Skip to content

Download Analysis of TCP Performance in Data Center Networks by Santosh Kulkarni, Prathima Agrawal PDF

By Santosh Kulkarni, Prathima Agrawal

This booklet addresses the necessity to enhance TCP’s functionality within info facilities through delivering recommendations which are either useful and backward suitable with regular TCP types. The authors procedure this problem first by way of deriving an analytical version for TCP’s functionality lower than ordinary information heart workload site visitors. They then speak about a few ideas which are designed to enhance TCP functionality by means of both proactively detecting community congestion via probabilistic retransmission or through heading off timeout penalty via dynamic resizing of TCP segments. Experimental effects exhibit that every of thoughts mentioned outperforms common TCP inside of an information middle.

Show description

Read or Download Analysis of TCP Performance in Data Center Networks (SpringerBriefs in Electrical and Computer Engineering) PDF

Best networking books

Implementing 802.11, 802.16 and 802.20 wireless networks: planning, troubleshooting, and maintenance

This isn't one other ebook approximately fitting a house or pastime wireless approach. in its place, this publication indicates you the way to devise, layout, set up, and function WLAN platforms in companies, associations, and public settings equivalent to libraries and resorts. In different phrases, this ebook is jam-packed with severe info for critical execs accountable for imposing powerful, excessive functionality WLANs masking parts as small as a espresso store or as huge as whole groups.

Metaheuristic Procedures for Training Neutral Networks

Man made neural networks (ANNs) provide a normal framework for representing non-linear mappings from a number of enter variables to numerous output variables, and so they could be regarded as an extension of the various traditional mapping ideas. as well as many issues on their organic foundations and their quite broad spectrum of purposes, developing applicable ANNs should be obvious as a very tough challenge.

NETWORKING 2011: 10th International IFIP TC 6 Networking Conference, Valencia, Spain, May 9-13, 2011, Proceedings, Part II

The two-volume set LNCS 6640 and 6641 constitutes the refereed court cases of the tenth foreign IFIP TC 6 Networking convention held in Valencia, Spain, in could 2011. The sixty four revised complete papers awarded have been rigorously reviewed and chosen from a complete of 294 submissions. The papers function leading edge study within the components of functions and companies, subsequent iteration web, instant and sensor networks, and community technological know-how.

Extra resources for Analysis of TCP Performance in Data Center Networks (SpringerBriefs in Electrical and Computer Engineering)

Sample text

Paxson, E. Blanton, TCP congestion control, RFC 5681 (Draft standard), Internet engineering task force, Sept 2009. [Online]. txt Chapter 3 Modeling Incast and its Empirical Validation Transmission Control Protocol is the transport layer workhorse for several application layer protocols like HTTP [1], FTP [2], SMTP [3], NNTP [4] and SSH [5–10]. As a result, TCP carries a significant amount of today’s Internet traffic [11]. Studies have shown that traffic from TCP and UDP [12] make for more than 96 % of the packets in the Internet.

In the example above, slow start phase ends when TCP experiences a timeout. Following the timeout, TCP sets its ssthresh, to half the number of segments that were in flight before the timeout. The protocol also sets the size of its cwnd to one. Since cwnd is now less than ssthresh, TCP resumes its data transfer with slow start. Like before, cwnd grows exponentially as long as cwnd ≤ ssthresh. When cwnd > ssthresh, TCP’s slow start phase ends. TCP then continues with its data transfer using the congestion avoidance algorithm.

The sender waits for a period of time defined by TCP’s timeout before retransmitting its unacknowledged packets. Following a timeout, the congestion window is reduced to one, and only one packet is resent in the first round after the timeout. However, because of the synchronized nature of the Incast traffic, the receiver cannot issue its next request until all the senders have finished transmitting their current blocks. Investigating the causes behind the aforementioned category of timeouts is beneficial in developing effective solutions that are capable of avoiding the ill effects of TCP Incast.

Download PDF sample

Rated 4.25 of 5 – based on 35 votes