Dynamic Right Sizing
Over the past decade, TCP has become the ubiquitous transport
protocol for the Internet. However, stock TCP performs abysmally
over high-bandwidth or high-delay links (or more generally, networks
with high bandwidth-delay products). As a result, the performance of
application infrastructures such as computational grids and
high-volume web servers, which are built on TCP, is crippled.
To address this problem, grid and network researchers continue to
manually optimize buffer sizes to keep the network pipe full, and
thus achieve acceptable performance. Although such tuning can
increase delivered throughput by an order of magnitude, it requires
administrator privileges and hence the changes cannot be made by the
end user. Instead, system administrators at the source and
destination hosts must separately configure their systems to use
large buffers; this process is tedious and time consuming.
Furthermore, this manual tuning only works well for the pair of
hosts that are manually tuned. If a TCP connection is split, as is
the case with some kinds of firewalls, the firewall administrators
must also get involved.
Dynamic Right Sizing is a straightforward modification to TCP that
automatically and transparently addresses the above problems while
maintaining connection semantics and the ubiquitously deployed
features of TCP.
