Market Data: Applying SRE Techniques to Legacy Designs
|Date||Tuesday, 25 October, 2022 - 15:10–15:30|
|Presenter||George Brighton, Goldman Sachs|
Market Data is the lifeblood of financial applications. Real-time ticker plants must distribute billions of updates per day quickly and robustly, with latency tightly correlated with business performance. Vendor products, rather than in-house software, are employed to handle this task. Due to the criticality of the platform and associated risk aversion, ticker plants evolve slowly, and are behind on modern operational techniques. This talk will convey Goldman Sachs’ key lessons learned while implementing a ticker plant in a cloud environment, including how we increased the number of failure domains, and the techniques we found most useful when overlaying observability.
- 1 of 4 people at Goldman Sachs works in engineering - that’s thousands of people
- This talk may be summed up as an explanation of how to wrap legacy services which don’t natively expose any metrics using blackbox probing techniques, as also explained in for example Google’s SRE books.