Newsletter sign-up

View all newsletters

Sign up for our technology specific newsletters.

Enterprise Java

Email Address:

Survival of the fittest Jini services, Part 2

Use transactions to coordinate the reliable interaction of Jini services

By Frank Sommers, JavaWorld.com, 07/20/01

Failure, according to Merriam-Webster's Dictionary, is a state in which something is "unable to perform a normal function." Inasmuch as a network's normal function is to transmit information between two or more hosts, experience shows that most networks often cannot perform that function as expected. In other words, failure is as much a characteristic of the network as is its normal operation.

Read the whole "Survival of the Fittest Jini Services" series:

• Part 1. Ensure the quality of Web services in the age of calm computing
• Part 2. Use transactions to coordinate the reliable interaction of Jini services
• Part 3. Implement transactional Jini services

In many aspects of life, we have learned to live in the presence of failure. In a large city, many new shops spring up all the time, while others close their doors for good. Unless you are a shop owner, you are not likely to lose sleep over that fact. Instead, you are interested in being able to obtain the goods you are looking to buy, at a reasonable price and in close proximity.

Taking a similar approach to building Jini-based distributed systems might be helpful. We cannot make a large network, such as the Internet, more reliable. But we can make the computations we wish to perform over that network as reliable as possible. Your users -- whether people or other Jini services -- are primarily interested in the computations your service provides. Ensuring the reliability of those computations in the presence of network and component failures will likely lead to your service's longevity.

By reliability I mean a set of guarantees that hold, no matter what. In other words, as long as the computation produces a result, that result should keep with a set of guarantees. If the computation cannot ensure those guarantees, then it should abort and not return a result.

We are all familiar with this notion of reliability: When people wish to accomplish a goal together, they typically agree to a verbal or written contract, which thereafter binds each party to its terms. Thenceforth, the participants perform all their actions related to the common task in accord with those promises. And, should the parties fail to keep their promises, all actions under contract will produce "unreliable" and unpleasant results.

The equivalent of such "rules of the game" between components of software-based distributed systems is the transaction: components participating in a computation agree to a set of rules, and each component thereafter adheres to those rules during the computation.

In distributed systems, such as Jini networks, components typically reside on distant network hosts. This is significant, because it means that no component can, by itself, ascertain whether the other components adhere to the rules. A component can only implement the rules and then communicate to the others that it, indeed, keeps with those rules.

Distributed transactions, therefore, are made up of the rules (semantics) by which the services must abide, and a coordination mechanism between the services that ensures that the rules hold for the whole computation. If even one service indicates that it cannot guarantee its promise, that mechanism will abort the transaction.

 

Improve efficiency with Visual Studio(r) 2010 Beta 2. - Microsoft

About Us | Advertise | Contact Us | Terms of Service/Privacy

Copyright, 2006-2008 Network World, Inc. All rights reserved.