In my last blog I had discussed about how Continuous Performance Optimization pillar has been tightly integrated within vROps 7.0 If you have’t yet read my earlier blog, I’ll request you to please browse the link to read it so that while I start my discussions onward in this blog you can relate to it.
In this blog I’ll be discussing about another important pillar of Self Driving Operation i.e. Efficient Capacity Management. The obvious question is why do require this factor while defining self driving operations and the answer to that is – No matter how efficiently I’m running my environment but if there is no space to accommodate new workload then that environment is of no good use. So it becomes utmost important to understand the existing capacity of my data center and if there are any chances of reclamation in it.
Forecast Prediction: Starting vROps 6.7 there was an built-in intelligence mechanism called Capacity Analytics Engine integrated in it. The basic purpose of this engine was to predict future forecast as to when the available resources will get exhausted and the when data center will run out of capacity. Remember the term Time remaining in vROps that was coming from this engine. With vROps 7.0 it has been further enhanced and there are few really cool features added to it. There are two features which I’ll be mentioning here in this blog:
(i) Calendar Awareness: vROps 7.0 is now calendar aware and knows exactly when a particular month is getting started and when is it ending. This can help customers to identify exactly on which particular day of a specific month they will run out of capacity.
(ii) Exponential Decay: In order to understand this feature let’s consider the below diagram of a data center capacity forecast. In the below example it is displaying data from 01st Dec to 01st June it is called as All Historical Data of the environment. Based upon this data vROps would have derived Prediction Confidence Level for this environment. However if we take a closer look we will notice that in recent past i.e from 01st Apr there have been some significant spikes in the resource demands (under Recent Data) which may alter the prediction made by vROps and may not give accurate result. In such situations vROps 7.0 will take sample from recent data and based upon that it will carve out the future prediction of the environment.
Capacity Assessment & Cost Savings: With vROps 7.0 customer gets the ability to look into all of their Data Center (DC) & Custom-Data Center (CDC) and its available capacity. If any of the DC or CDC is running out of capacity the information is displayed right on the widget and once you select that particular DC/CDC vROps will display various constitutes of that DC object and the chances of resource reclamation.
How does it do that..?
- How many snapshot are associated with the vm
- How many power-off vm are there in the environment
- How many Idle vm are there in the environment – Idle vm are those vm which are in power-on state but not much activity is happening on it. vROps finds out about this looking at the IO operations happening on vm for a period of time
Starting with vROps 6.7 VMware has also incorporated Cost engine in it, so now while providing reclamation information vROps can also say in terms of $$ value how much cost savings can be associated with a particular reclamation action. This cost saving can also be used while architecting your environment to comply with licensing guidelines for a particular OS or Application (like SQL, Oracle). While reclamation can give customer additional breathing period vROps Recommendation tab also gives OOTB information if they wish to add-in more resources (i.e. ESXi) how much capacity they need to fill in. So this gives a clear picture of available options to mitigate resource crunch situation.
What-If Analysis: This is I believe one of the most under-rated feature of vROps. During my interactions with customer I always demonstrate them what all information they can get just by using this feature. It gives them the ability to see how their environment will look after on boarding new workload without even actually running it. In order to understand this feature let’s consider this example:
An application owner comes to you and say that there is a huge build which they need to deliver by next week and in order to do unit-testing they will require 50 additional vm of a specific size. It is now your responsibility to answer him does your environment has that much spare capacity available or not? And if there isn’t enough capacity then what best can you do to accommodate this much workload ?
I’m sure all of us would have faced this situation during our day-to-day operation. In such situations vROps What-If Analysis can play a key role to showcase you what is your current utilization and after adding this much workload how much resource spike is it going to give in existing environment.
In the below example I have created a Test Scenario on my lab environment to check if I can accommodate additional 50 vm of a particular type or not. I can either select any of the existing vm to take a sample or I can define custom value under Configure option. Once I create this scenario then I’ll save it and click on Run Scenario
vROps will run this scenario in the environment and based upon that it will show that do we have this much capacity to accommodate this additional workload or not. In my case if I onboard these vm in my DC it will reduce my Time Remaining value from 10 days to 0 day(s). In this situation I have to look for reclamation possibilities in my environment so as to accommodate this new workload.
Customer can also run this scenario to identify how much ESXi hosts do they need if they wish to on board a specific workload. Once we select Add Capacity vROps can sense by itself about the existing hardware which is currently being used and based upon that it can showcase what will be the additional capacity after adding x number of ESXi hosts. Customer can also select from the list of hardware which are available in vROps database if they wish to select hardware other than which they are currently running in their DC.
Migration to Public Cloud: In situations where customer run out of capacity and even after reclamation there is no further scope to accommodate new workload then vROps 7.0 gives OOTB cost information to run that workload in AWS or VMware on AWS. So far customer can only get cost information for these two public clouds, but in future there will be cost information available for other cloud providers as well. In a Hybrid environment (i.e. VMC on AWS) vROps can instruct vCenter server to migrate these vm to Public Cloud in order to satisfy resource demand. This is a wonderful example where vRops+vCenter are taking decision on your behalf to find out best suited placements for business critical workload.
With this I end my blog here. I’ll be publishing another blog (under Self Driving Operations) to cover on the last pillar i.e. Intelligent Remediation.
I hope it was helpful for you.. till then Happy Reading 🙂