Hedge fund AHL Man Group has replaced a range of disparate relational databases with a single data platform built on NoSQL database MongoDB for financial market data.
While the average person might have a view of a trading floor as a cacophonous bear-pit full of aggressive men in red braces, the environment at AHL, says Gary Collier, co-CTO at the group, is “quiet and civilised”. There work the quantitative analysts – “quants” – whose job it is to devise models that can then feed algorithmic trading systems.
“We’ve changed the working model by which those researchers get new sources. It is now more of a self-service model for our users. Only a few years back, onboarding new data sources would have been met with a sharp intake of breath around lead times,” says Collier. “A lot of the data is now onboarded by users.”
The hedge fund has a “beefy compute cluster”, he says, “which was previously constrained by speed of data access”.
“We can now saturate the network and it is no longer a bottleneck,” adds Collier.
Interfacing with Python in MongoDB is very slick
Gary Collier, AHL Man Group
In 2012, Collier and his team of around 40 technologists took a hard look at their data management and analytics environment.
“We needed quick access to historical, financial and other types of data to draw it into a research environment where the quants, and others, can run analytics on it, and devise trading models,” he says.
There was, at that time, a range of computing languages in use at analyst level, including R and Matlab. Collier and his technology team decided to standardise on Python and, serendipitously, have found that their most recent data technology choice of MongoDB as a data store works well with it.
“Interfacing with Python in MongoDB is very slick,” says Collier.
When they had standardised on Python, in 2012, he says they were in a position to then tackle the historical data held in different forms of proprietary databases.
“Essentially, we had a broad landscape of data storage technology which wasn’t great for onboarding new things. Performance wasn’t good and there were lots of moving parts,” he says. That led to the search for a new data store.
The team completed a proof of concept with MongoDB at the end of 2013, then hit a 30 May deadline in 2014, obviating the need to relicense the technology they had been using before, which Collier declined to name.
For more on data management in financial markets
- Data mining techniques for financial markets
- Making money from data: Financial business intelligence at work
- Analysis: IBM discusses Financial Markets Framework
An immediate financial benefit was a 40 times reduction in licensing costs with the move to Mongo. And they have cut disk storage down to 40% of what it was previously.
Collier confirms that senior management at AHL are aware that the firm is using Mongo successfully, and was supportive of its use of as the new data store. For some type of analysis, quants at the firm are now modelling 25 times faster than they were before the full deployment of the NoSQL database.
As well as the use of MongoDB by the quants on market data, the firm is using the technology on the blackbox trading systems, trading hundreds of millions of dollars.
“Not every trading strategy is on there, but it is happening. It’s more chips on the table for MongoDB,” says Collier.