posted

1 Comment

Welcome to your lights-out datacenter!  With vRealize Operation 6.6 automated workload balance is easier and more controllable than ever.  This latest release of vRealize Operation gives you the ability to balance workloads across clusters and datastores, simple controls to govern how much balance you want based on your business needs, three ways to activate it (automated, scheduled or manual), and a powerful dashboard from which you can view and regulate everything related to balance.

Why is workload balance so important?

Workload balance ensures all clusters have enough resources to avoid future contention.  Contention is the bane of any VI/Cloud administrator, because it means applications and users are adversely affected.  Do these questions sound like something you have faced in the past?

  • I don’t want to have a contention issue with my business critical applications.  How do I ensure that clusters don’t get filled up beyond a target utilization level that will ensure there are enough resources for all?
  • If clusters and datastores get full, how can my team build confidence to move VMs around that will guarantee that the VMs and the applications get the resources they want?
  • Budget cuts are happening, I must ensure Windows and Oracle license costs are contained. Can I limit their movement to a few set of clusters that I pay for?

Workload balance attempts to prevent hot spots by intelligently spreading VMs across clusters and datastores.  If a cluster is filled to the brim with VMs, any resource spike will cause contention in the cluster.  However, if we balance the VMs across clusters, giving each cluster has a little bit of wiggle room, they can better handle with any sudden resource needs. Think of workload balance as an insurance policy against resource contention. Automating this process is KEY because it allows the system to automatically adjust when balance is needed, which means healthier applications, which means more time for you to concentrate on more strategic work and not application babysitting.

Wait a minute doesn’t DRS do that?

Distributed Resource Scheduler (DRS) and Storage DRS work to ensure balance and fight contention at the host level where as vRealize Operations workload balance does this at the cluster level.  You can think of it like this, DRS is the best solution to automatically mitigate contention WITHIN the cluster and vRealize Operations is the best solution to automatically mitigate contention BETWEEN clusters.  Together these create one tightly integrated cohesive solution working together to ensure your applications are always getting the resources they need.

How do other niche product move VMs?

Other products that claim to deal with contention will FIGHT with DRS causing a “ping-pong” affect.  This happens when their product moves a VM into a host, but DRS quickly determines its not the RIGHT host for the workload and then moves it to an appropriate host.  This causes the VM to be moved multiple times before it is properly placed. This is a waste of overhead and if the ping-ponging gets bad enough can actually CAUSE contention.  To resolve this, these products have you turn off DRS all together leaving you without the benefits of DRS like host maintenance mode or affinity and anti-affinity rules or many others.  Its like someone giving you a new pair of gloves and then asking you to cut your hands off.  It just doesn’t make sense.

So how does automated workload balance work?

vRealize Operations workload balance runs at the datacenter level watches the CPU, memory and disk space across the different clusters therein.  When it determines clusters are out of balance it steps in to provide its recommended balance plan highlighting which VMs should be moved to which clusters.  It also provides a view of the resource utilization before and after the balance occurs so you can visualize the benefits.  Again, this entire process can be automated and scheduled as we will see later to make it hands-free.

Once you accept the balance plan, vRealize Operations workload balance begins moving the VMs to their new cluster and allows DRS/SDRS to determine the proper host/datastore within the cluster in which to place the VMs. This mean no “ping-pong” affect.  We also get the added benefit of leveraging the DRS processes that ensure VMs are running on the right hosts (e.g. HW Compatibility checks, HA & Admission control policies, Reservations, Shares and Limits).

How do you control balance?

vRealize Operations workload balance complete control over how evenly you want to spread your workloads across clusters.  It provides two simple to use “knobs” that regulate how the balance process.

The first is the Balance Workloads slider which simply states how aggressively I want to pursue balance for a given datacenter.  A setting of CONSERVATIVE means not worry about balance until one of the clusters starts to get too busy to handle the load.  This might be something you would use in a very dynamic environment.  On the other hand, the AGGRESSIVE setting tries to keep things as closely balanced as possible.  Obviously, this later means more VMs will move, but will also mean the clusters will be better equipped to deal with any recourse spikes (remember this is an insurance policy again contention).

 

The Cluster Headroom setting allows you to control “how full is full” for the clusters and datastores.  This allows you to set a percentage of free space you want kept available and ensures clusters have a resource buffer for CPU, memory and disk space for which you are comfortable.  When a cluster breaches that Cluster Headroom barrier it is a good indicator that a rebalance may be needed.

When does workload balance take action?

There are three ways to engage the workload balance process and begin moving VMs manual, automated and scheduled.

The manual method can be run directly from one of the rebalance alerts which indicate a workload balance is needed based on your control settings for Balance Workloads and Cluster Headroom above.  It can also be run from the Workload Balance dashboard which gives you a quick view of the current workload of each cluster so you can determine if a rebalance is desired.

Either way it’s simply click the REBALANCE button, review the balance plan and hit START to run it.  Easy!

But running this process manually is old-school and in today’s datacenters we need to automate as much as possible.  Workload balance in vRealize Operations can be automated on a datacenter basis by simply switching the Automate flag to “yes” in the policy.  This means you can automate it in your test datacenter and not in production or visa-versa.  This means the next time a rebalance alert is generated instead of you having to deal with it it’s sent to the automation framework of vRealize Operations and automatically remediated.

The third option is the best one yet in my opinion.  Instead of manually inspecting your datacenter balance or waiting for an alert to fire and automatically fixing it (which could happen in the middle of a work day) why not create an ongoing rebalance schedule?  vRealize Operations allows you to ensure balance across your clusters using a schedules process that can be run during your standard maintenance windows, which is when these types of actions should be taken.   In a very dynamic datacenter you may want to do this once a week or once a month in a more static environment. You should of course still use the automated method as a backup in case its needed, but with the scheduled rebalance hopefully it never will be.

Everything in one place

Finally, vRealize Operations 6.6 provides you one place to go and do everything associated with balance.  The new Workload Balance dashboard gives you the ability to view current cluster balance in each datacenter, access to the workload balance settings, the ability to set up a rebalance schedule, and the ability to run a rebalance action at any time.  It’s your one-stop-shop for automated workload balance!

 

Try vRealize Operations 6.6 today and experience how automated workload balance can change your environment to a contention free lights-out datacenter!  Stop babysitting your applications and free yourself up to start thinking about bigger more strategic activities in the workplace…or maybe take a well deserved vacation.

If you want to see it all working you can watch this Automated Workload Balance video.