Data Dissemination part 2 : TBT VS MTBT data dissemination
This is part 2 of a 5 part series on data dissemination system of NSE during TCP TBT. In this series I’ll talk about how no technology is perfect and the problems claimed in TCP architecture remain even in MTBT. In part 2, I talk about the changes introduced with MTBT. Read Part 1 here.
One of the main themes in the NSE colocation controversy is the choice of TCP-IP protocol contributing to preferential access. It was showcased that the advent of Tick-by-Tick on multicast (MTBT) magically eliminated preferential access, which was supposed to be available when data was disseminated using TCP-IP. Such narratives appear to be misleading and correlation is often passed on as causation.
The NSE colocation HFT “scam” is like accusing Ford for producing Model T in 1908 which caused very high pollution when today’s cars are electric and don’t cause any pollution.
TCP VS MULTICAST
Data can be disseminated in an exchange environment in both TCP and multicast formats. TCP is a reliable and secure way of data transmission. Multicast is said to be unreliable but a faster way for data transmission. Tick by tick data was sent by the exchange in TCP format between 2010 and 2016. Multicast TBT data transfer was started in 2014.
In a single server multiple receiver system, TCP network data is sent from server to the clients sequentially. Each client establishes an Independent connection with the server. The data is then sent to each client one after the other by the server.
In a single server multiple client multicast network, data is sent from the server to a switch. Each client establishes an independent connection with the switch. The data is sent to each client one after the other by the switch.
Since hardware does the sequential data processing it appears to be simultaneous, in multicast. But in reality there are still differences between different clients for data dissemination.
TBT VS MTBT
One of the myths which have been propagated is the identical nature of ticks between TBT and MTBT. Apart from the protocol, TBT and MTBT were considerably different. The fundamental nature of tick difference between TBT and MTBT are listed below:
- TBT was disseminated on 1G network while MTBT was disseminated primarily on 10G network. Comparing data dissemination between the two is like comparing a single crowded road with a 10 lane expressway.
- TBT data compressed multiple packets into 1 LZO batch. MTBT data had no concept of batching. MTBT data was always sent uncompressed. This was like comparing sharpened pencil to an unsharpened one.
- NSE began sending out active tick information only in MTBT. TBT did not have any active tick information. This caused TBT data to be different in sequence from the MTBT data.
Thus MTBT differed from TBT in protocol, bandwidth, content and packaging. Any correlation of performance to just protocol is akin to saying the person standing closest to the gun in 100 meter race will win.
Was there is really a “scam” in NSE colocation controversy. The technology is an evolution. One can’t have a Bugatti supercar without having a Benz Motorwagen.
Next, I’ll write about role of network in tick data transmission.
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.
