Parallel Computing Using Linux
At the Supercomputing '92 conference, a committee, later known as the MPI forum, was formed to develop a message-passing standard. Prior to the development of MPI, each parallel computer vendor developed a custom communication protocol. The variety of communication protocols made porting code between parallel machines tedious. In 1994, the first version of MPI was released. The advantages of using MPI are:
It is a standard and should make porting among machines easy.
Vendors are responsible for developing efficient implementations of the MPI library.
Multiple implementations of MPI are freely available.
One of the problems with using PVM is that vendors such as Digital, Cray and Silicon Graphics developed optimized versions of PVM which were not freely available. The custom versions of PVM often did not include all the library routines of the PVM system distributed by Oak Ridge, and they sometimes included non-standard routines for better performance. Despite the problems mentioned, PVM is easy to use and developing a new parallel application takes little effort. Converting a PVM application to one using MPI is not a difficult task.
How do you go about parallelizing a program? The first step is to determine which section of the code is consuming the most time. This can be done using a profile program such as prof. Focus on parallelizing a section of code or a group of functions instead of an entire program. Sometimes, this may mean also parallelizing the startup and termination code of a program to distribute and collect data. The idea is to limit the scope of parallelism to a manageable task and add parallelism incrementally.
Run with two, four and eight processors to determine scalability. You can expect the improvement in performance to diminish as you use more processors. The measure of parallel performance often used is speed up. It is the ratio of the time taken to solve a problem using the best sequential algorithm versus the time to solve a problem using a parallel algorithm. If you use four processors and obtain a speedup of more than four, the reason is often due to better cache performance and not a superior parallel algorithm. When the problem size is small, you can expect a higher cache hit rate and a correspondingly lower execution time. Superlinear speedup is looked upon skeptically since it implies more than 100% efficiency.
While you may obtain good speedup, the results from the parallel program should be similar to the results from the sequential algorithm. You can expect slight differences in precision when heterogenous hosts are used. The degree of difference in results will depend on the number of processors used and the type of processors.
The most efficient algorithm is the one with the least communication overhead. Most of the communication overhead occurs in sending and receiving data between the master and slaves and cannot be avoided. Different algorithms to distribute and collect data can be used. An efficient topology which minimizes the number of communication hops between processors can be adopted. When the number of hops between processors is large, the communication overhead will increase. If a large communication overhead is unavoidable, then computation and communication can be overlapped. This may be possible depending on the problem.
If you are looking for a modest improvement in performance of your application, it is possible to use a cluster of PCs or an SMP (symmetric multiprocessing) machine. But, most applications do not see a dramatic improvement in performance after parallelization. This is true for a cluster of PCs, since the bandwidth and latency are still relatively high compared to the clock speed of processors. For an existing network, there is no additional hardware required and the software to run a parallel application is free. The only effort required is to modify code. Some examples of where parallelism is applicable are sorting a long list, matrix multiplication and pattern recognition.
Manu Konchady (email@example.com) is a graduate student at George Mason University and works at Goddard Space Flight Center. He enjoys flying kites and playing with his 2-year-old daughter.
Practical Task Scheduling Deployment
July 20, 2016 12:00 pm CDT
One of the best things about the UNIX environment (aside from being stable and efficient) is the vast array of software tools available to help you do your job. Traditionally, a UNIX tool does only one thing, but does that one thing very well. For example, grep is very easy to use and can search vast amounts of data quickly. The find tool can find a particular file or files based on all kinds of criteria. It's pretty easy to string these tools together to build even more powerful tools, such as a tool that finds all of the .log files in the /home directory and searches each one for a particular entry. This erector-set mentality allows UNIX system administrators to seem to always have the right tool for the job.
Cron traditionally has been considered another such a tool for job scheduling, but is it enough? This webinar considers that very question. The first part builds on a previous Geek Guide, Beyond Cron, and briefly describes how to know when it might be time to consider upgrading your job scheduling infrastructure. The second part presents an actual planning and implementation framework.
Join Linux Journal's Mike Diehl and Pat Cameron of Help Systems.
Free to Linux Journal readers.Register Now!
- SUSE LLC's SUSE Manager
- My +1 Sword of Productivity
- Murat Yener and Onur Dundar's Expert Android Studio (Wrox)
- Non-Linux FOSS: Caffeine!
- Managing Linux Using Puppet
- Doing for User Space What We Did for Kernel Space
- SuperTuxKart 0.9.2 Released
- Tech Tip: Really Simple HTTP Server with Python
- Parsing an RSS News Feed with a Bash Script
- Rogue Wave Software's Zend Server
With all the industry talk about the benefits of Linux on Power and all the performance advantages offered by its open architecture, you may be considering a move in that direction. If you are thinking about analytics, big data and cloud computing, you would be right to evaluate Power. The idea of using commodity x86 hardware and replacing it every three years is an outdated cost model. It doesn’t consider the total cost of ownership, and it doesn’t consider the advantage of real processing power, high-availability and multithreading like a demon.
This ebook takes a look at some of the practical applications of the Linux on Power platform and ways you might bring all the performance power of this open architecture to bear for your organization. There are no smoke and mirrors here—just hard, cold, empirical evidence provided by independent sources. I also consider some innovative ways Linux on Power will be used in the future.Get the Guide