So you would have thought that my first real post following being awarded a vExpert award would be all songs and praises for something to do with the big V? Well maybe it isn’t…I have noticed a lot of dialogue over the last week on twitter between a few VMware employees and VMware end users on whether DPM can add value and whether the Distributed Power Management feature in vSphere brings reduced TCO to an IT infrastructure. It appears mixed views are present, obviously the evangelists of the technology say it can be used to provide great cost savings and reductions in Opex cost of power and cooling requirements in the datacentre, others (mainly on the customer side) seem to be more negative towards DPM and advocate that strategy such as high density compute is better to sweat the underlying assets or DPM isn’t able to fit into there current IT model.
DPM technology is excellent and to be honest plain common sense, it has moved from being experimental into full blown production supportable within the later versions of ESX and now as a de facto proven product within vSphere. Core Main benefit of DPM is simple, it will dynamically turn off virtual hosts that are not needed at non peak times which is great, it avoids the cost that would have been occurred by running even vSphere hosts in an under utilised state. So I’ll get to the point do I think DPM is capability that can be used it to obtain the saving to anyone? well not really to be honest, I am in the non enthusiastic camp when it comes to DPM and the reasons I think this are as follows;
IT Service Providers
This is something that immediately springs to mind when aligned to my existing environment. Large amounts of organisations Outsource management of Virtualised environments, it is common for the agreed managed service contract to be setup over a Five year period with a pricing schedule built around things like Price of management of the VM’s in the environment, ESX host management etc etc. Organisations that outsource to managed service and hosting organisations very rarely share infrastructure with other customers. Based on this factor why should an organisation using a hosted DC care about the power costs being incurred? The cost of Power and Cooling is already present within the pricing schedule that they agreed with the outsourcer so it is extremely rare that DPM is going to be deployed in this type of environment.
Longer term and as with most IT infrastructure developments I expect service providers will start to push customers to adopt DPM to reduce there power costs (and the customers), however at the moment this is currently the case and most are subjected to older pricing models.
Organisations likely buy or leases underlying Infrastructure themselves or they potentially lease the hardware from the service provider. This of course means the organisation or the outsourcer has to purchase server hardware with CAPEX.
DPM falls short here, organisations simply don’t like to waste capital investment especially when they are paying a service provider to manage that environment 24/7 and not just at peak times. This dosn’t mean I agree with where DPM falls short here, it is merely a fact that is present in most outsourcer agreements.
This might seem harsh and vintage but some organisations are still very low on the maturity level when it comes to aligning change to the great dynamic capability available within VMware environments. Functionality such as Vmotion and SVMotion are just starting to be accepted as something that dosn’t require change. I am skeptical as to whether change boards would be happy with underlying server hosts turning on and off at certain periods of the day/night.
Additionally in hosted datacentre’s you can struggle to even be able to turn on physical infrastructure at certain periods of the day, this is mainly for M&E reasons such as potential power spikes or sudden dramatic increases of power draw from the UPS.
DPM is yet another dynamic feature within the VMware suite that will again need excessive volumes of education and evangelising from people like you and me for change boards to understand and accept.
The jury is out as to whether TCO cost savings in OPEX for power and cooling is percentage wise worthwhile the time and effort. Think of it this way, how much cost is incurred in designing, planning and operating an environment that is DPM enabled? As much as I trust VMware it is still software and still potentially prone to failure at some stage, so does a DPM environment need someone to be monitoring it more to ensure that it actually is working effectively?
Being an Architect I question whether it is more cost effective to plan for more higher density DRS pools in the first initial design phase as apposed to just installing hosts and using DPM to turn them off when not needed?
Also remember when your hosts are not being used and are turned off by DPM algorithms you still are paying for the Licensing and Hardware cost to sit basically turned off!
If DPM has true capability to reduce TCO then Maybe VMware should introduce a TCO calculator or build in something to highlight to give organisations the piece of mind?
So where is DPM suited?
At the moment I think DPM is suited to and should only be employed within some of the following areas;
- Organisations that internally manage there own Virtual environment from the Infrastructure level up,
- Are not subjected to OPEX cost for management of hosts and have to pay with facility budget for power and cooling.
- Has management and bean counters that embrace such technology
- Can demonstrate and measure the cost saving on Power and Cooling. Main crux of this is you should not install or enable technology if you cannot measure and justify the reason you are doing it.
- Has acceptance and clarity with CAB’s and other service management teams
Future hope is present for DPM to grow in popularity with the predominant growth of Cloud and multi tenant hosted environments. Cloud or shared utility models mean that turning off non used hosts within a cluster will become more realistic and crucial to the finance model of the cloud service provider, in a cloud IaaS environment the customer will only pay for VM’s and will not be concerned about wasting CAPEX, that will be the Cloud service providers problem!
Please do not think I think that DPM is rubbish, it is excellent innovation and great plain common sense thinking from VMware, I do feel though that awareness is needed to understand that it is not functionality that can simply be just ticked and used to reap reward, there are various factors that limit its potential within most larger enterprises and certain IT models.