Elastic Computing and a Sumo Wrestler Waistline

Dave Link


Last Thursday at the Interop Las Vegas conference, I sat in on an interesting panel discussion entitled “What Elastic Capacity means for IT Operations“.

What made this an interesting discussion, is the starkly different opinions of the panelists confirmed that we are a long way from consensus on the speed and growth of this emerging market.

The most insightful comments came from Michael Baum, co-founder of Splunk, who pierced the bubble of Cloud Computing Hype yet offered a glimmer of hope, including a consensus of clear ideas for which applications are appropriate for the Cloud.

To set the stage appropriately for this post, I have to give a quick background of the panelists – The discussion was moderated by Hooman Beheshti, VP Products Strangeloop and included Javier Soltero – CEO Hyperic, John Keagy – CEO GoGrid, and Michael Baum – Co-Founder, Splunk. Javier Soltero’s Hyperic recently bought by SpringSource and provides Cloud Application Monitoring solutions; John Keagy’s GoGrid provides a competitive solution to Amazon’s Ec2 offerings, and Michael Baum’s Splunk provides real world log analysis for traditional Enterprise IT and new Cloud upstarts.

All of the panelists have solutions that are positioned to assist with this amorphous Cloud Computing term that both the press and analysts have been hyping lately. I have to admit with complete transparency, we have also jumped on the Cloud bandwagon as it applies to our new EM7 G3 product.

First, a few Cloud Services Economy Building Block acronyms that may interest you: Acronyms – Examples SaaS (Software as a Service) – Salesforce / Gmail PaaS (Platform as a Service) – Google App Engine, Azure Services Platform IaaS (Infrastructure as a Service) – Go Grid, Amazon Web Services After the 3 panelists walked through their slide decks without even a single question from the audience, the moderator proactively asked the audience a simple question: Does anyone have any platforms running on App Engine or Go Grid?

Not a single person of 50+ seated raised their hands. The audience did not yet seem to be engaged in the presentation and or they were well behind the Cloud curve that everyone has been hyping.

My gut feeling was the Audience had selected this session in an attempt to find a little practical advice about how to get started. Michael Baum quickly summed up the audience’s (filled by traditional Enterprise IT technologists) lackluster mood by proclaiming that Enterprise IT and Cloud Computing are analogous to foreign antibodies trying to get together… his conclusion: they are probably still years apart.

Michael went on to talk about where are we all actually spending money on IT, compiled from several Forrester reports. In 2008 $2.1 Trillion dollars was spent on IT in the following categories:

  • 12% Security
  • 13% Capex
  • 33% Management costs
  • 25% Apps development + maintenance
  • 2.5% Storage
  • 17% Desktop management

His conclusion – you have likely heard this before, but the big opportunity in IT is to reduce the ongoing costs of managing existing IT infrastructure.

To be successful as an industry (Elastic Computing) they really need to breakdown capabilities and define the state of the industry – e.g., what kind of applications are a good profile to put in the Cloud, what kinds of apps and IT capabilities are applicable for the Cloud today vs.

3 to 5 years from now? No one is doing a great job at breaking down a matrix that would help guide IT professionals. One of the key problems with the Public Cloud today is that Cloud Service Providers are not providing the instrumentation visibility required to confidently run Enterprise apps.

James Urquhart (Cisco Market Manager for Cloud Computing and Virtualization – Data Center Solutions Group who is a prolific blogger on his own cnet blog, The Wisdom of Clouds, and on the Cisco blogs) was in the audience and he had some interesting comments.

We are at a place right now where education and demystification are necessary…We need to continue to define where the Cloud is today and how can we translate that into what people running infrastructure in mainstream industries can use… what will work today.

We have seen that IT operations have gotten very tactical. In certain situations, it comes down to people trying to protect their jobs. The general consensus is that unless you can give me something that gives me revenue assurance or dramatic costs savings – how can you do it cheaper, faster, better – I will stay put with my current solutions.

There are certain application classes for which the Cloud is a great fit, while there are plenty of others where it just doesn’t work. Figuring out the economics right now of internal vs Cloud hosting is very difficult.

Where IT budgets are strapped, new Web 2.0 applications represent the best candidates for Cloud services, to get them deployed with cost-effective scale and with better agility, especially if you have a stateless application.

However, that is a very low percentage of traditional IT applications that live in operations today. Additionally, traditional IT  applications have not been designed to easily scale horizontally, when it comes to web servers and application servers, so they are not perfect fits for Public Cloud architecture that supports elastic growth.

It is way too early to tell where we are headed. Javier Soltero’s advice: Get a good enterprise architect. It is not uncommon to hit some sort of Cloud Service limit and then have to migrate away from the Cloud offering.

Splunk hit a disk limit with Amazon and had to move back in-house because it didn’t work as the application grew over time. It’s a marketing myth that the Cloud can work for everyone and moving from public Cloud Service provider to another provider will be seamless. Michael Baum said:

Every Cloud Service Provider has made their own architectural tradeoffs, and they have made certain product roadmaps that will mean that you can be left out of control for your future as it relates to these services….Clearly a lack of standards is a real problem for broad, extensive acceptance of the Cloud.

Final comments from the panel: Michael Baum: IT Op ex is growing at 2.5 times faster than capex spending, and there is no slowdown. Everything is getting more complex.

It’s not getting easier to manage service delivery with the Cloud. Think about Cloud as another tool in the quiver of an IT warrior. Cloud is part of an overall strategy.

John Keagy: We are now at the peak of IT complexity and everything gets easier from this point forward because of the Cloud.Perhaps the Cloud is best suited for expansion / skunk-works applications where you need to quickly deploy new application resources.

Takeaway – So what’s with the title? Elastic Computing, Cloud Computing or whatever you call it, is an amorphous term that we continue to struggle with to get to a common definition.

Both a Sumo Wresler’s waistline and the Elastic Computing definition are moving targets which seem to be continually expanding before our very own eyes. Stay tuned for more blog posts on this topic!

Share This Post

Most Popular