Robert Pogson

Matt Asay quite properly predicts that “cloud computing” will be mainstream in a year or so. He quotes figures showing billions in annual revenue already. This has implications for all of IT. If billions are spent on cloud computing are those dollars in addition to current expenditures or are they a replacement for some current expenditures? Probably a bit of both.

There are a variety of cloud services that are imminent:

• storage, backup, file serving and the like that could replace some expenditures on local data centres
• load leveling, providing computing power or service where and when it is needed, e.g. buffering variations with time of day
• wholesale farming out of web applications like e-mail
• new kinds of web applications
• virtual desktops or desktop applications
• remote administration of whatever

Many of these are strictly practical solutions to re-inventing the wheel and over-staffing of server rooms. By that I do not mean people are playing cards in the server rooms but that you cannot have half a system administrator so you pay for a whole one. Further, if you give a system administrator the right tools there is almost no limit to how many system he/she can administer, so the cloud could be filled with sysadmins with fantastic tools unavailable to the in-house guy. It just does not make sense for a small to medium sized business to employ someone to administer an e-mail system when Google or Yahoo can do a better job at a lower price. Shifts from in-house to cloud make sense for almost any service that can be separated in some way from all the other services needed in a business. It is like software. Modularization is extremely helpful if not essential. Farm out a module and divide and conquer your IT problems/tasks.

Some of these overlap into desktop operating systems and applications. Almost no one objects to these things running on a local server. Why not on the cloud? You need some assurances about connectivity, security, backup and uptime. These can be managed. It may not make sense to depend on services a satellite-jump away but on a fibre line with redundant connections in a city, this could be a good plan for business. Already there are services providing basic office applications and thin client desktops on the web. More will come.

To the extent that M$ and other huge software providers cannot capture the flow of services to the cloud they will lose share in the cloud. The speed with which changeover can be made to cloud services is faster than migrating to a new OS because the user mostly has to be familiar with the browser and that is done. Any cloud supplier who has done his homework and provides good infrastructure in the cloud will get the business. M$ has built huge datacentres to capture some of this business but it is not like the desktop. They have no monopoly and no way of securing one in the cloud. People choosing a cloud supplier have lock-in on their due-diligence checklist. It is not 1990 all over again. M$ has an obscene cash-flow that can be used to buy hardware but have they a workable plan to supply software?

New software for the cloud involves innovation, not M$’s strong suit. They can try to mimic their desktop OS and applications but they have alienated large parts of the market with anti-competitive activities so the loyalty of customers is in question. Cloud computing is about lowering costs. M$’s business plan is about enriching M$, not providing service at lowest cost. They will not be able to sustain anything close to a monopoly in the cloud.

Netbooks on the low end and the cloud on the high end leaves M$ struggling to stay relevant in the middle somewhere. It will take a few years to know the result but I would bet that M$’s share of the pie will shrink drastically within five years. The dollars going into the cloud will come in a great measure from M$’s former share. They were charging $1000 plus $40 per client for in-house use of their OS. The cloud, running on FLOSS, can provide the same or better service for a lot less. They were charging $100 more of less for the privilege of using their bug-ridden OS on desktops. A GNU/Linux terminal server costs about $25 per client in quantity one. In cloud-volumes, the cost will be much less. You can buy hosted virtual desktops now. The lack of retail shelf space or OEM suppliers need not hold you back.