11-08, 11:40–12:20 (US/Eastern), Central Park West
Modern OLAP databases are amazing for doing quite simple calculations at scale, but is that what we need? I'll argue that online analytics processing isn't that useful for data science, and that the raw read and write throughput of these databases doesn't impress. OLAP is a business model, where 'Online' means 'someone else's environment', and as a result it gets marketing and mindshare. We need to be discerning about our choice of tools. I'll look at why and how we've optimised our workflow for throughput instead, along with working with data collaboratively where data-science (and in our experience Quant) is the focus, and what that means in terms of improved flexibility and productivity for users.
In this talk I'll present on my experience building systems for productive data-science, quant research and analytics. I'll look at where OLAP works well, and where other approaches are more productive. I'll also introduce ArcticDB, some of it's novel features and attempt a live demo.
Introduction - background to the use-cases, quant, data-science and analytics in finance
OLAP - what is it good for? Do we use it too much?
ArcticDB - it's design and how it addresses doing work in local/client environments whilst remaining a multiuser database
Demo - ArcticDB examples and performance
Q&A
No previous knowledge expected
James Munro is Head of ArcticDB at Man Group. ArcticDB is a high-performance data-frame database that is optimised for time-series data, data-science workflows and scales to petabytes of data and thousands of simultaneous users.
James was previously CTO at Man AHL between 2018 and 2023. He joined Man Group in 2011 as a quant developer and has worked with Man AHL’s portfolio management, FX, commodities, fixed income, equities and volatility teams.
James holds a PhD in Theoretical Physics from University College London.