Data Dissemination Part 1: Was Anybody First to Log-in Consistently?
This is part 1 of a 5 part series on data dissemination system of NSE during TCP TBT. Some of this article remains relevant for MTBT architecture also. In this series I’ll talk about what happens after the tick gets generated, how does the tick reach the member? In part 1, I talk about the 3 layer TBT architecture and how this architecture ensured there was no possibility of absolute first rank and why the very concept of rank appears to have no basis.
NSE’s TCP TBT tick data dissemination had a three layer architecture.
Layer 1: Primary Data Server (PDC) received information from matching engine and converted them into ticks. These ticks were disseminated to layer 2.
Layer 2: There were five POP servers which received ticks from layer 1. Each POP server independently sent this tick to layer 3.
Layer 3: There were three sender ports for each POP server. Each sender port independently sent ticks to clients (Brokers) connected to it.
Data from one layer to another was sent sequentially i.e. one after the other. Between layer1 and layer2 the order of data dissemination was fixed for a given day. But this order changed day after day. This meant no single POP server in layer 2 was given any preference.
Between layer2 and layer3, the order of data dissemination was not only fixed for a given day but was fixed throughout. Data was first disseminated to port 10990, then to port 10991 and finally to port 10992. Thus port 10990 might have had preference in data dissemination.
Between layer3 and client, data dissemination was based on first come-first serve basis. Based on the above architecture, few observations can be drawn.
Clients connected to port 10991 and 10992 couldn’t have received any preferential access
Since, port 10990 was disseminated data ahead of port 10991 and 10992 every day, it was impossible for clients connected to ports 10991 and 10992 to have been beneficiary of any alleged preferential access.
There was no fixed order of data dissemination across days
Since the data dissemination between layer1 and layer2 was different on different days and there wasn’t any preference given to any server, no single IP connection could have been disseminated data first. On one day port 10990 of one POP server might have been disseminated data first and on the very next day, it might have been disseminated data last. So any preference given to clients connected to port 10990 would have only lasted for a single day.
In NSE colocation “scam”, there is no concept of first-login. It was also impossible to manipulate the NSE system to gain any advantage. There are no such findings despite investigation by multiple agencies.
Many clients were disseminated data simultaneously
On a given day there is no way to determine the order of data dissemination between many clients. For example, let’s say for a given day POP 1 was disseminated data ahead of POP 2. Was port 10991 on POP 1, disseminated data ahead of port 10990 on POP 2? There is no way to know because second position on POP 1 and first position on POP 2 are theoretically simultaneous. Extending the same reasoning to fifteen server port combinations, several clients were disseminated data simultaneously.
There was no relation between order of data dissemination and log-in time for a POP server
Since, port 10990 was disseminated data ahead of port 10991 and 10992, a client connected to port 10990 at any point in time would have preference. For example, let’s say between 7:05 A.M. and 7:10 A.M., ten clients connected to port 10991 of a POP server. At 7:15 A.M. a client connected to port 10990 on the same POP server. Based on a pure log-in time wise ranking, client connected to port 10990 would be ranked 11th for data dissemination. This would be inherently wrong as the client should have been ranked 1st based on the NSE data dissemination architecture. Assuming that the client first to connect in time would be disseminated data first would thus be absolutely wrong. In fact, client who was first to connect to port 10990 could have been the only first to be disseminated data.
Was it possible for any Broker to crowd out the NSE TBT data dissemination system
For a member to crowd out the NSE TBT system he should not only have several allocations to each of the fifteen server-port combinations but also should be the only one to establish connections ahead of every other member. If any other member established even one connection during this time he would have been disseminated data simultaneously.
Next, I’ll write about the TBT vs MTBT protocols and data.
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.