Is Your Datacenter Truly Elastic?By Mike Berthiaume on January 28, 2014
In today’s modern datacenter efficiency goes hand in hand with the ability to be elastic – defined as the ability to quickly respond to the needs of the business by adjusting the supply of datacenter resources in accordance with their demand.
The theory behind this makes perfect sense and infrastructure technologies have helped the industry move us closer to making this reality.
There are two fundamental components required to make elasticity a reality. The first is abstraction. Abstraction has become an inherent part of the modern datacenter. Almost every technology implemented today has some level of abstraction. Server and Desktop Virtualization are an abstraction of datacenter compute resources, Storage Area Networks (SAN) and Network Attached Storage (NAS) are an abstraction of physical data storage and Converged Fabric is a combination of data center network abstraction and an additional layer of physical computer abstraction. Interestingly, the entire notion of “Software-Defined” is nothing more than a fun marketing term to describe this concept.
So abstraction provides us with the flexibility to “do” things in software, but until now has required people to move the levers and knobs in order to gain some efficiency improvements.
As mentioned, true elasticity requires two components – so what is the second? The answer is simple – Control. We need an automated system to make decisions for us based on the demands our applications place on our datacenter components. Once we have a control mechanism in place we can realize the benefits of abstraction within our datacenter components and truly become elastic Let’s describe this in a scenario:
Scenario: In our Datacenter, there are several clusters configured in a vSphere datacenter infrastructure, each providing different services critcal to our line of business. The underlying converged infrastructure is Cisco UCS and a NetApp SAN as a repository for all of the virtualized applications. Each component of the infrastructure provides different layers of abstraction.
Our control mechanism is VMTurbo Operations Manager which has defined all of our components (virtualization layer, converged infrastructure and SAN) and is currently controlling them in a desired state. VMTurbo has determined that one of our clusters is significantly overprovisioned and inefficient, and has driven actions to evacuate two hosts in this under-utilized cluster. Ultimately our control system will suspend these hosts and remove them from vSphere. Here is where it gets interesting – because our control system understands the constructs of our converged fabric (Cisco UCS) when the resources are removed from vSphere, our control system “disassociates” the UCS service profiles from the physical blade servers and adds them into the server pool, essentially making them available for use.
The following day, a new application ”goes live” in another production cluster in our vSphere datacenter. Our VMTurbo control system “sees” increased memory demand and impending memory contention across the cluster. Based on this information, VMTurbo decides that an additional host is needed. This decision drives automation through the Cisco UCS abstraction layer to provision a new service profile based on the service profile template that was used to provision all of the clusters’ original blades. VMTurbo powers on the blade and through integration with Cisco UCS Manager API’s and VMWare API’s, the host is provisioned and customized according to the requirements of the contended cluster. Once the process is complete our control system recognizes that the host is available for consumption and begins to migrate workloads to prevent a contention event.
The scenario above describes how we can achieve elasticity in the modern datacenter through the combination of abstraction and control with VMTurbo Operations Manager TODAY. It is apparent that elasticity has a significant value proposition and is imperative for organizations to improve efficiency and agility ultimately resulting in a significant savings to the bottom line.