Welcome to my 2^nd TTITITA (Ten Things I Think I Think About…) post.  The concept of "I think I think" is used to reflect the fact that things change so much in IT, I’m never 100% sure what I think.   I received a bunch of good feedback on my last post on Agile development.  This time I’m going to hit another hot topic – cloud computing.

Cloud computing – the first problem is you can’t even get a solid definition of what it is.  Wikipedia’s entry is a good start.  However, it’s hard to get a handle on what is really new and different about it.  In Accenture’s Internal IT, we have a private cloud that is used for our email and collaboration platform (i.e. exchange and sharepoint) – although much of the rest of our infrastructure is virtualized, I wouldn’t consider it cloud.  We have also used a few public clouds ranging from Software as a Service (such as Taleo and Workscape) to Platform as a Service (such as Microsoft Azure) to Infrastructure as a Service (Amazon AWS).  The thoughts below are based upon those experiences and on discussions I have had with IT leaders at many of our clients.

1. Scale Play: While I’ve not always been a huge fan of Nicolas Carr, I do like the core analogy in book The Big Switch: That computing today is like electricity in the middle 1800’s – every factory had it’s own coal burning power plant to supply electricity to make the factory run.  Today, in the early 21st century, every office has it’s own computing power plant (called a data center) to supply information to make the office run.  Carr posits, and I agree, that just as the electricity grid became a scale play utility, so will computing power.  I do not, however, agree with his belief that nearly all computing technology will go in this route.



2. Lower==Easier: The lower on the technology stack, the more applicable cloud computing is.  While there certainly are many exceptions, the closer to the infrastructure you get, the better it can serve as a common platform.  Telecom providers have been doing this for years with the network.  AOL gave many enterprises a “cloud based” instant messaging platform long before it was hip to call anything cloud (likely in 2006, but perhaps as early as 1996).



3. IAAS – good to go: Infrastructure as a Service (IaaS) is the most exciting part of the cloud. If you believe in the promise of cost savings promised by the cloud, you need a way to move some legacy applications there – without business case destroying remediation/redesign projects.  Putting only new applications in the cloud means a 15-25 year journey for most companies.  With IaaS, you must also be clear what  you are getting: Who’s patching, who’s administering, who’s accountable for what?



4. PaaS – good, scary: Platform as a Service (PaaS) is attractive and scary.  I like the added value run-time services it provides, the speed of development, and the scalability design patterns it brings. (Accenture has long been a proponent of investing in common run-time architecture even in the pre-cloud era).  The lock-in does, however, make me pause.  I don’t like that I can’t bring back in-house an application written for on some PaaS offerings.  In Accenture’s IT, we’ll likely do some smaller new development on PaaS, but we won’t go all in.



5. SaaS - not just packaged software: While Software as a Service (SaaS) on a technical level is nothing more than “packaged software” running on an infrastructure cloud, there is a bit more to it than that.   Yes, SaaS has many of the advantages of IaaS (although I’d argue few of them will truly make it as a scale IaaS – see point #9).  The commercial model of SaaS is dramatically different from traditional packaged software model.  Low, or no, upfront costs combined with some for of pay by the use/user/transaction is very attractive.  Given the diversity in the jobs and roles of Accenture’s nearly 200,000 employees, we lean toward some type of usage/transaction based model rather than user. Additionally, true SaaS offerings are often simply designed better – modern architectures, more modular and built with integration in mind.  This design approach can dramatically shorten the time to deployment (and therefore business value).  One last caveat around SaaS – make sure you can get your data back if/when your contract ends!



6. Rise of commodity: A big part of the cloud advantage that I don’t see talked about much is the ability of cloud providers to give us enterprise class performance, scalability and reliability from commodity hardware and open source software.  If you walk around Google’s data center, you don’t see a lot of high end Tier 1 storage or specialized servers.  Many Fortune 1000 companies are paying 10x the storage cost and 5x the computing cost for specialized hardware and software.  The trend we have seen over the last 20 years will continue – more and more computing will be done on commodity (read: “cheap”) hardware and software.



7. Public vs. Private: There is much talk in the industry about public (external) and private (internal) clouds.  While Accenture today is using both, I believe it is predominantly a transitory state.  While there will certainly be exceptions, I believe that the scale play of the major cloud providers will be too hard to overcome, even for Global 1000 companies.   Remember, the biggest proponents of the private clouds are the companies that don’t want to lose you as a customer. 



8. Lions and Tigers and Bears, Oh my: Security FUD around cloud is overblown.  Most people I talk to don’t get it.  Yes, there is risk in entrusting your “crown jewels” to a 3rd party.  But many companies already do this: they engage in 3rd party non-cloud hosting, they hire contractors who don’t even encrypt their laptops.  I’m not saying there is no risk in signing up for Cloud Computing, especially with some of the startups that may lack in experience in running enterprise class IT.  But you should talk to Google, Microsoft, Verizon or Amazon about their security practices and expertise.  In my experience they will stack up very well against your internal capabilities.  Where do you think the best minds in security are going? 



9. SaaS to Iaas?: If I ran a SaaS company, I would run my offering on someone else’s IaaS or PaaS.  Strategically, I believe my company’s value would be in the IP tied up in the software, not in running the hardware.  Let the scale plays handle that.  As enterprises embrace IaaS, I think you will see SaaS (and PaaS) companies moving their infrastructure towards the IaaS of one of the major scale players.



10. Accenture’s Cloud Experience: Accenture’s experience with the cloud providers has been quite positive.  On the SaaS front, our implementations have gone smoothly and arguably (we don’t run parallel projects to race) more quickly than had we gone a non-SaaS route.  The commercial models have been acceptable even though they continue to evolve.  Our IaaS and PaaS pilots have made us quite bullish on pushing more of Accenture internal IT to the cloud.  While it is all a moving target, our modeling activities with Accenture’s Cloud Computing Accelerator have shown that we could cut a very significant percentage of our hosting costs by going “all in” on cloud computing.   As a matter of fact, we have recently agreed to move our internal Exchange and Sharepoint to Microsoft BPOS.