Wednesday, July 16, 2014

Book Review of Java Performance: The Definitive Guide

While tools and technologies change rapidly, and looking up information online is sometimes the best way to get the information you need, I can be useful to occasionally read a book to get oriented in a subject and discover what you didn't know that you didn't know. For me a good technical book sets the context for the problem and gives you enough information to apply what you learned to harder problems that the book covers, but which also gives you information you can apply immediately.  Java Performance: The Definitive Guide does a good job of both.

Not just about JVM params, the book covers application and algorithm issues, database connectivity, as well as JVM issues such as garbage collection algorithms. What is useful, and sadly rare,  is  that this book not only tells you things you can do,  it also tells you things that you can skip, since not every thing you do has a good cost/ benefit payoff.

With an emphasis  on standard tools, including open source and those that come with the JDK,  you can apply what you learned immediately. The book does not cover every tool, and may not include your favorite, but the discussion has enough  tools to get you started.

This book will be useful for both those new to programming in Java (since performance and resource use area rarely emphasized who you are learning to program) and those who have been programming a while but have not spent as much time thinking about performance as one might like.

I got my copy of this book through the Amazon Vine program.

Saturday, March 22, 2014

Estimation as a requirements Exercise

I explored the role of estimation in a couple of articles on Techwell recently.

In the first article I discussed how teams balance the cost of estimation (in terms of time it takes) with the value it brings to the project. Some argue that estimation isn't very useful at all, where other's say that it can be useful, but that all stakeholders may not have the same vision of the value estimation brings.

In a follow up article I explore the debate about whether to estimate in hours, which reflects effort, and time, or  points, which reflect complexity. This is a common debate, since many articles on agile advocate points, to step away from the concepts of estimate, and focus on the complexity of a feature, and also to help teams move towards the model of the estimate being valid for the team and not just a particular person. And stakeholders are often concerned about schedule and deadlines, so tend to prefer hours initially.

Different teams will come to different decisions about what works best for them. To me the most important part of the estimation discussion is the part many teams don't have, namely asking (and answering) the question of why they are estimating, and what value the estimates bring to the project given that they are now working with an agile process.

As I think more about what value estimation has brought to teams that I have worked on, I realize that the biggest value is that of having a a discussion of scope. When you have an planning meeting, a few questions come up:

  • Why people have different estimates
  • Why the estimate is larger, or smaller, than the product owner expected
  • Whether the team really understands what the story means.
Give this I'm leaning towards thinking of estimation as being more about requirements definition rather velocity, effort, or even complexity.  To that end, maybe the approach to use is one where you spend all of your planning effort defining stories in terms of small, fixed size units, and your velocity is about how many stories you finish off of a prioritized backlog. I link to a description of what this means in the points and hours article.

I'm  interested in hearing about creative approaches your teams have used for estimation. Please comment on  the Techwell articles, or here if you want to share lessons you have learned.