Stuff The Internet Says On Scalability For May 6th, 2011

Submitted for your reading pleasure...Hi Mom!...

• We don't need no stinking servers says the W3C. This Could be Big: Decentralized Web Standard Under Development by W3C by Marshall Kirkpatrick. Browsers talking to directly to other browsers. Marshall is right, this could be very big.
• Quotable Quotes for Pi Alex:
  • @eric_brewer The Amazon outage & CAP theorem: http://aws.amazon.com/message/65648/ (partition is the root cause)
  • @kylecordes A problem with cloud hosting (EC2) is that it brings the problems of scalability to systems that *don't* need scalability.
  • @virtualpete Last month everyone was a nuclear physicist. Today everyone is a web scalability architect
  • @jfelipe We cannot overlook migration/federation issues (scalability) in cloud tech: open standards are a plus compared 2 closed (Amazon)
  • @lapsu Stored procedures aren't so bad if you write them in Javascript & they do MapReduce. That makes them cool. #nosql
  • @mohinish  "Scalability" never sounded sweeter.. wishing this is one unicorn that's real!
  • @joelhelbling OH: Scalability has a cost. It's called "joy".
  • @1am1r Relying on someone else's IT to keep your business up is like walking on Times Square without looking at traffic
• Adapteva wants your tablet and phone to have 64 processors. What can you do with all that power? Process the world around you in real-time. Analyzing sound, video, making sense of it, embedding you in a data enchanted world. That's one option anyway. More at Epiphany promises CPU performance breakthrough: The only way to get there is to specialise - you can't do it by putting down more and more general purpose processors. It doesn't scale.
• Gabriel Weinberg of Duck Duck Go with the details on how to create RAID0 ephemeral storage on AWS EC2. He's moved to AWS as his primary data center selecting ephemeral storage instead of EBS, and he's looking at ways to optimize performance, durability and availability. The bottom line for me is that the RAID0 on the m1.xlarge is fast enough such that IO is no longer a significant bottleneck for my usage patterns.
• Unlocked Achievements: Pandora Is Now 10 Billion Thumbs Strong
• How big is the world of cloud computing? Awesome infographic. A Data Center 11.5. times the size of a football field would make a great roller hockey field.
• Jeff Darcy with 10 ways on How To Ruin a Project by turning a promising and fun project into a soul-sucking wreck. Starting a project wasn't in the list, but if you can't find stuff you do in the list then get a different mirror.
• How to Scale Rails to 100 million requests per month by Dan Singerman. The progression: squid cache, memcached, varnish.
• Quora: From a scalability perspective at startup, Do you think we need to embrace NoSQL from day 1 or should we do it incrementally when the user-base(data) increases ?
• Pigs, Bees, and Elephants: A Comparison of Eight MapReduce Languages. Antonio Piccolboni with a good look at the many paths to MapReduce. The way is many but the result is one.
• Response Time, Throughput and Horizontal Scalability. Proper indexing is the best way to reduce query response time.
• With 3-D Transistors, Intel Keeps Moore’s Law Ticking. Stacey Higginbotham lays out this major chip design advance. 37 percent increase in performance, 50 percent decrease is power usage. 3 percent increase in costs. And they look like waffles.
• Scalability Isn’t About the Size of Your “Equipment”, it's a mindset of doing more with less says Kyle Brandt, and that doesn't take a lot of hardware. Also  Use Histograms to Visualize Response Time
• Roberto Scoblo follows up his previous epic Facebook datacenter posts with one of his patented videos: Video tour of Facebook datacenter now up
• Michael Ho gets us thinking what we could do with our own bite sized satellites in DailyDirt: Satellites For Everyone! Please don't say SkyNet. If only the latency was so bad...
• List of Current and Upcoming Cloud Platforms. Andy Hu with a surprisingly large and diverse list of PaaS offerings for a wide range of environments: PHP, Python, Ruby, Java, Node JS, .Net, Perl, Ringo JS, and Javascript. Lots of niches to grow in.
• Putting and Getting Data from a Database. Marko Rodriguez with a quick look at the basics of getting and setting in databases of the primitive, key-value,  document, and graph variety. What distinguishes one database type from another is the structure of the data they store.
• If you are considering Azure then take a look at A Real-World SQL Azure Implementation - An inside look at Quosal’s migration to Microsoft’s cloud computing platform by Megan Keller and Michael Otey. Pro: Access data from anywhere. Con: Backups in Azure are very difficult.
• Are SSDs from Mars? Jeff Atwood finds that solid state hard drives are great until they fail, and they fail a lot.
• Square Pegs and Round Holes in the NOSQL World. Jim Webber with a wonderful exploration of the emergent property of graphs - simply store all the data you like as nodes and relationships in Neo4j and later you'll be able to extract useful business information that perhaps you can't imagine today, without the performance penalties associated with joins on large datasets.
• Good discussion of the differences between scaling and sharding on the NoSQL group. Also take a look at How to speedup Hbase query throughput and Efficient Faceting.
• Picking a database is just a start, data must still be modeled. Bryan Keller explores options for the one-to-many question.
• 10 EMERGING TECHNOLOGIES 2011. The next big thing will likely be confound and be something else entriely, but these are the contenders: social indexing, homomorphic encryption, cloud streaming, synthetic cells, crash-proof code, separating chromosomes, smart transforms, gestural interfaces, cancer genomics, solid-state batteries. I'm surprised something about education isn't in the list.
• Memory Bandwidth and GPU Performance. David Kanter with a deep on GPU performance. In some cases, the GPU with the lower GFLOP/s actually delivers the best performance – which is totally counter-intuitive. 
• Node.js gets a lot of hype, but Ruby has had evented interfaces for a long time. Mathias Meyer takes a look at how the EventMachine works in more detail. EventMachine is based on the idea of an event loop, which is basically nothing more than an endless loop.
• Google wants to Increase TCP's Initial Window. Several large scale experiments showing that the higher initial window improves the overall performance of many web services without risking congestion collapse. 

Comments

[John]

"How to Scale Rails to 100 million requests per month" -- LOL that's not exactly high scalability. 100M requests per month is 3.3M req per day. That's about 38 request per second on average. let's say peak time is 4 times the average -- that's still only 150 rps. Not exactly a lot of traffic.