consolidation ratio

[vcpguy]

Under the Dashboard - Efficiency badge, I see that Density VM:host ratio is shown as 418.6:1
The way I interpret - This host can have 418 more Vms. I understand this number depends on the host configuration and  workloads but still 418 Vms per host does not makes sense to me.

I think, vcops should do a better and show more realistic numbers. They should thiink about the cores and sockets and show numbers which are as per the best pratices. Showing such high numbers can be a problem, if the upper management sees them.

What are your thoughts ?

[admin]

Could you elaborate on why specifically the number does not make sense to you?

I guess you are implying the limitations of the vShpere infrastructure itself. But I'm not sure.

Anyhow, If you consider some of the underlying vSphere/hypervisor limitations it is entirely possible to hit VM densities in VCOps beyond current VMware capabilities. For instance, ESX 5.0 can only support up to 512 VMs per Host, but based on the VM workload analysis VCOps could tell you that you might achieve density of, say, 1024 VMs per host.

What does that mean in VCOps 5.0? It means that the results of the worlkload analysis are just that. So if your infrastructure supports it that is how many workloads you could put there.

Note: different ESX versions have different limits but one trend is obvious: the limits are always growing and catching up with the industry demand.

Future versions of VCOps may take the infrastructure limits into acount but the feature is currently somewhere in the backlog.

[vcpguy]

Welcome to the forumn. All though I don't see a vm icon next to your name. I am assuming either you are working with VMware OR you have very close knowledge about future developments.

From a VMware admin point of view it puts them on a tight spot. vCOPS is a VMWare product and in my humble opinion, you should pratice what you preach. Today, when Management sees these numbers, they start to doubt the consolidation ratio and they think, more money can be saved if we add more VMs of similiar capacity on this host. But who is going to think about failure of that host, who will consider Vms per socket? This just adds more confusion.

So, if the software recommends something, I think they should follow the best pratices. I understand there are lots of things that needs to be considered before those VMs per host are displayed but it should not show such outrageous numbers.

Thanks

[admin]

Sorry, I thought you meant vSphere infrastructure limitations but now I'm not sure if I guessed the source of your concerns right and what exactly makes the 400 number of VMs "outrageous".

I'm not a VMWare admin and it is not obvious to me If I understood you right or completely missed the point. But I work in VCOps R&D so I might be able to help. 😉 Well... hopefully.

- Could you tell me what specific "VM per socket" best practice VCOps should honor?

- What do you mean by "host failure"? Which best practice would you like us to follow?

- Could you refer me to the list/examples of best practices VCOps should account for?

Cheers.

[vcpguy]

The way, I look at vCOPS is a product which will recommend the users with best pratices. I know, VMware has purchased this product and made changes to it but still I think we can make improve it considerably.

Could you tell me what specific "VM per socket" best practice VCOps should honor?

Please refer to this link. This article is written by http://searchservervirtualization.techtarget.com/tip/Sizing-server-hardware-for-virtual-machines

Here the max number of VMs per core is about 10. I am sure,if we start to dig in VMware's manuals we will find some number which will be close to 10.

- What do you mean by "host failure"? Which best practice would you like us to follow?

- Could you refer me to the list/examples of best practices VCOps should account for?

When, we initially deploy vcops we shoudl consider the vSphere version and accordingly change the suggestions from vcops.

Ideally, it will be nice during initial deployment, there shold be a widget/Badge asking us to about the cluster its requirement its design. Each and every environment is unique and every customer has their own requirements, so we should ask failover capacity, we should look into best practices and accordingly make suggestions. In vSsphere 5, per the config max, the max no of VMs per cluster is 3000. If the end user has a powerful system with 12-15 hosts per cluster, I dont see a point we telling them that they can host more than 3000 VMs in this cluster.

My few suggestions for vCOPS

1) Need more How-to's. It should be very specific to the topic and not a 150-200 page guide.

2) The interface should be more user friendly. right now in my humble opinion it is overwelming to see all the numbers.I think vfoglight does a good job here.

3) When, vCOPS is making a suggestion, it should not bombard the end user with too many numbers/graphs. It should be direct and to the point.

4) Preferably, we should also link the suggestions with the KB articles from VMware's site.

5) There should be more online videos/help about vCOPS. Avoid putting it on you-tube. Lot of companies block them.

6) Allow users to download entire vCOPS suite on trial basis and not to force POC.

7) I think, for mass usage of vCOPS, the adapters that can help them to link to various other system should be made and given free of cost to the community.