Cloud computing raises a lot of issues, from information crossing borders to homogenous IT infrastructure, and our bloggers weighed in on the issues at stake, including the fact that cloud computing, in one form or another, may have been around for a while, and the many nagging questions faced by IT managers around this out-of-the-box method.
This was the second topic of ComputerWorld Canada’s Blogging Idol contest, and we have published extracts from IT bloggers’ entries on our site.
The complete posts can be viewed at the Blogging Idol Web site.
Let’s use my Services-as-a-Service definition—this includes just about any computing function provided on a pay-as-you-go basis by a third-party supplier. Thus, we could include basic PC applications on-demand, software access, server hosting, storage hosting, common applications, collaboration services, and so on.
How comfortable are companies with this level of outsourcing? How much of anyone’s IT infrastructure and how many of their applications are sufficiently generic? Are IT executives convinced that services provided by one (or more) suppliers in cloud form will meet required service levels?
How about storage? If you stored data in the cloud, would you want a back-up copy on a different cloud (or even in-house) to be sure it won’t get lost? What about the old trans-border data flow issues—some companies or governments would not want the cloud to be located outside the country.
How about security? Is the cloud any less susceptible to security issues than in-house systems? Will the cloud meet military or banking security requirements?
How about development/test systems? Would we be able to use the cloud for development systems, test systems and acceptance testing systems?
How about …… you name it!
Big deal. I’ve been using hosted internet web services for years.
It amounts to essentially this: I host my applications on someone else’s equipment; that equipment is shared (i.e. not a dedicated server); and I don’t need to worry about maintenance, renewal or provisioning.
It is all part of the hosting fee.
The challenge is a little loss of flexibility (can’t just put anything on that server), and not being able to go and reboot it when I want.
Giving up that level of control, or being dependent on a third party and your WAN connection, can be intimidating. It’s not for everyone, but if you don’t have the square footage, are growing dynamically or just don’t want to have to deal with the complexity of building your own data center, I’d say yes, it is the right way to go. For shops that are already completely virtualized… well, you are already 99 per cent there, just get a decent link to two or more Tier 4 carrier-based data centres, turn on VMWare HA, and you are done.
Overall, I agree with Don Sheppard’s post calling cloud computing not much more than marketecture. It’s basically a marketing department’s way to present an old idea.
Perhaps the single greatest advantage of cloud computing is that it easily handles peak load situations without the need for additional hardware infrastructure that will spend most of its time underutilized. Resources can be virtualized and presented to the customer as virtual servers that the customers themselves manage. Physically, the resources might span multiple computers or even multiple data centers.
I believe that cloud computing is here to stay, but also that it will have slow growth at first. The slow growth is partly due to a large installed user base of data centre servers that would be difficult to port to the cloud, and partly due to general discomfort with hosting the company’s data on someone else’s hardware. In fact, Canadian law requires that, for certain types of data (such as health records), the data must be hosted on machines that are physically located within the national boundaries. The reason given by the government for this is simple: the U.S. Patriot Act. Placing physical limitations on where data can be located defeats the purpose of a computing cloud.
As usual, the big players in the industry have all been pretty much stuck for a while, so, for now, they are all watching to see how Google is going to push the edges again. It will be particularly interesting to see if Google tackles the toughest part of the problem: integrating the cloud with mobile devices (you thought I was going to say the online/offline issue, didn’t you?) This is something they are positioning themselves to take on with Android, their (relatively) new mobile platform.
It will take a few years, but eventually cloud computing will reach critical mass and really take off in terms of popular use. In the process, it will make a lot of Internet empires, and a lot of empires bigger; the developers and companies that are able to build applications that seamlessly integrate the cloud with desktop computing and mobile devices will make the first fortunes. The ones that bring ease-of-use to the end consumer are the ones that will really turn clouds into rainmakers.
One of the questions that has to be asked when talking about cloud computing is whether the comparison between computing and electricity really works, or at least how far it can be pushed as an analogy.
In general, ubiquitous electricity became possible only through standards—what would we do if people wanted to buy electricity by the volt? You buy 10 volts, I buy 25 and someone else wants 110—well, we do have 110 and 220, but we don’t have a hundred different choices. If it weren’t for standards, would we have a single type of plug and receptacle? Would fuses be easy to buy? How many different types of light bulbs would we need to have in the stores?
Will “computing” ever reach the same state (where) I can buy the computing equivalent of 110 and 220 volts?
To answer that question, we can refer back to the OSI 7 layer model: at the physical to network levels, we do have a reasonably small set of choices—Wintel platforms, Ethernet networks, Internet links, etc.
For layers 4-7, it might even be possible to claim there are a reasonably small number of standards for highly-distributed, multi-organization computer processing—SOA comes to mind, as do things like Google searching, Web shopping and payment services, Facebook and many others. But, even here, this is more like having many kinds of electricity, or many types of water supply.
And it even gets more complicated when you talk about business applications and databases. Most companies would be reluctant to homogenize their customer-facing systems for competitive reasons. And potential suppliers really need to distinguish themselves in the marketplace.
So can we ever have word processing be like electricity? I doubt it! So, what does this mean? First of all, it says we should be careful about pushing analogies to far. Second, it suggests we need to worry about how different cloud suppliers would compete. And, finally, we probably need to worry about what corners are being cut (read: security flaws) in order to minimize costs and maximize speed to market. What do you think? Are you using network-based data storage? Are you using Google’s net-based services? Are you planning to? And, if so, will you think of it as being just like electricity (always