Impressive, wonder if they were running this on a single node or more widespread? In a wide spread environment I think Erlang would be the true winner, though it does not natively have macros :-( There is an implementation of Lisp for Erlang called LFE (lisp flavored Erlang) which I looked at, which does have macros and a real engine underneath. But clojure is an awesome combination
On Sep 7, 7:32 am, Sean Corfield <seancorfi...@gmail.com> wrote: > I just wanted to share this experience from World Singles... > > Back in November 2009, we started developing with Scala. We needed a > long-running process that published large volumes of changes from our > member database as XML packets published to a custom search engine. > The mapping from half a dozen tables in the database to a flat XML > schema was pretty complex and the company had tried a number of > solutions with mixed success in the past. I introduced Scala based on > the promises of performance, concurrency and type safety - and > conciseness (especially with XML being a native data type in Scala). > > We've been running the Scala publishing daemons in production for most > of two years. Generally they work pretty well but, under stress, they > tend to hit Out of Memory exceptions and, after a lot of poking > around, we became fairly convinced it was due (at least in part) to > the default actor implementation in Scala. Scala is going to fold in > Akka soon and we had been considering migrating to Akka anyone... > > But having introduced Clojure this year (after experimenting with it > since about May last year), we figured we'd have a short spike to > create a Clojure version of the Scala code to see how it worked out. > > It took about 15 hours to recreate the publishing daemon in Clojure > and get it to pass all our tests. Today we ran a "soak test" > publishing nearly 300,000 profiles in one run. The Scala code would > fail with OoM exceptions if we hit it with 50,000 profiles in one run > (sometimes less). The Clojure code sailed thru and is still happily > running - so we'll be replacing the Scala code during our next > production build. > > The other aspect that's interesting is that the Scala code totaled > about 1,000 lines (about 31k characters of code). The Clojure > replacement is just under 260 lines (around 11.5k characters of code). > Neither code base has much in the way of comments (*ahem* - I'm not > proud of that, just pointing out that there's no "noise" offsetting > the code comparison). That doesn't include unit tests either, it's > just the raw production code. The form of the Clojure code mostly > follows the form of the Scala code, most of the same functions - it > was very functional Scala - with some refactoring to helper functions > to make it more modular and more maintainable. > > The net result is (obviously) that we'll be taking the Clojure > publishing daemon to production and we'll be dropping Scala > completely. > > Kudos to Rich Hickey and the Clojure/core team for creating a great > general purpose language that can solve big problems - thank you! > -- > Sean A Corfield -- (904) 302-SEAN > An Architect's View --http://corfield.org/ > World Singles, LLC. --http://worldsingles.com/ > Railo Technologies, Inc. --http://www.getrailo.com/ > > "Perfection is the enemy of the good." > -- Gustave Flaubert, French realist novelist (1821-1880) -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
