Most of us at RabbitMQ HQ have spend time working in a number of functional languages in addition to Erlang, such as Haskell, Scheme, Lisp, OCaml or others. Whilst there is lots to like about Erlang, such as its VM/Emulator, there are inevitably features that we all miss from other languages. In my case, having spent a couple of years working in Haskell before returning to the RabbitMQ fold, all sorts of features are "missing", such as laziness, type classes, additional infix operators, the ability to specify precedence of functions, fewer parenthesis, partial application, more consistent standard libraries and do-notation. That's a fair list, and it'll take me a while to get around to implementing them all in Erlang, but here are two for starters.

In our previous blog post we talked about a few approaches to topic routing optimization and described the two more important of these in brief. In this post, we will talk about a few things we tried when implementing the DFA, as well as some performance benchmarking we have done on the trie and the DFA.

RabbitMQ 2.4.0 introduced an extension that allows publishers to specify multiple routing keys in the CC and BCC message headers. The BCC header is removed from the message prior to delivery. Direct and topic exchanges are the only standard exchange types that make use of routing keys, therefore the routing logic of this feature only works with these exchange types.

I decided to run some benchmarks of my AMQP encoder/decoder (AMQ Protocol gem) against the old one in the AMQP gem to see whether it performs better or not. So far I did only the most basic optimisations like storing reusable values in constants, nothing special (yet).

I did two sets of benchmarks: CPU time benchmarking using my fork of RBench with support for custom formatters (like writing results into a YAML file) and memory benchmarking using Object.count_objects (Ruby 1.9).

In many messaging scenarios, you must not lose messages.  Since AMQP gives few guarantees regarding message persistence/handling, the traditional way to do this is with transactions, which can be unacceptably slow.  To remedy this problem, we introduce an extension to AMQP in the form of Lightweight Publisher Confirms.

RabbitMQ 2.3.1 introduces a couple of new plugin mechanisms, allowing you much more control over how users authenticate themselves against Rabbit, and how we determine what they are authorised to do. There are three questions of concern here:

1. How does the client prove its identity over the wire?
2. Where do users and authentication information (e.g. password hashes) live?
3. Where does permission information live?

Question 1 is answered in the case of AMQP by SASL - a simple protocol for pluggable authentication mechanisms that is embedded within AMQP (and various other protocols). SASL lets a client and a server negotiate and use an authentication mechanism, without the "outer" protocol having to know any of the details about how authentication works.

SASL offers a number of "mechanisms". Since the beginning, RabbitMQ has supported the PLAIN mechanism, which basically consists of sending a username and password over the wire in plaintext (of course possibly the whole connection might be protected by SSL). It's also supported the variant AMQPLAIN mechanism (which is conceptually identical to PLAIN but slightly easier to implement if you have an AMQP codec lying around). RabbitMQ 2.3.1 adds a plugin system allowing you to add or configure more mechanisms, and we've written an example plugin which implements the SASL EXTERNAL mechanism.

From time to time, on our mailing list and elsewhere, the idea comes up of using a different backing store within RabbitMQ. The backing store is the bit that's responsible for writing messages to disk (a message can be written to disk for a number of reasons) and it's a fairly frequent suggestion to see what RabbitMQ would look like if its own backing store was replaced with another storage system.

Such a change would permit functionality that is not currently possible, for example out-of-band queue browsing, or distributed storage, but there is a fundamental difference in the nature of data storage and access patterns between a message broker such as RabbitMQ and a generic database. Indeed RabbitMQ deliberately does not store messages in such a database.

I'm happy to announce that the AMQP 0.7 is released, as I promised in the previous blog post. So what are the changes?

In the past year development of the AMQP gem was practicaly stagnating, as its original author Aman Gupta (@tmm1) was busy. A lot of bugs stayed unresolved, the code was getting old and out-dated and no new features or documentation were made.

At this point I started to talk with the RabbitMQ guys about possible collaboration on this. Actually originally I contacted VMware when I saw Ezra Zygmuntowicz looking for people to his cloud team, but when I found that VMware recently acquired the RabbitMQ project in London, I got interested. I signed the contract, switched from script/console to Wireshark and the RabbitMQ Tracer and since November I've been happily hacking on the AMQP and AMQ-Protocol gems.

We have been prototyping support for a new protocol, as is our wont. This one is called "AMQP 1.0 R0", and it is the new issue from the AMQP working group (of which RabbitMQ, and latterly VMware, are a member). The "R0" indicates that it's the first revision of a recommendation. The specification is incomplete: there are many TODOs, and to a large extent it is unproven. Those two facts are part of what prompted this prototyping.

The prototype code is mirrored at github: http://github.com/rabbitmq/rabbitmq-amqp1.0. It is built just the same as all our plugins.

The AMQP 1.0 R0 specification differs from the specification of previous versions of AMQP, in that it does not define a broker model; i.e., it doesn't define exchanges queues and bindings, or their equivalents. The protocol is really only about transferring messages from one agent to another, and then agreeing on what the outcome was. That means it is amenable to bolting on to a message broker implementation, among other uses -- the idea is that one can adapt an existing model to suit.

In our case, the incumbent model is that of AMQP 0-9-1, with some generalisations and extensions (for example, chained bindings). Our target with the prototype is therefore to be able to get something useful done with both 1.0 clients and 0-9-1 clients connected at the same time.

Well, the good news is, we've achieved that. In fact the plugin can be set up to replace Rabbit's usual network listener, and will happily talk to AMQP 0-8, 0-9-1, and 1.0 clients. We did have to do some invention along the way, and there are some parts of the specification that we are conspicuously not implementing. These will be detailed in the README soon.

One large part of the invention is to fill in semantics where the specification is silent. Some of these are detailed in this client-broker protocol work we did for the AMQP working group. We're hoping the prototyping will help fill this out some more.

Next week I'll be taking our prototype to the AMQP 1.0 "Connectathon", where it'll be tested against other implementations of the core protocol (not all of which are open source). Again, this will help to flush out barriers to interoperability in the specification.