3 ms·
No, that had absolutely nothing to do with it. It was based on a bad assumption about volume of trade entries/cancellations during the initial price calculation
by hueving 10y ago
No, that had absolutely nothing to do with it. It was based on a bad assumption about volume of trade entries/cancellations during the initial price calculation.
- xapata 10y agoSounds like you know more. Care to explain what the assumption was and how it caused the malfunction?
- hueving 10y agohttp://www.system-logic.com/commentary/2013/06/04/NASDAQ-and-the-Facebook-IPO-part-1-The-Race-Condition.html http://www.system-logic.com/commentary/2013/06/04/NASDAQ-and...
- xapata 10y agoThat doesn't contradict my conjecture that the issue was improper sharding. The jump in duration of cancellation detection from 2ms to 20ms could have been because they were running that calculation on a single machine. Although... Ethernet latency would probably make it tough to stick to 2ms.