Multithreading and code concurrency were once the preserve of elite programmers, but the combination of multicore processing power, complex requirements, and the readily available javax.util.concurrent package has changed that. Today, enterprise application developers are expected to be knowledgeable about the various synchronization mechanisms and constructs available in the Java language. The level of expectation is even higher where the problems being solved require non-textbook and highly innovative concurrency constructs. It isn`t enough in those situations to understand the Java language`s concurrency mechanisms and those included in the standard SDK; you also must be able to use these tools as building blocks for custom-made concurrency controls.
In this article, we`ll explore a concurrency pattern that isn`t widely discussed, generally called thread gates. Like its real-world counterpart, a gate instance opens or closes, thus allowing or preventing thread progress. It does so based on the truth value of some predicate.
I`ll start with an overview of thread gates based on the traffic-flow model, then explain how to set up your development environment for the example application, a multithreaded prime-number generator. The remainder of the article will be hands-on, as you learn about the thread gate pattern by implementing it.
Thread gates: An overview
A good metaphor for thread gates is the traffic light system that operates on many public roads. When the signal is red, cars have to wait at the stop line until the signal changes. On the other hand, cars are free to run past the signal when it is green. There is no limit to the number of times that the signal can go from green to red, given a statistically significant observation timeframe. Traffic lights are designed to enable crossflow of traffic, and are redundant where crossflow does not exist. In a programmer`s terms, you can imagine the light as a control that lets bidirectional traffic share a small section of road that might otherwise cause the paths of traffic to intersect in an unsafe manner.
In the same way, thread gates are generally best used for scenarios where one set of threads has to be prevented from proceeding beyond a determined point, whilst another set of threads is active. To put it another way, the competing sets of threads are dependent on the value of some truth predicate, where each distinct value of the predicate strictly triggers one (and only one) set of threads, forcing the others to remain suspended. Note that the emphasis here is on sets or groups of threads rather than individual threads. In essence, the focus is on scenarios where multiple threads share access to an underlying resource, and the threads can be partitioned into sets depending on the actions that they perform in relation to that resource.
A good example of this is in producer-consumer flows, where some threads are responsible for producing data that is consumed by another group of threads; the shared resource is most likely the handoff mechanism (data bus) used by the disparate thread groups; and the truth predicate that determines thread progress is the quantity of data in the handoff mechanism. An in-memory request queue can sometimes fit such a pattern if the data is enqueued as part of a process that is analogous to a production process, and the data is dequeued or consumed by a separate process.
The producer-consumer analogy is a good one for formulating a problem that explores the concept of thread gates. An example problem should be easily recognizable to the vast majority of programmers. In this case, the problem should also be one that lends itself easily to task splitting and parallelization, as the emphasis is on creating a multithreaded solution to it. This article`s sample application will fulfill all of these criteria and more.
The multithreaded prime-number generator
This article`s example application will tackle the age-old problem of finding all the probable prime numbers within a given bound (between one and a million, say). Specifically, the task is to implement a software component that exposes a method for retrieving all the probable prime numbers in a given range (with an upper and lower bound given). Assume that the component`s client mandates that the method return a thread-safe handle for accessing the results as soon as it is invoked, and then performs the actual task of finding the prime numbers in the background or asynchronously. Assume also that the handle must provide a blocking method that allows primes to be accessed and returned as soon as possible, such that the client does not have to wait for the search to complete before accessing results. To simplify the task, no restrictions are placed on the order in which the handle returns the results.
Source:
http://www.javaworld.com/javaworld/jw-03-2009/jw-03-java-concurrency-with-thread
In this article, we`ll explore a concurrency pattern that isn`t widely discussed, generally called thread gates. Like its real-world counterpart, a gate instance opens or closes, thus allowing or preventing thread progress. It does so based on the truth value of some predicate.
I`ll start with an overview of thread gates based on the traffic-flow model, then explain how to set up your development environment for the example application, a multithreaded prime-number generator. The remainder of the article will be hands-on, as you learn about the thread gate pattern by implementing it.
Thread gates: An overview
A good metaphor for thread gates is the traffic light system that operates on many public roads. When the signal is red, cars have to wait at the stop line until the signal changes. On the other hand, cars are free to run past the signal when it is green. There is no limit to the number of times that the signal can go from green to red, given a statistically significant observation timeframe. Traffic lights are designed to enable crossflow of traffic, and are redundant where crossflow does not exist. In a programmer`s terms, you can imagine the light as a control that lets bidirectional traffic share a small section of road that might otherwise cause the paths of traffic to intersect in an unsafe manner.
In the same way, thread gates are generally best used for scenarios where one set of threads has to be prevented from proceeding beyond a determined point, whilst another set of threads is active. To put it another way, the competing sets of threads are dependent on the value of some truth predicate, where each distinct value of the predicate strictly triggers one (and only one) set of threads, forcing the others to remain suspended. Note that the emphasis here is on sets or groups of threads rather than individual threads. In essence, the focus is on scenarios where multiple threads share access to an underlying resource, and the threads can be partitioned into sets depending on the actions that they perform in relation to that resource.
A good example of this is in producer-consumer flows, where some threads are responsible for producing data that is consumed by another group of threads; the shared resource is most likely the handoff mechanism (data bus) used by the disparate thread groups; and the truth predicate that determines thread progress is the quantity of data in the handoff mechanism. An in-memory request queue can sometimes fit such a pattern if the data is enqueued as part of a process that is analogous to a production process, and the data is dequeued or consumed by a separate process.
The producer-consumer analogy is a good one for formulating a problem that explores the concept of thread gates. An example problem should be easily recognizable to the vast majority of programmers. In this case, the problem should also be one that lends itself easily to task splitting and parallelization, as the emphasis is on creating a multithreaded solution to it. This article`s sample application will fulfill all of these criteria and more.
The multithreaded prime-number generator
This article`s example application will tackle the age-old problem of finding all the probable prime numbers within a given bound (between one and a million, say). Specifically, the task is to implement a software component that exposes a method for retrieving all the probable prime numbers in a given range (with an upper and lower bound given). Assume that the component`s client mandates that the method return a thread-safe handle for accessing the results as soon as it is invoked, and then performs the actual task of finding the prime numbers in the background or asynchronously. Assume also that the handle must provide a blocking method that allows primes to be accessed and returned as soon as possible, such that the client does not have to wait for the search to complete before accessing results. To simplify the task, no restrictions are placed on the order in which the handle returns the results.
Source:
http://www.javaworld.com/javaworld/jw-03-2009/jw-03-java-concurrency-with-thread
