Oversized VMs report help

[LittleNickey]

Hi all,

We are looking at a migration project (to move to new HW) and want to take the oppertunity to resize the VMs in conjuction with this. The "Oversized VMs" report is of great help for this, but what interval is best to use?

As I understand it, when selecting "days" it uses more measure points than when using for example "months", thus better catching peaks. For example, a VM might be assigned 32GB RAM but normally uses no more than 16GB, but at peak times it needs 24GB. Will I have a better time catching such peaks by using 30 days instead of 1 month, so we don't lower it to 16GB and immediatly face issues?

What are the pros/cons for the different periods and what best practices/recommendations are there? Does anyone have experience with this?

[RomanK]

Hello LittleNickey,

As far as I know, there is no additional logic in period filters. I've checked in my lab and see the same data and graphs for 30 days and for 1 month.
If there is a huge discrepancy I'd recommend opening a support case and providing case ID in this thread so we could collect all the additional details and investigate it accordingly.

Answering the second question, there are no specific best practices or tricks. If you know that your VMs are relatively stable for a few months - check the data for these months. Rather, the situation is much better visible from your side.

Thanks

[jorgedlcruz]

Hi all,

We are looking at a migration project (to move to new HW) and want to take the oppertunity to resize the VMs in conjuction with this. The "Oversized VMs" report is of great help for this, but what interval is best to use?

As I understand it, when selecting "days" it uses more measure points than when using for example "months", thus better catching peaks. For example, a VM might be assigned 32GB RAM but normally uses no more than 16GB, but at peak times it needs 24GB. Will I have a better time catching such peaks by using 30 days instead of 1 month, so we don't lower it to 16GB and immediatly face issues?

What are the pros/cons for the different periods and what best practices/recommendations are there? Does anyone have experience with this?

Hello, I speak from memory, but I am aware that we have much more sample data for a week, every 5 minutes in normal dats collection mode, and for the rest of period after we should show every 2 hour aggregated data. Here is some info:
https://helpcenter.veeam.com/docs/one/r ... mance.html

I would need to check if this every 2 hour period is forever, or we even roll up the data to 24 hours after x months, like VMware does.

I am not aware of any way where VeeamONE, or for that matter VMware vSphere can give you low interval sample data like every 5 mins, for a long period of time, as the fragmentation and store of this data could be quite heavy.

VMware offers every 20 seconds data for an hour, rolling this to 5 mins later on, and keeps reducing it, so you can not see granularity for a long period of time I am afraid https://docs.vmware.com/en/VMware-vSphe ... CF49B.html

Keep us posted with the migration. Another interesting tool is Capacity planning reports, and the Deployment Projects, you can grab a bunch of VMs on a single project, and produce and estimation of resources based in consumption, with detailed reports, quite nice.

I will check the details on the stored procedure with different data, and let you know, will reply next week.

[LittleNickey]

Thanks for the replies!

@jorgedlcruz much appriciated, looking forward to that confirmation. My guess was that Veeam ONE used the same statistics as VMware, but it sounds like the vSphere Raw Performance Data report differs from VMware. Does that mean that also the Oversized VMs use the same sample size, or could that be alligned with VMware metrics (every 30min for a week instead of every 5min)?

[jorgedlcruz]

Hello,
That is what I want to double check on my lab next week. I know we have something similar to VMware vSphere for the ONE Client, but want to check the report.

Will answer next week around thursday

[jorgedlcruz]

Hello, I have some news! Basically for the RAW performance report the helpcenter is right, with the addition that it depends of your data collection mode, let me put it in writing:
Typical

• 5 mins chunks for a week, only selecting the option called last week specificially
• 2 hours chunks for everything after

Advanced

• 30 mins chunks for a week, only selecting the option called last week specificially
• 2 hours chunks for everything after

Link - https://helpcenter.veeam.com/docs/one/r ... ml?ver=120

Now, for oversized VMs Report, or for other time selections that are not "last week", we only have every 2 hour chunks forever, meaning it doesn't matter the time selection.

Veeam ONE Monitor behaves differently, and you will see there same as VMware vSphere Client, last hour every 20 seconds, etc.

Hope it helps!

[LittleNickey]

@jorgedlcruz thanks!

I missed the notification for your reply last week, but that clarifies a lot.

Basically, going for 1 month rather than 1 week should give the same average but for a longer period of time, thus making it more "correct" since a VM during one week may have extraordinarily high or low load, but that should cancel out when checking over a month instead.