I was a bit disappointed at this talk as I felt it had a great deal of promise to deep dive into complicated Big Data architectures and offer real world tuning advice. Instead what we got was a description of the ways to approach tuning and optimisation (test, capture, change, test, capture) which is a standard way to approach performance tuning - nothing earth shattering there. The talk went on to describe the use of vTune, an intel product for capture hardware counter data to determine the 'under the hood' actions being taken by the hardware in response to the JVM's actions. There are practical uses of this tool, but it is not free and is hardware specific. Towards the end the guys used this to diagnose scenarios such as CPU LLC cache misses, poor compression performance in map/reduce, and over allocation of GC threads in an old JVM instance.
GlassFish 4: From Clustering to the Cloud
Steve was also attending this talk, see his Tuesday blog for a good overview.
Scala Tricks
Venkat Subramaniam
This was a lively talk that kept the crowd entertained throughout. I daresay that if you were choosing your JavaOne talks based on the quality of presenters, you would not be disappointed attending anything that Venkat presents on.
Technically this talk covered some of the basic language features of Scala and how you can use them to build applications. Venkat's focus was on removing 'ceremony' - the concept that boilerplate Java code gets in the way, and that the standard ways of creating Java code ("...inheritance is like a gorilla. Cute when you see them at the zoo but don't bring it home") are holding us back from new ways of solving problems.
In my opinion Scala is powerful when hooked into Akka (for distribution and concurrency), but one of the strongest cards Java plays is it's low barrier to entry - a card which Scala's syntax negates. Whilst I don't think that Java 8's lambda syntax will completely bridge the two camps of new problem solving paradigms and ease of use, I will be keeping my eye on functional programming going forwards as I think it has a place in future scalable architectures. Long may Scala blaze a trail in that space and find innovative ways of bringing these features to the JVM.
Dealing with JVM limitations in Apache Cassandra
Jonathon Ellis Steve's Tuesday blog covered this talk.
Meet the experts: JMS 2.0 Expert Group
Nigel Deakin
Reza Rahman
Me!
Our BOF seemed to go quite well. We summarised the changes to the specification and accepted questions and answers from a crowd of approximately 30-40 people.
Questions from the crowd were focussed on the lack of monitoring and management APIs for JMS given their ubiquity in most vendor implementations, and the future of JMS post 2.0.
Meet the Java EE 7 Specification Leads
Bill Shannon
Linda Demichial
(remaining specification leads from EJB, CDR, JMS, JSF, JAXRS, JSONP)
The aim of this birds of a feather was to get input and feedback from the spec leads as to their feelings on the future of JEE going into, and post, JEE7.
Given the keynote focuses of Simplicity and Community involvement I asked what the spec leads thought the biggest barriers were to greater community involvement and what could be done to break down these barriers. The answers ranged from comments on encouraging individuals to get involved to exposing JCP activity transparently on wikis and open forums. I believe that we need to go a step further and give the process of following specifications as low a barrier to entry as possible, perhaps via periodic podcasts and interaction with non-JCP bloggers as the spec goes on. This will help translate developments to people who do not have the bandwidth to follow a specification wholesale.
I'll be keen to keep tabs on these developments in the coming months. Nick Wright
No comments:
Post a Comment