Comments for Sword Systems

Comment on Auto-create User Domain Object with Spring Security by An Army of Solipsists » Blog Archive » This Week in Grails (2011-50)

An Army of Solipsists » Blog Archive » This Week in Grails (2011-50) — Tue, 20 Dec 2011 08:31:03 +0000

[...] Auto-create User Domain Object with Spring Security [...]

[WORDPRESS HASHCASH] The comment’s server IP (74.208.16.10) doesn’t match the comment’s URL host IP (74.208.33.17) and so is spam.

Comment on Auto-create User Domain Object with Spring Security by » Blog Archive

» Blog Archive — Mon, 19 Dec 2011 13:32:11 +0000

[...] Auto-create User Domain Object with Spring Security [...]

[WORDPRESS HASHCASH] The comment’s server IP (62.149.141.126) doesn’t match the comment’s URL host IP (62.149.140.124) and so is spam.

Comment on Xerces and xml-api Dependency Hell by stefan

stefan — Tue, 29 Nov 2011 10:28:08 +0000

Great post – thanks a lot!

xml-apis:xml-apis v1.4.01 was released 20-Aug-2011 – would be great if you could write a follow-up post that includes this newest artefact.

Comment on GPars performance test by eric

eric — Mon, 10 Oct 2011 14:24:53 +0000

The jsr-166y libs are installed with groovy 1.8.x, but you don’t need to remove them in order to use the GParsExecutorPool as far as I know. Just use the GParsExecutorPool,withPool statement rather than the GParsPool.withPool.

Comment on GPars performance test by nooruddin

nooruddin — Mon, 10 Oct 2011 10:02:43 +0000

hi,
Great Article. I am working on similar issue wherein i need to test between GParsPool and GParsExecutorPool. The reference guide of GPars stated in 3.1.2 section that “It needs to be states, however, that GParsPool performs typically much better than GParsExecutorsPool does.”
It also says that GParsExecutorPool doesn’t need the jsr-166y and uses JDK’s executor service to achieve concurrency. CAN ANYONE TELL ME WHEN DOES THE jsr-166y jar GETS installed.?? I mean does it gets installed when i execute install-plugin command from my spring source toolsuite.? I want to try GParsExecutorPool so i need to remove the jsr-166y jar file for that.

Thanks in Advance. – Noor

Comment on GPars performance test by eric

eric — Mon, 12 Sep 2011 00:24:18 +0000

GParsPool.withPool has worked fine for us so far, though we have only used it with some limited calls.

Comment on GPars performance test by Conor Mullen

Conor Mullen — Fri, 09 Sep 2011 20:36:08 +0000

Great article, very helpful with getting started.

I had noticed then when I use a GParsPool.withPool instead of a GParsExecutorPool.withPool that the GParsPool version of the code does not destroy the threads and seems to run forever without joining back to execute the next piece of code. Whereas with GParsExecutorPool, once each thread is complete it exits.

Have you ever seen this behaviour before?

Comment on GPars performance test by eric

eric — Mon, 29 Aug 2011 20:21:18 +0000

Thanks for the reminder. I forgot to include the pool size info in my write up because I tried it early on in my tests (using a size of 10) with the math and file operations and found that the results were 30% and 10% slower, respectively. I just tried it with the URL retrieval test and didn’t get any difference with my local test server. I tried again using a remote page that’s about 300k in size and got these results:

Normal: 10 in 10574 (1/s)
GParsPool(3): 10 in 11775 (1/s)
GParsExecutorsPool(3): 10 in 7062 (1/s)

GParsPool(5): 10 in 13895 (1/s)
GParsExecutorsPool(5): 10 in 7378 (1/s)

Each iteration (of which their were 10) retrieved 5 pages. The first test run used a pool of 3 and the second used a pool of 5 (one for each page). I don’t have a good explanation for why the jsr166-based GParsPool was slower than the executor-based pool. I’ll post again if I have time to come back to it and dig deeper.

Comment on GPars performance test by Vaclav pech

Vaclav pech — Mon, 29 Aug 2011 16:45:56 +0000

Thanks for posting your results. GPars gives you one more knob to try turning for better parallelism – the withPool method takes an optional int argument specifying the number of threads in the pool.

E.g. GParsExecutorsPool.withPool 10, {…}

Since you’re running on a dual-core, the pool holds three threads by default. While this is close to the optimum for CPU-bound calculations, networking may benefit from higher parallelism. You may try to increase the number of threads till you start hitting the bandwidth wall.

Comment on Representing an XML qualified name as a string by Graig Wiley

Graig Wiley — Tue, 14 Jun 2011 13:46:37 +0000

Oh oh!
This is a very good post, simple and easy
Excellent article, bookmarked for future referrence.
Thanks!

[WORDPRESS HASHCASH] The poster sent us ’0 which is not a hashcash value.