It's remarkably close to what I'd write today. The biggest difference now is the numbers. Now on a typical weekday we get about 150k unique IPs and 1.5m pageviews.
Anecdote: My brother is a construction engineer with Ellis Don and he goes to HN even though he has no interest in startups because he likes the "quality of thought and discourse for current events" and is working his way through learning python the hard way in order to script some of his more mundane tasks.
Excellent story! Discussion on HN is indeed second to none.
Would you brother mind sharing what kind of tasks he would like to script? I am a big fan of Python and it is always interesting to see how people use it in various industries.
Sure. Since Ellis Don is a huge company they often end up doing join partnerships with other huge companies because large projects often need companies with multiple deep fields of expertise. This results in needing to use both companies systems to track things like concrete truck tickets, personnel on site, meeting minutes, etc.
Usually the tools that are built for very large construction companies use fairly open file formats, like xml, or CSV files. Instead of having to double enter the data, he'd like to be able to enter in the information once, then have it propagate to the other system. This would free up about 5 hours a week for him on his current project that is expected to take about 3 years. So the total time saved would be about 750 hours, minus the time it took him to program the system.
That would be the most complex case. Others include needing to rename hundreds of thousands of site photos, or to include in an email only photos that matched a certain pattern (for example: duff-site_2011-09-14_4201.jpg would match a pattern looking for all the photos on the duff site taken in September but only between floors 40 and 45).
Other programs he would like would be to import into their in house tracking system any invoices his subtrades email him, again, matching a certain pattern (dig-con_final-invoice_duff-site_req-345.pdf).
How has the scaling effort gone? In the first post you mentioned that you were just rewriting code to remove bottlenecks. Have you had to make any more exotic and interesting changes to the code base since then to support the growth?
Point of comparison, as of March 2011 Stack Overflow, which got about 2x the traffic for page hits/month, used 12 cores for the web tier and 12-24 cores for their SQL Server 2008 database(s?). Source - http://highscalability.com/blog/2011/3/3/stack-overflow-arch...
Does anyone know what db engine serves as backend for HN?
A Stack Overflow page I'd imagine would be so much more complex than a Hacker News page to render. Would likely require lots of complex queries for everything going on around the primary question and answer mechanism. By contrast Hacker News seems fairly simple.
These are the results of different constraints. SO wants a slick, interactive interface with a lot of access to long-tail content. HN wants inexpensive and easy-to-maintain that still allows discussion of active topics that fade from view fairly quickly.
This is great to hear. Growth without a corresponding decline in quality. I must admit, when I read the original article, I was a little skeptical it would work out well.
Are there any design choices you would particularly attribute the success to?
