Efforts by Web heavyweights such as Amazon and Google to entice companies into tapping into the power of their datacentres are being slowed by a number of factors, according to Interop panelists.
Analyst Alistair Croll of BitCurrent said there are specific applications for which grid/cloud computing is perfect. For example, The New York Times recently rented Amazon's grid to create searchable PDFs of newspaper articles going back decades. The Times estimated that the project would have taken 14 years if the Times had used its own servers. Amazon did the entire project in one day, for US$240.
But those examples are few and far between, as most companies are still in the 'kicking the tyres' stage when it comes to grid computing. Reuven Cohen, founder and CTO of Enomaly, said his customers are primarily using grid computing for research and development projects, rather than production applications.
Kirill Sheynkman, head of start-up Elastra, said the early adopters of grid computing are Web 2.0 start-ups who want to get up and running quickly and without a lot of capital expenses, independent software vendors that want to offer their applications in a software-as-a-service model, and enterprises who have selected specific applications for the cloud, such as salesforce automation or human resources.
"Equipment inside the corporate datacentre isn't going away anytime soon," added Sheynkman. Companies remain reluctant, for a variety of reasons, to trust the cloud for their mission-critical applications. Here are some of those reasons:
- Data privacy. Many countries have specific laws that say data on citizens of that country must be kept inside that country. That's a problem in the cloud computing model, where the data could reside anywhere and the customer might not have any idea where, in a geographical sense, the data is.
- Security. Companies are understandably concerned about the security implications of corporate data being housed in the cloud.
- Licensing. The typical corporate software licensing model doesn't always translate well into the world of cloud computing, where one application might be running on untold numbers of servers.
- Applications. In order for cloud computing to work, applications need to be written so that they can be broken up and the work divided among multiple servers. Not all applications are written that way, and companies are loathe to rewrite their existing applications.
- Interoperability. For example, Amazon has its EC2 Web service, Google has its cloud computing service for messaging and collaboration, but the two don't interoperate.
- Compliance. What happens when the auditors want to certify that the company is complying with various regulations, and the application in question is running in the cloud? It's a problem that has yet to be addressed.
- SLAs. It's one thing to entrust a third party to run your applications, but what happens when performance lags. The vendors offering these services need to offer service-level agreements.
- Network monitoring. Another question that remains unanswered is how does a company instrument its network and its applications in a cloud scenario. What types of network/application monitoring tools are required.
While many of these questions don't have answers yet, the panelists did agree that there is a great deal of interest in grid computing. Conventional wisdom would say that small-to-midsize businesses (SMB) would be most interested in being able to offload applications, but, in fact, it's the larger enterprises that are showing the most interest.
As Google's Rajen Sheth pointed out, when Google started its messaging and collaboration services, it thought SMBs would be the major customers. "Lots of large enterprises are showing interest," he said, "but it will take a while."