Path: csiph.com!x330-a1.tempe.blueboxinc.net!usenet.pasdenom.info!weretis.net!feeder4.news.weretis.net!news.musoftware.de!wum.musoftware.de!fu-berlin.de!uni-berlin.de!individual.net!not-for-mail From: Giovanni Azua Newsgroups: comp.lang.java.programmer Subject: Re: NIO multiplexing + thread pooling Date: Sun, 25 Sep 2011 11:13:28 +0200 Lines: 37 Message-ID: References: Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit X-Trace: individual.net DIpD4xTmdoEClqM+ZzMz9gzkbJo4rfF5cSGs9vBOL6oHdAMwlJ Cancel-Lock: sha1:YcZYuD/TRr9pteRelkR+thpovwE= User-Agent: Microsoft-Entourage/12.30.0.110427 Thread-Topic: NIO multiplexing + thread pooling Thread-Index: Acx7Y2MCksJpEldZRkq3dM0sdgHRlQ== In-Reply-To: Posted-And-Mailed: yes Xref: x330-a1.tempe.blueboxinc.net comp.lang.java.programmer:8286 Hello Matthews, On 9/25/11 6:26 AM, in article "John B. Matthews" wrote: > This related thread may be helpful: > > /cd054f226e2d5c82> > I found that presentation and following threads in internet yesterday and I must admit I was surprised :) I actually don't exactly know where this NIO being faster stuck from but it is widespread knowledge already that it is supposed to be faster :) I also read about blocking NIO which might be more efficient than classic IO. So apparently there are three choices to consider here. The meat of course I am taking "Advanced Performance System Analysis" (which is a mandatory CS Master subject at ETH) is not so much about building the architecture but more about modeling it quantitatively i.e. Modeling throughput, runtime, model the performance using Queuing Theory, Little Law; do experimental analysis and ANOVA to find which parameters (CUT) more drastically affect performance e.g. might be number of threads in ThreadPoolExecutor or the number of database servers. That's why I was thinking supporting both means of communication would be interesting for me so I would 1) finally learn NIO and 2) have this choice of communication as another "degree of freedom" in the system as part of the performance experimental analysis i.e. Try to answer whether NIO is faster than classic IO but I guess this will depend a lot on the quality of my implementation. Thank you. Best regards, Giovanni