Hello Rogrigo, I'll take a shot at a response to this, bearing in mind there are quite a few salient details you haven't yet provided in this post.
--As a principle, I'd recommend keeping your solutions architecture as simple as possible, and use either PostgreSQL or SQLite, but not both, barring really good reasons to the contrary. --Integrating MPI (MPICH2 etc) with any relational database begs for very careful partitioning of the problem and the workflows used to address the problem; yes, storing computational results in a either SQLite or PostgreSQL makes perfect sense, but I'd look for clear process-boundaries and checkpoint/milestone events (say, after a salvo of concurrent tasks have been synchronized) to perform your updates to any database, and assure that any application I/O buffers are completely flushed when integrating any relational database within any parallel or concurrent application. PostgreSQL is similar to many other sophisticated RDBMS in that it maintains its own internal, multi-level caches. --Bear in mind, I'm a huge fan of SQLite for many applications, but its a question of balancing its relative strengths and weaknesses over PostgreSQL, and of course, assessing which environment you know better as well (or which you could learn the necessary subtlies, if you don't really know either). If PostgreSQL serves some roles that you find SQLite cannot easily serve, I'd just go with an all PostgreSQL solution. I suspect you'll have plenty of other technical and scientific areas of your work to ascertain the correctness of (and possibly debug), than to have to also take on the added complexity of understanding the interaction effects of two different SQL engines, if a problem emerges. --I would say it is "possible" to "employ SQLite" in an MPI application, but again, I think it depends a lot on how you are structuring your particular problem. If the runtime platform is "dual-core", I presume you're at least planning to deploy on (2) cores concurrently, but there are different mechanisms in MPI addressing for addressing multi-core architectures (on the same CPU and memory bus) than those addressing parallelism via interconnects between many separate CPU/memory buses (or blades, racks etc). Absent more information on your execution architecture, I'm guessing you're looking at some sort of multi-threaded (or multi heavy-weight process) design on a master bus. If so, you'll definitely want to factor in some very deterministic thread synchronization relative to when you perform writes to any database, whether it is SQLite or PostgreSQL. Generally speaking, concurrent reads on a backing store (SQL database) are often not as critical to schedule, but still bear careful consideration if you use these to trigger the start of a subsequent parallel process gang. OK, I'll stop there, but just wanted to address the question in overview. Some of the others on this list may be able to address this from a more SQLite centric perspective. good luck. Joe On Tue, Jun 2, 2009 at 10:33 AM, Rodrigo Faccioli < faccioli.postgre...@gmail.com> wrote: > Hello, > > I'm very new user about SQLite. I'm working with Structural Bioinformatics > and everybody knows that this area requires a lot of computational > resources. So, I'm developing an Evolutionary Algorithms (EA) for some > protein analysis (I don't know what will do exactly, because I've just > started my PhD). > > However, I know that I'll work with C++ and mpi. I've worked with > PostgreSQL > and Python with SQLAlmy and I've enjoyed. I've known about sqlite in > SQLAlmy > website. I'm enjoying the sqlite although I have a question about sqlite > and > mpi application. > > Is it possible to employ sqlite in mpi application? Example: I have a > computer which is a dual-core and my program works with mpi. Its goal is to > calculate and store it on text file. So, I want to change from text file to > SQLite. > > Another question refers to PostgreSQL and SQLite. I've read that SQLite > doesn't work with client-server applications. So, if my first question was > correctly, I have the idea using SQLite and PostgreSQL. The goal of > PostgreSQL is to attend my web-service applications and SQLite goal is to > store the results of calculation. It's clean that PostgreSQL will store the > results were created by my program which works with SQLite. Therefore, I > would know good way to work PostgreSQL and SQLite. > > I thanks for any help. > > Regards, > > Rodrigo Antonio Faccioli > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > ----------------------------------------------------------------------- Joseph Glassy Lead Software Engineer (contractor) NASA Measures (Freeze/Thaw),Rm CFC 424 CFC, Univ. Montana, Missoula, MT 59812 and: Research Analyst/Programmer University of Montana NSF EPSCoR Program Missoula, MT 59812 _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
