The Holy Java

Building the right thing, building it right, fast

Posts Tagged ‘jvm’

Most interesting links of November

Posted by Jakub Holý on November 30, 2010

This month has been quite interesting, among others I ‘ve picked up several blogs by Adam Bien. I really like his brief, practical and insightful posts :-)

Java, Jave EE, architecture etc.

  • EJB 3.1 + Hessian = (Almost) Perfect Binary Remoting – “With hessian it is very easy to expose existing Java-interfaces with almost no overhead. Hessian is also extremely (better than IIOP and far better than SOAP) fast and scalable.” See also the discussion regarding a “DynamicHessianServlet” (which would be more comfortable then creating a new servlet for each service).
  • Binary XML – Fast Infoset performance results – FI is the name of a standard for binary XML encoding and also of its OSS implementation; according to these measurements (with default settings, compared to Xerces 2.7.1), on average: The FI SAX parser  is about 5 times faster than the Xerces SAX parser, the FI DOM serializer (using default settings) is 25% to 30% faster than the Xerces DOM XMLSerializer, and FI documents are 40% to 60% smaller than XML documents (i.e. it can provide modest to good compression). And: “The use of external vocabularies can be a very effective way to increase the efficiency of parsing, serializing and size at the expense of the fast infoset documents no longer being self-describing … .”
    For the interested ones: FI is built on ASN.1, a standard for describing data structures in a way that is independent of machine architecture and implementation language, it also includes a selection on different binary encoding rules, e.g. DER (triplets tag id – length – value).
  • EJB 3.1 And REST – The Lightweight Hybrid – Why to use @Stateless – there have been recently discussion about Spring vs. EJB and whether @Stateless is of any use. According to this article, the added value of EJB (though some are provided also by Spring) include injection capabilities, transactions, single threading model (for why that is good see #2 on Why I like EJB 3.0/3.1), visibility in JMX, concurrency restriction via thread/bean pools.
  • Why Service Isn’t A ServiceFacade, But ServiceFacade Is Sometimes A Service…
  • Experiences from migrating Hyperic 4.5 from EJB/JBoss to Spring/Tomcat, what good has it brought – simplified unit and integration tests, simplified code thanks to Jdbc/JmsTemplate, … (My comment: EJB 3.1 would certainly bring at least some of the advantages too.)

Performance – large JVM heaps are very much feasible

  • Tuning the IBM JVM for large heaps – tuning 64b IBM JVM 6 with 100GB heap to have acceptable GC times (~ 1/2s as opposed to 25s with the default settings)
  • BigMemory: Heap Envy – there have been recently a lot of fuss about Terracotta’s new BigMemory, a  GC-resistent many GB cache space (using NIO byte buffer). This post discusses it’s disadvantages and compares it with a solution based on ConcurrentHashMap, comming to the conclusion that the old good ConcurrentHashMap outperforms BigMemory considerably.

Other

Posted in j2ee, Languages, Top links of month | Tagged: , , , , , , , | Comments Off

Most interesting links of September

Posted by Jakub Holý on September 30, 2010

The most interesting articles and other IT resources I’ve encountered the previous month, delayed a bit due to my holiday in Andalusia. In no particular order. Included some performance stuff, few tools and few general SW engineering things.

  • The Open-Closed Principle , i.e. open to extension, closed to modification – one of the basic principles in OOP underlying many best practices such as “make all member variables private” nicely explained, with code samples with and without O/CP applied.  Noteworthy: The principle is implemented with abstractions (abstract classes defining the constant part with subclasses being the extensions). You can’t “close” your design against all changes and must thus choose the ones that are more likely, i.e. create a “strategic closure”.
  • Java EE 6 xor Spring by A. Bien – the main difference is in the philosophy behind – Java EE is based on Convention over Configuration. The decision factor is usually the support policy, though. Also the tc Server is certainly better than a “custom” Tomcat with Spring.
  • ThoughtWorks Technology Radar – to help decision-makers from CIOs to developers understand emerging technologies and trends that affect the market today; though their assignment of GWT under hold is disputable (http://www.dzone.com/links/r/thoughtworks_radar_demystified_gwt.html); TW answers: “As it turned out the conciseness of the text didn’t allow us to adequately make our points so that they were not misunderstood. We are interested in a discussion but our opinion about the suitability and usability of GWT has still not changed.”
    • JavaScript as a 1st-class language w/ the same best practices (unit t., refact.,…); functional languages (Clojure > Scala)
    • WS-* beyond the basic profile, GWT, RIA on hold
  • “Agile Business” for a startup with “Virtual Assistants”: outsource (to an “Virtual Assistant”) what you can, only do what’s necessary at the time even if that means doing manually (outsourced) st. that could be automated; this helped to decrease time from 160 MH to 10 MH.
    “The lesson is that before you launch your product, think about the processes you can avoid automating. How about reminder emails? How about monthly billing? Could a human being run a report once a month and send emails or charge credit cards?” , “Every hour spent writing code is wasted time if that code could be replaced by a human being doing the same task until your product proves itself.”
  • InMemProfiler: Identifying Memory Allocators – tool to track memory allocation capable of attributing it to the classes (i.e. packages) of interest without blurring the results with char[] and all the java.lang.* classes
  • Easy Performance Analysis with AppDynamics Lite – I’m fond of performance troubleshooting tools and AppDynamics Lite looks really cool. The post includes two very short yet very informative and nice screencast showing its installation and usage. App. D. automatically discovers Struts actions, JDBC calls, webservices etc. and captures slow operations with the necessary details; this is quite similar to what the open-source Glassbox.com does. The monitoring web UI is very nice and user friendly. See the white box on the right side at http://www.appdynamics.com/lite.php to see what JVMs, ASs and frameworks it supports. Check also the comparison of the lite and standard versions (max 30 transactions, max 2 hours if diagnostics data, …).
  • Google Relaunches Instantiations Developer Tools – Now Available for Free – incl. the static code analysis tool (Eclipse plugin) CodePro AnalytiX
  • A blog about Terracota’s new BigMemory (commercial) claims that 64b JVM with heap over few GB may be a nightmare – “… not all people know about 64-bit JVMs and the nightmare these things can cause. Contrary to what other vendors are claiming, most shops such as Unibet, PartyPoker, Expedia, Sabre Holdings, Intercontinental Hotels Group, JP Morgan, Goldman, and more will tell you a 64-bit JVM pauses unpredictably and for minutes at a time. even when a 64 bit JVM is small (<2GB) it takes 30+% more RAM than a 32-bit equivalent JVM running under the same app.
  • Presentation Real Software Engineering by Glenn Vanderburg  – the talk is pretty interesting and I recommend it. Few, subjectively selected and interpreted points:
    The SW engineering as taught in universities doesn’t work, it’s actually a caricature of “engineering” (that is, established practices that work). The reason is that SwE is unreasonably fascinated by (ideally mathematical) modeling and precise, repeatable processes. This is not how real SW development can or does work. Given the complexity and uncertainty, an empirical process, based on frequent feedback and continual adjustment, is much more suitable. Also we don’t need complex models because prototyping and testing is nearly “free”, compared e.g. to spacecraft engineering. And with BDD and tools like RSpec and FitNesse we may have both readable and executable specifications – the code becomes the model.
  • Monte Carlo Analysis of the Zero Defect Mentality of TDD – conclusion: TDD pays off in the long run even though being slower [learning curve; 0-value bringing defect fixing] Fow short life time, TDD may be not worth it. Don’t argue, simulate :-)

Posted in General, Languages, Testing, Tools, Top links of month | Tagged: , , , , , , , , | Comments Off