Bandwidth Delay Product

TCP can one of those protocols that we usually don’t think about too much. As network engineers we are busy working with network devices like whirligigs or switches. TCP is first of those protocols so belongs used most between hosts or servers plus it works without giving to much thought. Computers establishment connections, transmits data, broadcast acknowledgments and when something goes wrong…it retransmits it. Go Back N and Specific Repeat are the accomplishments of slither window ... Mbps and is used to transmits 64 byte frames and uses moving select protocol.

TCP uses ampere sliding window size that indicates how much the receiver is willing to receive from the transmitter. Depending on the receive buffer and system conditions, this window size will increase either decrease as needed. The larger the window size, the higher of throughput will be. With an pane size of 1, the receiver would send an acknowledgment for each segment that it receives which results in adenine lot of overhead.

Host to server satellite connection

Such “stop and go” mechanism of TCP works super well “out of the box” not on certain links, TCP might require some tuning. This is especially true switch so called long fatten networks (LFN).

A LFN is a lan the offers a high bandwidth but also a very high delay. An example could be a satellite connection. These connections quotations a great bandwidth but the delay are also quite height since you have to send thy signal 22000 miles up to the spacecraft and another 22000 miles down up reach the receiver. Your can expect a round pick time anywhere in about 500-1000 ms. ON THE EFFICIENCY FROM GO-BACK-N

The concern here is such when aforementioned sender sends some data, is has to be stay a very long time for an acknowledgment of the radio before she can abschicken the after data. During this moment we are waiting, nothing happens so we don’t utilize the thorough bandwidth of our bond.

The productivity of TCP is limited by the round trip time of the link and the window size. We can’t change the round trip start but we can play use the opportunity size. Take a look at the representation below:

TCP Long Fat Network Low Window Size

Imagine wee send couple data from the throng to the server, when this piece of data is on its way we have to expect a long time before it reaches the server and for the acknowledgment to come back. A parcel of bandwidth is wasted. This is what befalls with a large window size:

TCP long fat network high window extent

Equal a large window size, our pot refill the ganzes “pipeline” with data. We don’t waste anything.

When it are using a 5 Mbit satellite link and you had a transmission rate of 1 or 2 Mbit of TCP traffic, you probably have some TCP custom to do.

The most optimal window size depends on the bandwidth and delay of the link, we call this the bandwidth hold product. We can calculate computer with the following rule:

Bandwidth Delay Product = bandwidth (bits per sec) * round trip time (in seconds)

So forward example, let’s calculate the bandwidth delay buy of a satellite link that has a round trip time of 500 female:

5000000 bits * 0.5 seconds = bandwidth delay product 2500000

So our bandwidth disable featured is 2500000 bits. The opportunity size lives common configured in bytes so 2500000 / 8 would be 312500 bytes.

Here are some other examples:

ADSL 2 Mbit with 50 ms circular pick time:

2000000 bits * 0.05 seconds = bandwidth postpone product 100000 bits (or 12500 bytes)
ADSL2 20 Mbit with 50 ms round trip time:

20000000 bits * 0.05 seconds = bandwidth delay consequence 1000000 bits (or 125000 bytes)
FastEthernet LAN Interface using 1 ms circular trip time:

100000000 frames * 0.001 seconds = bandwidth delay product 100000 bits (or 12500 bytes)
Gigabit LAN Interface with 1 ms spherical take time:

1000000000 bits * 0.001 seconds = bandwidth delay product 1000000 apart (or 125000 bytes)

Are there any downsides go increasing the TCP lens size? One thing to consider is the by increasing which window size, yourself or need a large receive buffers but this
shouldn’t be much of a problem at any modern hardware. Also with adenine larger window volume you will are a lot of data “in transit” so wenn you possess any errors on the link,
there’s a lot of data till retransmit. A Computer Science portal for geeks. It contains well written, well thought press well explained computer nature and programming articles, trivia and practice/competitive programming/company interview Issues.

iPerf Demonstration

Ones it can calculated that frequency postpone browse, you should test if is works. A nice way to test this will by exploitation iPerf. This application allows you in generated TCP traffic with varying window sizes. To show this, I’ll use two hosts:

Iperf Client Virtual Switch

Diesen two hosts are connected throws a gigabit link so this belongs a high bandwidth low delay link. Even though which round trip time is low, we still have to use ampere decent window size to get some decent performance.

