Book Excerpt and Interview: The Joy of Clojure | Hacker News Reader

Book Excerpt and Interview: The Joy of Clojure

infoq.com

16 pointsKototama10 comments

I have been curious about Clojure for a while. I have the Manning MEAP for his book and I have watched Rick Hickey's introduction videos. My first impression is that is a very well thought out language; my biggest concern is the performance benchmarks : http://shootout.alioth.debian.org/u64q/benchmark.php?test=al...

Clojure seems to be consuming more memory and is generally slower by significant margins compared to Scala ( or java). I would be interested in hearing from anyone using this in a high performance mission critical application like in a bank or in financial services ? What was your response time like under heavy load ? How did you justify this to your management?

I don't think its fair to expect Scala or Java performance from Clojure. Clojure is only three years old. Scala had much more time to work on performance. Clojure will get faster. The community is very performance aware.

In a Bank or financial institute I would rather write something save that can scale but these institute prefere static typing most of the time so Scala may be a better fit.

Sad but true.

What that banks like to use static typing? I can kind of understand it.

UPDATE: This post by the fellow leading the Clojure Alioth effort shows that the JVM GC settings are probably off for some of the benchmarks, http://groups.google.com/group/clojure/msg/8de18e3719aa376a. This shows just how early days it is for Clojure microbenchmarks :)

Those are benchmarks for Clojure 1.2.0. The Clojure 1.3.0 alphas introduce quite a few things that specifically make numerical code quite a bit faster as well as boosting overall performance.

Being competitive with Scala and Java is a real goal for Clojure so expect the performance story to continue improve for some time.

While I don't know of any real-world Clojure cases for mission critical applications - Clojure has been run Azul boxes where the GC is collecting some like 6gbs a sec. Turns out that only takes up about 6% of CPU time. So I think Rich is onto something here about the memory cost of immutable data structures in high performance code and the incredible guarantees of the JVM's GC.

On Alioth microbenchmarks you probably won't see those benefits.

Notice that the other JVM language implementations haven't tried to reduce memory use by tuning JVM GC settings.

The excerpt is at http://www.infoq.com/resource/articles/the-joy-of-clojure/en...

Do you know or have an estimation of when your book will be printed?

According to http://manning.com/fogus/ in January 2011.

dho seems to be correct. Sadly, we as the authors seems to be the last to know the answer to these types of questions. :-(

Reply on news.ycombinator.com