Ideally there should be differnt tiers of data centers for differnt tiers of applications and Cloud (public and/or private) with different SLAs gives us these tiers
Your development and testing/qa does not belong to production grade data centers. And there are many types of workloads which can viably sit in cloud with great efficiencies.
I originally wrote this in response to a blog post by Rob Hirschfeld on his blog post titled “Why cloud compute will be free”. My take on this is that in general higher volume of data at a cloud provider less likely that client will move to another provider, but in some cases cloud usage can be compute intense (compute can cost more than data storage). For example for testing a microprocessor simulation you can spin up 200 nodes for 4 hours and bring these down, input data more or less will remain constant and outputs are not heavy either. In these types of scenarios cloud is a great fit, as it cuts down time of testing a microprocessor simulation from 4 days to 4 hours (for example sake).
I don’t think stickiness of clients will be due to their data volume, I think stickiness will be due a mix of several factors (of which data can be one) and these will differ based on verticals and use cases. Few things which are important to enterprises are: quality of service, consultative approach to service delivery, integration with existing enterprise infrastructure, vendor’s ability to innovate and vendor’s understanding of client’s requirements/challenges. Having said that, I strongly agree that cloud providers which throw out some free services (testing the waters) will do much better in attracting and retaining clients. After all, humans will pick which cloud to go to and if they are well versed in one vendor’s technology/process/offerings they will prefer that vendor over others. Technique of throwing a free evaluation licensing out there (for six months or so) was mastered by Microsoft and that was secret behind dominance of languages like VB in 1990s. Later this technique was copied by other vendors like Oracle and they are benefiting from it (SAP ignore it and see where they are now).
Another twist I would like to add here is; a vendor’s treatment of developer community which is closely related to my point on free service for limited time, vendors who will make it easy for developers to write code to their cloud services (from integration to new applications through programmable infrastructure and platform as a service) will win in the long term. Keep in mind; what makes data valuable is the processing of data which is done through applications which are written by developers at large. More love a cloud provider shows to developers more love they will get back…
When we talk about people side of things in cloud context, we can divide it into three categories:
First, people who consume these applications as end users. In my view, impact on end users will be minimal as long as application performance does not deteriorate and end users are assured of their data security and privacy. To address needs of this set of people (where enterprise application is consumer driven), during assessment phase, deeper emphasis on data classification and performance analysis should be placed.
Secondly, it’s set of people who build, host, operate, manage and service enterprise applications. Cloud is most disruptive for this set of people. In this area, age and architecture of application will play a major role. Newer/younger applications will be easier to handle than old/legacy (it’s a relative term) applications. One way to approach this area is to generate heat maps for enterprise applications (application tier by financial impact on business). Other factors which should be taken into account are relationship with vendors for licensing, servicing & maintenance contracts, you want to make sure that these applications will still be supported be vendors who wrote them and can be serviced and maintained by either the exiting vendors or an alternate vendor (may be cloud service provider). Day to day upkeep of these applications is done by people (through technologies) that includes, monitoring & backup (amongst other things). Another are under this set of people is application dependencies, in bigger enterprises different set of people procure and manage different applications but these applications have to work with each other. People alignment is a big task in big enterprises, difference of opinion or people ego in one group can impact cloud adaption for other applications. When it comes to this set of people (in second category), a careful consideration should be given to retraining existing workforce to work in new paradigm. From my observations in the industry, I can say this with certain level of confidence that (besides best efforts of all businesses on policies and procedures) there is a lot of tribal knowledge buried in these enterprises which makes them move. Abruptly changing roles or letting people go can be very disruptive. Retention while retaining of existing people for businesses will be a key to smooth transition to cloud.
Thirdly, moving enterprise applications to cloud will impact business folks especially on two fronts; one is cost structures which will change. In short run there will be costs associated with either moving the applications to cloud as is or for retrofitting/refactoring, two is that move to cloud will have impact on company’s balance sheet. In an ‘old data center’ model servers/switches/routers were considered assets and in cloud world (especially if cloud is provided by a service provider or a separate entity) business will not be able to show these assets on their books. In this area some of industry thought leaders are pushing for legislatures where company’s IP (intellectual property which pertains to expensive software applications) will carry certain money value on balance sheets, which is not the case currently.
When it comes to moving to cloud, technology is about 30% of the work, rest is people and processes. If people and processes side is not handled properly, move to cloud can be a bitter and expensive experience. I will highly recommend enterprises to engage external (third party) assessors for moving applications to cloud. Drawbacks of putting internal people on this task can cause problems like; biased opinions, entity misrepresentation and too much emphasis in certain areas and too little emphasis in others (missing forest for trees).
I am eager to hear everyone’s input on this one!
Cloud lowers barrier to entry for new software startups. What do you think?