Data Dissemination part 3 : Role of the network
This is part 3 of a 5 part series on data dissemination system of NSE during TCP TBT. In this part I’ll talk about network through which the tick passes before reaching the member. Data sent is NOT data received. TCP was invented just to solve this problem. In Part 3, I talk about delays and variability in the tick traversal. Read Part 1 and Part 2 here.
Imagine you are going from Delhi to Mumbai on a flight. One person is seated on the first row, another person is seated on fifteenth row. Both of them want to reach their homes which are equidistant, say about 40 KMS from the airport. Would it be even logical to think that person sitting on the 1st row would have an unfair advantage and reach home earlier? It would be foolish to discount the variability caused by traffic, red lights, tolls and other stoppages.
Just like city traffic, NSE’s network on which TBT data was disseminated was highly congested. TCP packets had to pass through tolls (firewall), red lights (router), traffic congestion and narrow roads (1G network) and speed bumps (switches). TCP packets ultimately are like vehicles travelling through traffic where sometimes they have to wait in a queue while at other times they are lucky to arrive just when the light has turned green. Just like a red light with multiple lanes, waiting time for a TCP packet in a queue depends not only on the arrival time but also on the speed of the packets ahead of the packet in the queue.
Red lights have different duration of stoppage time based on traffic during the day. Similarly network components take variable time to process a packet. This variability in network components is unavoidable and affects every packet. It makes one packet pass through faster while it delays the other packet. This variability is also known as jitter and is present in all software and hardware components. Every packet is affected by jitter and the impact is different on every packet.
In real time traffic terms, NSE network would equate to peak hour Mumbai traffic congestion, distances of over 100 KMS and scores of traffic stoppages. In such a system, where a car would be on the road for hours altogether, would it be possible to ensure that one car reaches the destination always, or even mostly, ahead of others? Only if a person starts earlier by an hour, maybe 30 minutes, or maybe even a few minutes. Definitely not a millisecond earlier. For two cars which start about 5 minutes apart to cover a distance of 50 km, there is absolutely no guarantee that the car which started first which reach the destination first. Similarly, in a congested traffic network like that of NSE, if a data packet starts a few milliseconds before another one, there is no certainty that it will reach its destination before the second packet. When it reaches, it could be at number 2 or 3 or 4 or even 10.
In NSE colocation, the packet which started from NSE servers underwent significant network congestion before reaching member. This network was an effective load balancer. Therefore, shouldn’t we agree that there was absolutely no advantage which can be derived out of such a congested network.
Next I’ll write about the role of the operating system.
All articles are here. The author advises market participants in legal matters related to securities markets and has advised some noticees in this matter also.
