take exception to the last part of your question: "don't support features like VMotion".  This is not true, as I mentioned in my original reply on this thread.

Let's try again.  The original question was very specific: is DT automatically recalculated when certain events (say VMotion) at *any* point in the day?  The answer is "no".  DTs are calculated once a day, at midnight or 1am or whatever based on all the historical metric data vC Ops has collected thus far.  So a particular event (again say VMotion) happening at any random time of day does NOT trigger a recalculation of DT.

I answered why in the original post but I'll state it here again: DTs are analysis of historical behavior to determine cyclical patterns that are then projected forward and applied to real-time data as it comes in.  Any deviation from those projections will result in an anomaly.  So DT is not calculated when some event occurs.  Why?  Because you want to know whether the event is normal or not.  DTs will tell you this.  The event will either cause the metrics to change to a non-normal value (and generate an anomaly) or the metrics will stay within their normal ranges (in which case everything is ok).  If DTs were to be recalculated on every event occuring, then every event would appear to be "normal", which would completely defeat the purpose of DTs in the first place.

So, to be absolutely clear, VMotion and all events and behaviors and properties and everything else of the environment *are* taken into account by vC Ops.

Now, to your actual question: we may have some documentation on our website, but I did a deep dive on vC Ops' analytics last year at VMworld and that talk has been posted on YouTube: http://www.youtube.com/watch?v=LTesggnOocE.  That should answer most of your questions about how the analytics work.  In terms of the badges, in the vSphere UI, DT is used in the Anomalies badge, which rolls up into Health.  In the Custom UI, DTs are used in the "Health" badge, which is basically equivalent to the Anomalies badge from the vSphere UI.

Hope that helps.

[vkaranam]

Hey KitColbert,

I got what your saying. Thanks for the info. Correct me if i am wrong. As per your explanation The DT in VSphere UI is used in Anamolies badge where as in Custom UI it is used by haelath Badge. Also what about Risk and Efficiency badges in vSphere UI whether they use DT?

Appreciate your help Guys

Thanks

VK

[kitcolbert]

VK,

The DT algorithms are for understanding real-time performance issues.  The Risk and Efficiency badges are both describing forward-looking issues such as capacity trending.  So the DT algorithms themselves are not used in the Risk and Efficiency areas.  However, we do have some capacity analytics that are used in those areas.

[vkaranam]

Thank you for your valuable info

Thanks

VK