TCP INCAST PROBLEM
TCP Incast problem attracts much attention due to the
catastrophic goodput drop. In this paper, a goodput model of the problem is
built to understand why goodput collapse occurs and a solution to the problem
based on the theoretical analysis is proposed. We found that the TCP Incast
goodput deterioration is mainly caused by two types of timeouts, one happens at
the tail of data blocks and dominates the goodput when the number of senders is
small, while the other one at the head of data blocks and governs the goodput
when the number of senders is large. The proposed model describes the
relationship between these two types of timeouts and the
Incast communication pattern, block size, bottleneck buffer size, and so on.
The simulation results indicate that the model well characterizes the features
of the TCP Incast problem.
TCP Incast has
risen to be a critical problem recently in data center networks due to its
catastrophic goodput collapse. Incast, a communication pattern, was first termed by Nagle et al. in file storage systems. In the Incast
communication pattern, multiple senders concurrently transmit data blocks to a
single receiver, and any sender cannot send another data block until all the
senders finish transmitting the current ones. As the number of senders increases,
the goodput of the receiver will become lower than the capacity of the
bottleneck link in one or even two orders of magnitudes.
No comments:
Post a Comment