Application Layer’s Approaches for TCP Incast Problem at Data Center Networks?

Data Centers have become very popular to store a huge amount /volumes of data. Many companies like Amazon, Google, Microsoft, IBM, Yahoo, and Face Book use the data center for storage of Web Search, E-Commerce, and large scale computation. High Speed Links, Low Propagation Delay, Limited sizes Switch Buffer are the main characteristic of Data Center. Data Center in these days have hundreds of thousands of servers to store the data across many thousands of machines. TCP is the most popular transport layer protocol that is currently used in the internet. Data Center faces the different set of problems than internet. TCP Incast is the main problem that the data centers faces. TCP incast problem is refer to the TCP throughput collapse. When the multiple data sender s simultaneously respond to a single receiver, that is called many to one communication pattern, the burst data overload the buffer of receiver’s switch. This causes the throughput collapse that degrades the performance and packets loss. That is so-called TCP incast problem. Many techniques, approaches and algorithms have been introduced to resolve the throughput collapse problem of TCP. Due to the switching to cloud computing culture or cluster base switching system those are data centers, the researchers join the race of research to avoid the throughput collapse issue. Multi-layer approaches & techniques are proposed for incast problem. Application layer is the most important layer in the TCP/IP structure. This layer has the capabilities to handle or control to congestion. In this research work techniques for incast problem at Application Layer level is discussed. In this research work attempt is made to discuss available possible solution for researchers who want to work on the incast problem at application layer.