Hard to imagine? Our grandparents days of networking across systems work reliably over TCP and that is what we have seen throughout. The systems at each end of the network was not just how the TCP connection was established so that TCP basic definition was "a single connection between two hosts." While the researchers designed suite of TCP / IP protocols, they did an awesome job looking through requirements that may come in the next few decades. Given their vision until today, we are able to communicate over TCP.
But what changed between the two? The network devices or the Internet grew at an unexpected pace and broke all predictions. The backbone Internet traffic in 190 was close to 1 terabyte, which grew to nearly 35,000 by 00. What Terabyte exceptional growth and large companies have started to turn on the Internet. TCP was designed to take as much load without getting slower and get to a point where it starts breaking? While all this growth was happening in the background the researchers continued to work on simplifying congestion control problems with TCP and many new RFCs came and were adopted as well. Today, we are all able to work effectively using these congestion control algorithms and complex avoidance.
Beyond the scope of TCP, routing are critical in networking today by finding the shortest multiple paths available between two hosts. Once the top layer defines the properties that must be based on the best path selection, most of the best times is not really one that is picked up. Once the path is taken then all communication is linked to this single track or some equal cost paths for transmitting packets. The result certainly ignoring the possible alternative paths that become available for data transfer between a same set of computers. Even when the routers try to make intelligent decisions on packet distribution for the same TCP stream on multiple paths, there is a packet of concern up to order. This starts the congestion control algorithms and after the TCP threshold is estimated that the packet was lost and retransmission arrives. It also reduces the rate of sending packets halving congestion window. To avoid this problem Routers are trying to ensure that packets belonging to single streams are sent by the same path.
You must be wondering why not even worth the multiple paths? Good question and the answer is with you ... do you bring a smartphone with an Internet connection? Does your phone to connect via WiFi? And it must have Bluetooth and USB ports? Then you have the answer, is not it?
Each unique ability here would allow your smartphone to connect to a remote host that is great news. You can have 2 or 4 parallel channels for communication ... but that takes the data from one end to the other? Same old TCP and TCP session can use one of the available paths. So how will you use other connected streams and on scenarios where moving, disconnect / re-connect to a particular network? Let us understand some realistic image around Mobile. Mobile data usage for 2012 was nearly 12 times for Internet data volume connected with 00. Mobile devices have crossed the number of devices connected by large fold and it is becoming more explosive rate. Therefore, we must revamp how the data transfer has worked in fixed Internet line over how it should work with mobile devices around.
With the understanding of the problem at hand, consider how different data transfers would be if the TCP session could make use of all available paths between two hosts. Enumerate quickly down the benefits of such a scenario:
- Better performance network layer
- Using the best path that is less crowded
- Using the path which has the largest bandwidth for bulk transfer
- much robust App connectivity while the device loses its way
- cause trouble to the least already overcrowded roads and congested
- Finally much better end user experience ... causing Wow factor
Looks great and something that we should go to but how? Should we build a new suite of TCP was not designed to take care of these scenarios? Will it be acceptable, what happened to the adoption of SCTP over the Internet? Many times, a great technology does not get the same acceptance because it is trying to change the foundation which is used by billions of devices.
interesting problem and leave you for a little brainstorming ... stay tuned for the next blog where we take the discussion to the next level ...
0 Komentar