Live data from Hacker News

How exchanges turn order books into distributed logs

quant.engineering

71–73 of 73 posts

Re: How exchanges turn order books into distributed logs

#71
post #37

Earlier quoted context omitted.

B/c, by design, you want the archived stream of events to include everything. e.g. a lot of these systems have a "replay" node that can be used by components that just restarted. You want the replay to include ALL of the messages seen so you can rebuild the state at any given point. (There are, of course, tradeoffs to this so I'm just commenting on the "single sequencer" design philosophy)

by definition: an exchange doesn't need any reference to outside market data even for systems built on a sequencer which do (e.g. an OMS), the volume is too large the usual strategy is for processes which require it, is to sample it and them stamp it on commands which maintains the invariants (my background: I have been a developer on one of Mike Blum's original sequencers)

Fair point that you don't actually need the market data you send out if you have the original order book data internally.

Re: How exchanges turn order books into distributed logs

#72
These distributed sequencer solutions are for resilience, and they add a lot of latency because each node needs to do something like RAFT. Exchanges generally don't care aggressively about low latency, they care about resilience and fairness. It's the hedge funds etc looking for an edge.

One thing often missed here is that most orders, even from most hedge funds and prop trading shops, still go via broker systems. Direct Market Access is getting more common but it's often a pain in the arse from a regulatory and disclosures perspective, and means you lose out on short locate (shares that you can borrow from your broker to short sell).

"Sponsored Access", where you connect directly to an exchange but your broker monitors your activity via a drop copy, is a happy middle ground.

Surprisingly though, I've heard of at least one trading venue where going direct is slower, because the venues own risk checks are slower than the the ones implemented by at least one broker, and the broker themselves are allowed to bypass the risk checks put in place at the exchange for general DMA clients. "Direct" is clearly subject to negotiation.

I've also heard of brokers who tried to implement their gateways in FPGA, and have later shuttered the project, having gone back to relatively slow software gateways for the flexibility.

A lot of trading still happens via FIX, which is a slow ASCII protocol. Most prop shops will have aggressively optimized FIX parsers and serialisers out of necessity.

People think all trading happens in these elite, bleeding-edge hardcore sub-microsecond systems, but a lot of it is just dogshit.

Things are a bit more optimised in the derivatives space because of the insane volumes (Options trading just for US equities is easily into the petabytes of storage per year).

Re: How exchanges turn order books into distributed logs

#73
post #50

The article says it's not enough to accurately timestamp orders at the various order entry portals. I didn't understand why that's not enough. GPS can provide fairly accurate timestamps. There's a few other GLONASS systems as well for extra reliability.

GNSS is the generic term, GLONASS is the name for the Russian system.

Thanks for the correction. I just know the Russians and Chinese have their own systems.
Post reply on HN