Post to Twitter |
Comments (6) |
Views (22616) |
As described in part 1 of this blog series, the cloud is not the answer for every enterprise or every workload. This is true based on the current economics even without considering additional factors like security, control and SLA's. On the other hand the economics do point out a major cloud advantage when it comes to short term or variable workloads. ( pointed out by Michael Keen and Billy Marshal as well ) Perhaps this is not very surprising to many, but it does help to look at the numbers to put it in perspective. The largest Intrinsic cost advantage of the Cloud is the ability to share infrastructure among multiple customers ( i.e. Multi-Tenancy and/or Multi-Instance ). This comes into play when many customers have variable workloads that are not likely to overlay at the same time. A Cloud infrastructure can load balance this workload on-demand significantly reducing the cumulative infrastructure required to support N number of customer workloads.
A Premise only solution will typically deploy the infrastructure required to accommodate the anticipated peak demand plus a factor of safety. As a result excess capacity is built into every deployment even if it is rarely ( or never ) utilized. This formula gets very expensive for many scenarios such as implementing a redundant DR solution across multiple data centers or a retailer building infrastructure to accommodate the Christmas shopping season but paying for it all year. As noted in the Cloud 101 example however, when a premise based is well utilized it can be the most cost effective solution to stay with especially if the on-site facilities can accommodate the anticipated growth.
The following Premise Plus Cloud scenario provides an optimized view of where a fully utilized premise infrastructure is used for constant predictable workloads and the Cloud is used for the variable workload. To put some simple numbers to it based on the original example, let's assume that the constant workload is roughly equal to 5 Quadcore server capacity. The variable workload on the other hand peaks at 160% of the base requirement, however it is required only about 400 hours per year, which could translate to 12 hours a day for the month of December or 33 hours per month for peak loads such as test or batch loads. The cost for a premise only solution for this situation comes to roughly 2X or $ 15,600 per year assuming existing space and a 20% factor of safety above peak load. If on the other hand you were able to utilize a Cloud for only the peak loads the incremental cost would be only $1,000. ( Based on Amazon EC2 )
| Premise Only | |
|---|---|
| $ 15,600 | Annual cost ( 2 x 7,800 from Part 1 ) |
| Premise Plus Cloud | |
|---|---|
| $ 7,800 | Annual cost from Part 1 |
| $ 1,000 | Cloud EC2 - ( 400 x .8 x 3 ) |
| $ 8,800 | Annual Cost Premise Plus Cloud |
As noted for this example the server cost of using a Premise Plus Cloud solution could save as much as 44%. This does not factor in many costs that either the Premise only scenario or Premise Plus Cloud would face but those costs vary according to the situation. So the challenge is how to identify variable workloads that can be placed in the Cloud or split between premise and cloud. The CSP ( Cloud Service Provider ) must also provide the proper infrastructure and remote administration to enable corporate IT to control and manage applications and images in the extended cloud as a virtual private network of their own. Economics aside, the CSP also needs to address the SLA's and security concerns that corporate IT has identified as prerequisites for adoption. Given the intrinsic cost savings possible as portrayed in this example, there is little doubt that CSP's will fill the gaps and the industry will move to Premise Plus Cloud solutions.
Comments (6)
Oct 13, 2008
Anonymous says:
Good analysis, it really depends on the task however. This may make me look hard...Good analysis, it really depends on the task however. This may make me look harder before buying more servers.
Oct 15, 2008
Anonymous says:
Great stuff, Chris. This concept of "Premise plus Cloud" maps very well t...Great stuff, Chris.
This concept of "Premise plus Cloud" maps very well to the concept of a hybrid cloud that I've written about from time to time. I think its inevitable that some workloads will remain "in house" while others are shipped to "the cloud". What I like about what you have written in parts 1 and 2 is that there is an economic basis to justify the hybrid nature of the cloud in an organization with enough need. Choose to distribute load not project by project, but strictly according to "cost of service". Very cool.
I would note, however, that this screams for software fluidity, the ability for applications to continue to function even as they are moved (or replicated) from data center to data center. The new "cloud OSes" from Citrix, VMWare and others will assist greatly in this respect, but to really make this work the network (especially use of TCP/IP) also has to undergo some changes. The concept of Premise plus Cloud is spot on on paper, but I think "private clouds" will have significant traction until this fluidity exists.
There is much to ponder here. Cloud Economics 101 is required reading for those making IT purchase decisions in years to come.
(Let me note the irony that the "captcha" word I had to type to post this was "limiting"...)
Nov 12, 2008
Tarkan Kocoglu says:
Chris, very comprehensive insights and details around cloud computing, however ...Chris,
very comprehensive insights and details around cloud computing, however the concept is not new as stated in a previous comment and I
am curious what the next name for this concept could be.
One important item that has not been considered yet is the cost related to keeping company specific data local to the country that can add
another chunk of costs to having either a data center in that specific country. If a globally acting organization wants to leverage cloud
computing concept, they need to consider regulations and politics for their specific branches in other countries. Usually, data should not
leave the country such as in several European countries. Considering this as part of the model is crucial.
Nov 26, 2008
Anonymous says:
Chris, Any chance you could update this model with the hidden costs? ... Sys a...Chris,
Any chance you could update this model with the hidden costs?
... Sys admin costs, facilities mgmt, data center build-out, leasing the land, renting the security guards, paying the procurement guy to buy the stuff, lost productivity waiting for your instance, decomissioning old hardware, configuring your SAN to work with the lastest server, etc.
I know that everyone else is rather congratultory on your analysis here... but I think you've missed the mark. In all seriousness, I'd love to see an update to this model.
Thanks,
Jeff Schneider
MomentumSI
www.MomentumSI.com
Nov 28, 2008
Chris Fleck says:
Jeff, thanks for the feedback. All valid concerns however I wanted to make ...Jeff, thanks for the feedback. All valid concerns however I wanted to make the point just based on incremental "out of pocket" costs as a start, too often the simple analysis is overlooked and the soft costs are discounted because they vary by the situation. ( These assumptions were identified in Part 1 )
The additional costs you point out are real however and generally favor the Cloud solution, if have you seen some industry accepted calculations for those costs, please let me know.
Good topic for a future post...
Nov 30, 2008
Anonymous says:
Chris, Agreed - it would make a great post. The value proposition changes based ...Chris,
Agreed - it would make a great post. The value proposition changes based on the unit of analysis, such as:
1. I want to move a single physical unit (server, disk, etc)
2. I want to move a set of physical units with shared overhead (rack, disk array, etc.)
3. I want to move a set of sets (closet, data center, etc.) with shared overhead (facility, security, etc.)
The analysis you performed looks about right if we're talking about #1. And for some people, that might be the right calculation. Good fodder for another day...
Jeff
Add Comment