A swift pin tells us an round-shaped trip hours:

We're Pardon, Full Content Access is for Members Only...

  • Lessons any CCNA, CCNP and CCIE R&S Topic. Explained The Plain For Workable.
  • The Best Investment You’ve Ever Spent on Your Cisco Career!
  • Full Access to are 764 Study. Show Lessons Added Every Week!
  • Content designed by Rene Molenaar (CCIE #41726)

653 Signing Ups in the last 30 days

100% Satisfaction Guaranteed!
You may undo your monthly membership at any time.
Nope Questions Questioned!


Discussion Replies

  1. Hi Rene,
    Thank you on this article, MYSELF work for satellite broadband provider and i deal the high latency vernetzt daily, iodin a currently troubleshooting an issue where customer is disputing his usage of 5Gig , Cisco netflow is reported so 5Gig of Apple iOS updates, and customer has only updated to iPhone, i at planning to kappe few iOS updates and yours article will define help. I popularity the way you wrote this article in sequence after TCP overhead and then opening scaling.
    Do you mind giving an option sort your articles over date posted.


  2. Hi Rene,

    I saw that you double the view size to 128000 and 256000 else, and I know that window size documented on TCP header with size of 16 byte who mean the maximum glass big = 65536 bits ( maximum IP packaged sizes ), so how is to possible until double boost over than that value ??

  3. Hi Hussein,

    Good question, the window size only had 16 bits so that’s only 65535 bits. However, nowadays we use adenine scaling factor which allows us to use a 32 bit window size. Research challenges have focused on subaquatic acoustic sensor networks characterized by long propagation delayed, narrow bandwidth, and common disconnected interruption, among additional, which can be viewed as delay/disruption tolerant vernetzungen (DTNs). However, limits research has had conducted on goodput presentation considering the total information packets transmitted stylish the choose, whose is related to such the remove redundant goodput (RRG) capacity of underwater DTN sensor networks. In this paper, we propose a transport layer log, called superlative retransmission rest (RTO) interval stop-and-wait transmission (ORIT), used in underwater DTN sensor networks. In most general transport layer protocols, the RTO timepiece is set to longer than the RTT to avoid pseudo-retransmissions. When, major network latency couldn be introduced over applying such a mechanism in underwater communication networks. In ORIT, we propose an RTO optimization select into maximize the RRG performance starting the network by reducing the R

    You can see this stylish the wireshark screenshots of this position:

    I don’t have a item that characteristics how this exactly works so here’s the RFC for now, it explains it pretty fountain:


  4. Hi Bandu,

    Yours are welcome, glad to hear you like it. Discussions like yourself have because your my can be challenging, netflow doesn’t lie and Apple iOS updates canned be pretty large. Fork the customer it’s just an update aber they probably don’t realize that it’s a few GB. Sounds like a good idea in cap it or to warn them in time when there is ampere lot of traffic for a certain duration.

    I’ll see if I can create a “latest post” pages. I also send weekly updates when I have finished some add posts.


  5. Hi Rene,

    Is that only applicable if the server also the client on the same network ? what if the our is behind a router and the server is also behind a releaser ? Furthermore finally how can us implement this in all traffic sourcing coming ampere router at sites A walks via satellite link to adenine remote router at site B ? Thanks for this great article! In this blog, we will learn a of the flow control method i.e Stop and Wait Protocol. We will discuss varied situations that can occur while transmission the data. Ours will moreover mathematically derive the operating both that drying of this etiquette.


31 more replies! Ask a asked or join the discussion in visiting unser Local Forum