<antirez>

Yesterday distributed systems expert Aphyr, posted a tweet about a Redis Sentinel issue experienced by an unknown company (that wishes to remain anonymous):

“OH on Redis Sentinel "They kill -9'd the master, which caused a split brain..."
“then the old master popped up with no data and replicated the lack of data to all the other nodes. Literally had to restore from backups."

OMG we have some nasty bug I thought. However I tried to get more information from Kyle, and he replied that the users actually disabled disk persistence at all from the master process. Yep: the master was configured on purpose to restart with a wiped data set.

The first commit I can find in my git history about Redis Cluster is dated March 29 2011, but it is a “copy and commit” merge: the history of the cluster branch was destroyed since it was a total mess of work-in-progress commits, just to shape the initial idea of API and interactions with the rest of the system.

Basically it is a roughly 4 years old project. This is about two thirds the whole history of the Redis project. Yet, it is only today, that I’m releasing a Release Candidate, the first one, of Redis 3.0.0, which is the first version with Cluster support.

Queues are an incredibly useful tool in modern computing, they are often used in order to perform some possibly slow computation at a latter time in web applications. Basically queues allow to split a computation in two times, the time the computation is scheduled, and the time the computation is executed. A “producer”, will put a task to be executed into a queue, and a “consumer” or “worker” will get tasks from the queue to execute them. For example once a new user completes the registration process in a web application, the web application will add a new task to the queue in order to send an email with the activation link. The actual process of sending an email, that may require retrying if there are transient network failures or other errors, is up to the worker.

-----------------
UPDATE: The algorithm is now described in the Redis documentation here => http://redis.io/topics/distlock. The article is left here in its older version, the updates will go into the Redis documentation instead.
-----------------

Many people use Redis to implement distributed locks. Many believe that this is a great use case, and that Redis worked great to solve an otherwise hard to solve problem. Others believe that this is totally broken, unsafe, and wrong use case for Redis.

The strong reactions about the recent OpenSSL bug are understandable: it is not fun when suddenly all the internet needs to be patched. Moreover for me personally how trivial the bug is, is disturbing. I don’t want to point the finger to the OpenSSL developers, but you just usually think at those class of issues as a bit more subtle, in the case of a software like OpenSSL. Usually you fail to do sanity checks *correctly*, as opposed to this bug where there is a total *lack* of bound checks in the memcpy() call.

Generally speaking, I love randomized algorithms, but there is one I love particularly since even after you understand how it works, it still remains magical from a programmer point of view. It accomplishes something that is almost illogical given how little it asks for in terms of time or space. This algorithm is called HyperLogLog, and today it is introduced as a new data structure for Redis.

Counting unique things
===

Usually counting unique things, for example the number of unique IPs that connected today to your web site, or the number of unique searches that your users performed, requires to remember all the unique elements encountered so far, in order to match the next element with the set of already seen elements, and increment a counter only if the new element was never seen before.

Yesterday and today I managed to spend some time with linenoise (http://github.com/antirez/linenoise), a minimal line-editing library designed to be a simple and small replacement for readline.
I was trying to merge a few pull requests, to fix issues, and doing some refactoring at the same time. It was some kind of nirvana I was feeling: a complete control of small, self-contained, and useful code.

There is something special in simple code. Here I’m not referring to simplicity to fight complexity or over engineering, but to simplicity per se, auto referential, without goals if not beauty, understandability and elegance.

The title of this blog post is an apparently trivial to answer question, however it is worth to consider a bit better what performance really means: it is easy to get confused between scalability and performance, and to decompose performance, in the specific case of database systems, in its different main components, may not be trivial. In this short blog post I’ll try to write down my current idea of what performance is in the context of database systems.

A good starting point is probably the first slide I use lately in my talks about Redis. This first slide is indeed about performance, and says that performance is mainly three different things.

Today Redis is 5 years old, at least if we count starting from the initial HN announcement [1], that’s actually a good starting point. After all an open source project really exists as soon as it is public.

I’m a bit shocked I worked for five years straight to the same thing. The opportunities for learning new things I had because of the directions where Redis pushed me, and the opportunities to learn new things that I missed because I had almost consistently no time for random hacking, are huge.

In this blog post I’m going to describe a very simple distributed algorithm that is useful in different programming scenarios.
The algorithm is useful when you need to take some kind of information synchronized among a number of processes.
The information can be everything as long as it is composed of a small number of bytes, and as long as it is idempotent, that is, the current value of the information does not depend on the previous value, and we can just replace an old value, with the new one.