CAP Twelve Years Later: How the "Rules" Have Changed

"Systems have a tendency to self synchronize as operations line up and try to destroy themselves. Fascinating to watch."

In the modern era, software is commonly delivered as a service: called web apps, or software-as-a-service.

"The implication: high performance equals high tolerances, which means your entire system must be designed to exacting standards."

There's a whole lot in this that just seems right to me.

"The Maginot Line was a huge fortress that ran the length
of the Franco-German border and was constructed at great
expense between World War I and World War II. It
successfully kept the German army from directly crossing
the border between France and Germany. It was quickly
bypassed by the Germans in 1940 who invaded through
Belgium"

We consider the problem of migrating user data between data centers. We introduce distributed storage
overlays, a simple abstraction that represents data as
stacked layers in different places. Overlays can be readily
used to cache data objects, migrate these caches, and migrate the home of data objects. We implement overlays as
part of a key-value object store called Nomad, designed to
span many data centers. Using Nomad, we compare overlays against common migration approaches and show that
overlays are more flexible and impose less overhead. To
drive migration decisions, we propose policies for predicting the location of future accesses, focusing on a web
mail application. We evaluate the migration policies using
real traces of user activity from Hotmail.

"Detecting failures is a fundamental issue for fault-tolerance in distributed systems. Recently, many people have
come to realize that failure detection ought to be provided as some form of generic service, similar to IP address
lookup or time synchronization. However, this has not been successful so far. One of the reasons is the difficulty
to satisfy several application requirements simultaneously when using classical failure detectors."

"The most important contribution of SEDA, I think, was the fact that we made load and resource bottlenecks explicit in the application programming model."

When your systems are distributed, with varying throughput and resiliency, you can't treat pieces of your architecture as function calls, returning data instantly, consistently, and reliably. The "load and resource bottlenecks" are integral parts of the logic of your software and contracts have to be written and programmed against.

SEDA: An Architecture for Well-Conditioned,
Scalable Internet Services

How Etsy built their activity feed. Mention of how a certain part is a hack and won't scale is interesting...

A little light on details but talks a bit about how they use Amazon SimpleDB, Cassandra, and HBase.

"Multiplayer gaming is about shared reality: that all of the players feel they are in the same world, seeing from differing viewpoints the same events transpiring within that world. As multiplayer gaming has evolved from the little 2-player modem games that characterized Doom, into the large, persistent, more free-form interactions of games like Quake 2, Unreal, and Ultima Online, the technologies behind the shared reality have evolved tremendously."

"We show that some of the problems are NP-hard yet admit effective heuristics..." 

Effective heuristics are always good.

Paxos is a family of protocols for solving consensus in a network of unreliable processors. Consensus is the process of agreeing on one result among a group of participants. This problem becomes difficult when the participants or their communication medium may experience failures.

Just great stuff.

Facebook also has an extensive MySQL infrastructure, but they found performance suffered as data set and indexes grew larger. And they could have built their own, but they chose HBase.

Speaking respository of a former flickr engineer, Cal Henderson. Lots of information about PHP and scaling, obviously.