Tuesday, August 11, 2009

The coming of virtual appliances

Everyone knows that web 2.0 has revolutionized people's private lives, with online social and entertainment applications like Facebook and YouTube. However, web 2.0 so far only has limited success for business applications in offices where the real money is at. While people are willing to entrust personal data to web 2.0, businesses can't do the same about their critical business data. Even well-known brand name like Google is fighting a hard fight selling the Apps for Domain service.

The business concerns are manifold. Privacy: do I get to control who has access to my data? Safety: once data is stored, will it persist and out-survive any forms of failure? Liability: is there someone I can sue and recover some damage when worse comes to worst? Being a business model still in its infancy, many web 2.0 companies have terms of service that are against all these concerns. Only after a while do people realize that these are non-issues with certain brand name companies simply because of the sheer number of user base and the lack of negative publicity.
Digression. In fact, some negative publicity became confidence boosters. When lightning stroke one of Amazon's data centers for EC2, it brought down a rack of machines. All their customers had to was to restart the instances on other machines that are still working. They lost only the temporary instance data that have not been committed to permanent storage, the same kind of loss if you unplug the power to your servers. It took only a few minutes to resume normal operation.
It is different if you are a small web 2.0 startup that wants to focus on business customers. How do you build a customer base when nobody wants to trust their data with you, and how do you build trust without a customer base? Furthermore, the remoteness of the web means it's more difficult to hold you physically accountable, so that is one more reason not to trust you. This is an obvious reason why web 2.0 business-oriented applications failed to flourish so far.

Also, as a small web 2.0 startup, you want to focus on developing the application, not on the plumbing. You can go back to pre-Internet model, selling software on a disc and have your customers install it on their own machines. However, some web 2.0 applications are built on top of many dependencies (web server, database, programming language runtime) as well as being intricately integrated with the operating system (disk access policy, user privilege, network firewall rules, performance tuning). It might not be your customer's best interest to run IT in-house either.

Cloud computing is a solution to this problem. Here comes a brand name utility provider like Amazon with a service like EC2 that you will more likely trust. And a third-party software company with a web 2.0 application you need. They give you a stateless virtual machine image (such as AMI) and charge you a license fee for using their software. Amazon provides the hardware, storage, and network infrastructure, and bills you for the operational costs. All the costs are fully transparent: license fee for the development and support of the application, and Amazon for only what you use.

When it comes to a customer who has built an infrastructure around VMware technology, the third-party simply rereleases their web 2.0 application as a VMware virtual appliance. It is easy to adapt to another virtualization technology because the only difference is the tool used for packaging a virtual appliance; the software stack---the operating system and all the dependencies---stay very much the same.

Once the web 2.0 company grows large enough, they could consider selling hardware appliances; or if they're willing, provide a hosted solution for their apps when someone asks for it.

To summarize, cloud computing with virtual appliances is going to be a great platform for business oriented web 2.0 technology, particularly lowering barrier to enter market due to lack of trust of critical data. And now, web 2.0 can finally enter the business sector.

No comments: