Scalability

http://techblog.netflix.com/2010/12/5-lessons-weve-learned-using-aws.html Notes from netflix tech blog about some key points of their EC2-based infrastructure design. Note that netflix is not affected by the current EBS problems in AWS US-EAST region. Their key point seems to be to equally split your infrastructure between 3 AZs (availability zones) and run at ~30% capacity, so if 2 AZs fail (as we have now) you would still running in one AZ @ 90%

http://www.kriha.de/krihaorg/dload/ultra.pdf a draft of the book “Scalability and Availability Aspects” by Walter Kriha. Many up-to-date info on theory and practice of building scalable high-loaded web/internet systems

http://www.softimage.net/2009/05/09/converting-a-unix-com-site-to-windows/ Notes from Hotmail team about conversion of hostmail apache/bsd -based webfarm to windows 2000. (they have converted CGI to ISAPI filters running on IIS 5.x )

http://www.insight-it.ru/masshtabiruemost/arkhitektura-vkontakte/ [ENG] http://translate.google.com/translate?u=http%3A%2F%2Fwww.insight-it.ru%2Fmasshtabiruemost%2Farkhitektura-vkontakte%2F&sl;=ru&tl;=en&hl;=&ie;=UTF-8 some info about internals of vkontakte.ru - russian facebook clone, most popular social network in Russia, #34 in top100 websites according to alexa (btw, Flickr.com is #35).

http://www.imperialviolet.org/2010/06/25/overclocking-ssl.html Notes from google ppl about their optimizations for SSL connections.

http://events.linuxfoundation.org/slides/2010/linuxcon2010_wheeler.pdf PDF slides that show what happens if you create a file system with one billion (10**9 ) files

http://3.rdrail.net/blog/memcached-based-message-queues/ Blog pots shows and explains how to implement simple queue system on top of memcached

http://wiki.secondlife.com/wiki/Message_Queue_Evaluation_Notes Second Life notes about Message Queue Systems

http://jeremy.zawodny.com/blog/archives/010511.html notes on existing open source queue/messaging systems

http://highscalability.com/blog/2009/10/26/facebooks-memcached-multiget-hole-more-machines-more-capacit.html also see http://dormando.livejournal.com/521163.html Describing some problems with performance/scalability you may have if you partition/shard your data by hash. ( that’s what memcached does.) You may hit the cpu bound limit from the amount of requests that few very popular keys receive. Currently, there is now common way around this (perhaps because very very few sites scale to the point when this becomes an issue )