No Recent MPI Books: Is It Dying?

Question

I never used the messaging interface (MPI), but I heard how his name appeared recently, with Windows HPC Server. I quickly looked at the Amazon to see if there were any books on it, but they all date back 7 or more years. Is MPI still the current technology choice for new applications, or is it significantly superior to other distributed programming alternatives (like DataSynapse GridServer)?

How is this not an implementation, but rather a standard, what is the probability (assuming that it is not dead) that its study will lead to a better design of distributed programming systems? Is there anything else I should look at instead?

Answer 1

For what MPI is good for him, remains a good choice. It is simply possible that there are no recent books on this subject, because the existing ones are good enough, and most of us using MPI don't need anything else.

I would not characterize MPI as a distributed programming standard, more standard for parallel programming on computers with distributed memory, which now covers most of the largest computers in the world.

If I were betting on a replacement, I would look at Chapel , X10 or, most likely, Fortran 2008.

What you should look at depends on your requirements, but if they include a high-performance number-crunch for scientific and engineering codes, Fortran or C / C ++ with MPI should be in your views. I've never heard of DataSynapse GridServer, the fast Google suggests that it targets a completely different class of computational problems.

EDIT: I just checked Amazon for "MPI" books. While the books of Gropp et al. A bit outdated, there are still many other published books that cover MPI. This is, in particular, a reflection of the use of MPI. This is not very interesting for computer scientists, so there are few MPI for MPI books, but it is interesting for many computer scientists, so there is a steady stream of “physics with MPI” and “engineering with MPI”. If they are outside the sphere of interest, perhaps MPI too.

Answer 2

Update

2016. MPI is still the king for programming distributed memory on low latency networks of reliable computing nodes. I think the question is posed correctly in that MPI is probably not the protocol layer where fault tolerance should take place. In 2006, we had MPI over SunGridEngine. Recently becoming a popular MPI on Mesos .

---

The MPI standard is under active development: http://meetings.mpi-forum.org/MPI_3.0_main_page.php

The main problem is that now we have several machines with more than 10,000 processors, and MPI itself is complicated scaling. A lot of open research problems. http://www.springerlink.com/content/q11r042317g88230/

Answer 3

Why do you need a book? The API is well documented.

In distributed systems, you have no option other than MPI.

Some Fortran compilers, such as Cray and G95, support coarrays. Then you have a UPC, but I have not seen anyone use it.

Answer 4

Well, probably because it’s not enough for this (or the user base is still too small or too smart) to describe the API and a few examples to support a single book. However, many parallel programming books cover it as one of several parallel methods. One from the latest (February 2010): “Parallel Programming: for Multi-Core and Cluster Systems” by Thomas Rauber, Gudula Runger. I have not read it, I mention this because it is the last and the real experts in this field (both => MPI not are dead). As for the best book to help you think about how to use MPI, I can only refer to people’s reviews on Amazon, but look for a “parallel” in the title.