Stuff The Internet Says On Scalability For April 8, 2011

Submitted for your reading pleasure on this tomato killing frosty morn...

• Now we really know why vampires feed on blood...they are elastically acquiring more compute power. Your Next Computer May Be Made of...Blood! It's those memristors again. Ancient vamps are really just giant super computers.
• Twitter now at 155 million tweets a day, up from 55 million a year ago.
• 10,000-core Linux supercomputer built in Amazon cloud By Jon Brodkin. The 10,000 cores were composed of 1,250 instances with eight cores each, as well as 8.75TB of RAM and 2PB disk space. The cluster ran for eight hours at a cost of $8,500.
• Quotable Quotes for $273 Alex:
  • @davidklemke: Holy balls Windows Azure Tables is awesome. Man am I regretting not getting into this cloud stuff sooner, it's scalability heaven.
  • @nik: The volume of tweets we are flowing into HBase is truly staggering #bigdata #datasift
  • @wattersjames: One of the key points I mentioned before: Scalability is being able to deploy on the smallest systems that meets your req's
  • @swiftriver: Great #data quote: "More is not better. Better is better." #bigdata #datascience
  • @marcelomilo: YouTube is almost entirely written in Python. Oh! think of the scalability of a Python based system...
• Quora: How to achieve high scalability of Distributed Transactions within SOA? and What are the main mistakes Google has made in scaling its organization?
• Resilience Engineering: Part I shows you just can't keep John Allspaw down. The core idea: builders of safe complex systems must embrace the messiness of the human element, not just blame all failures on our lack of machine like qualities: it's better to ask why humans could make an error than to blame humans for the error; complex failures do not originate from linear causes; learn from what goes right; optimize for recovery time, not failure interval; safety is adaptation. For information about resiliency in a larger context take a look at Resilient Communities. 
• Guy Steele & Richard Gabriel: 50 in 50. Either the strangest beat poetry you've ever heard or the stragest programming language history you've ever heard. In either case, it's interesting. 
• Cliff Click of Azul Systems knows his stuff. In A JVM Does That? he reveals the 5 illusions of  the JVM. These illusions allow programmers to think: bytecodes are fast and have a reasonable cost model, you can quickly change the program at any time, infinite memory, consistent threading and memory model, quick time access, that these other illusions all work on all sorts of machines from cell phones. But we want to be beguiled even further with illusions of infinite Stack, code is data, Integers are as cheap as ints, memory supports atomic update, a JVM run utterly different languages. Remember we just wanted mirrors that gave us the illusion of youth and beauty? Also take a look at Clojure: STMs vs Locks.
• Fantastic overview of all the Key Takeaway Points and Lessons Learned from QCon London 2011. With over a 100 speakers something there should increase your wisdom.
• Optimizing Cache Performance on a Rapidly Growing Site. Shows how Posterous was able to use tools to diagnose and fix their caching problems. Very useful bit of technique.
• What should a machine learning class include?  John Langford tells us in The Ideal Large Scale Learning Class. Looks like fun.
• Paul Tero with a very nice introduction to Speeding Up Your Website’s Database. Covers all the bases in a quality way.
• Virtualization and IO Modes = Extra Complexity By Peter Zaitsev. The point is Virtualization adds complexity and there are at least some cases when you may be lied to about IO completion.
• Looking for bloggers to read? The Top 100 Bloggers on Cloud Computing
• Level-triggered and edge-triggered by  Mike Burr. In a distributed system you should move state information, not work items. Distribute information about the desired state and let each node determine how best to get to that state.
• Google Groups: Which database is most appropriate for sensor logging and Ditching Cassandra and Latency on a node based massive multiplayer web lightcycle game
• Dzmitry Huba shows some code for Parallel merge sort. Glad we got that sorted.
• Stack Exchange: Windows Azure vs Amazon EC2 vs Google App Engine. Really good exchange on the tradeoffs between the systems.
• Can Rob Kalin Scale Etsy? "We want to allow the makers of the world to claim authorship for what they're making. This is what Etsy stands for: The little guy being able to organize a better marketplace."
• MySQL Performance: Reaching 100M(!) Transactions/sec with MySQL 5.5 running on Exadata!