tag:blogger.com,1999:blog-9111971326981479563.post7396862630371443966..comments2023-10-10T14:20:22.306+01:00Comments on ArkAngelOfKaos: London Java Community - Open Conference 2014Unknownnoreply@blogger.comBlogger2125tag:blogger.com,1999:blog-9111971326981479563.post-22347450893417162542015-03-26T01:12:18.328+00:002015-03-26T01:12:18.328+00:00Thanks for adding all that information Mani! I did...Thanks for adding all that information Mani! I didn't see this reply until just now :(<br />Looking forward to the LJC Unconference this year :Darkangelofkaoshttps://www.blogger.com/profile/13857337949466524155noreply@blogger.comtag:blogger.com,1999:blog-9111971326981479563.post-8694228580356672582014-12-01T23:51:51.179+00:002014-12-01T23:51:51.179+00:00I'll +1 the post on the discussions about the ...I'll +1 the post on the discussions about the said topics. It was a pleasure to be in the same room that covered these topics:<br /><br />1) disruptor pattern<br />2) pushing the complexing of concurrency and threading to the language, platform / OS.<br />3) The Actor framework - Akka, uses event buses to exchange messages<br />4) When evaluating technology consult the tech evaluation matrix from Matt Raible - http://bit.ly/jvm-frameworks-matrix/, also a presentation worth watching - https://parleys.com/play/52aee01be4b0cc758a75811b/chapter0/about/<br />5) When possible avoid hard-core tech like threads and low-level, OS level, hardware level stuff, use a framework instead.<br />6) Lot of these technologies were interesting to work with but hard to test, or the tests would be hard to maintain than the implementation themselves.<br /><br />I was happy to plug in a few of my own experiences.<br /><br />I'll add my person experience which I didn't mention on the day are:<br />One of my projects long ago was to process a large ETL transfer from a unknown format to PostgreSQL database, one of the colleagues attempted to do that, too 9 hours of stalling of processes, as the database took forever to write a large chunk of data due to all the memory being hogged by all the processes.<br /><br />So I decided to slice the number of records to process per OS task instead of dumping all the records to the database. And when I did that, the OS and database worked hand-in-hand and within under an hour we were able to transfer all the data into the database, see my experience in a post - http://neomatrix369.wordpress.com/2013/09/24/my-experience-of-learning-r-from-basic-graphs-to-performance-tuning/<br /><br />We didn't need to resort to any high-tech or performance tuning rather than tuning our process and programs a bit.Anonymoushttps://www.blogger.com/profile/07467548077580344443noreply@blogger.com