Boards.ie

techguy · 25-07-2012, 01:13

Hi,

I'm putting this in here as it is a little more technical than most of the threads in the Nets/Comms forum.

Anyway, here is a question on an assignment i'm working on:

Quote:

What effect does packet size have on total transmission time for a given data size? use a diagram to show such an effect when transmitting same quantity of data using different packet sizes.

I am a little confused because to me there is no straight answer for this. It really depends on a lot of routing factors.

The question isn't worth many marks so i'm not really sure how to go about answering it. What do you think?

Thanks.

srsly78 · 25-07-2012, 08:45

http://en.wikipedia.org/wiki/Maximum_transmission_unit

If packet goes over the MTU it will get split up, once that happens there are more packets so more things that can go wrong and causes a resend etc (depending on whether tcp or ucp etc).

Would expect graph to move in steps, each step corresponding to a multiple of the MTU. Would expect insignificant difference for 2 packets that both fit into MTU. Example: would expect 1000 byte packet to take same time as 500 byte packet (assuming 1500byte mtu).

jmcc · 25-07-2012, 12:51

Overall data (I) would be a constant (error free connection). Total amount of error control/framing/signalling etc (C) variable depending on the number of packets. Thus smaller packets would have more overhead so that theoretically more ec/framing/control data would be sent with smaller packets increasing overall transmission time. The things to consider are the section of the packet available for data and that taken by the ec/framing etc. The largest packet would be where the full set of data could be transmitted.

Each packet could be represented as (D + C). I would be the full set of data and D would be space for data in each packet. So the number of packets, crudely, would be I/D but the sum of C would increase according to the number of packets required. Not sure if this is the academic solution or exactly what you or your lecturer wants. That might take T as being the transmission time and you would have to show how the C overhead impacts on T.

I definitely need more coffee as this is giving me flashbacks to dialup days. And there's no data size specified in the question above.

Regards...jmcc

**stevenmu** · 27-07-2012, 15:26

It's something that you could easily write a paper on, and include lots of tests with graphs and so on.

But for a short answer I'd just mention how larger packets will make for less packets, which will reduce the delays between packets caused by latency, but going over the MTU will increase packets. And possibly also mention that because the internet is a heterogeneous network, affects could vary from route to route.

edit: oh, and also mention that the protocol would have an impact too. Latency and the number of packets would have a greater impact on a controlled protocol like TCP, but less of an impact on an uncontrolled protocol like UDP.

techguy · 28-07-2012, 03:56

Thanks guys.

I think (hope) all your answers are a little too detailed for my question as I haven't really come across anything like that in the course notes.

I just went with the fact that there would be more data if the packets were smaller as there would be header data for each packet. Does that sound like something that might get me a few marks?

Thanks.

25-07-2012, 08:45	#2
srsly78 Registered User Join Date: Oct 2010 Posts: 3,758 Adverts \| Friends	http://en.wikipedia.org/wiki/Maximum_transmission_unit If packet goes over the MTU it will get split up, once that happens there are more packets so more things that can go wrong and causes a resend etc (depending on whether tcp or ucp etc). Would expect graph to move in steps, each step corresponding to a multiple of the MTU. Would expect insignificant difference for 2 packets that both fit into MTU. Example: would expect 1000 byte packet to take same time as 500 byte packet (assuming 1500byte mtu). Last edited by srsly78; 25-07-2012 at 08:47.

25-07-2012, 12:51	#3
jmcc Hackwatch Join Date: May 1999 Location: Waterford,Ireland Posts: 2,737 Adverts \| Friends	Overall data (I) would be a constant (error free connection). Total amount of error control/framing/signalling etc (C) variable depending on the number of packets. Thus smaller packets would have more overhead so that theoretically more ec/framing/control data would be sent with smaller packets increasing overall transmission time. The things to consider are the section of the packet available for data and that taken by the ec/framing etc. The largest packet would be where the full set of data could be transmitted. Each packet could be represented as (D + C). I would be the full set of data and D would be space for data in each packet. So the number of packets, crudely, would be I/D but the sum of C would increase according to the number of packets required. Not sure if this is the academic solution or exactly what you or your lecturer wants. That might take T as being the transmission time and you would have to show how the C overhead impacts on T. I definitely need more coffee as this is giving me flashbacks to dialup days. And there's no data size specified in the question above. Regards...jmcc Last edited by jmcc; 25-07-2012 at 12:54.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

27-07-2012, 15:26	#4
stevenmu Moderator Join Date: Jul 2003 Posts: 8,474 Adverts \| Friends Mod: Spirituality	It's something that you could easily write a paper on, and include lots of tests with graphs and so on. But for a short answer I'd just mention how larger packets will make for less packets, which will reduce the delays between packets caused by latency, but going over the MTU will increase packets. And possibly also mention that because the internet is a heterogeneous network, affects could vary from route to route. edit: oh, and also mention that the protocol would have an impact too. Latency and the number of packets would have a greater impact on a controlled protocol like TCP, but less of an impact on an uncontrolled protocol like UDP.

28-07-2012, 03:56	#5
techguy Registered User Join Date: Jun 2004 Location: Dublin Posts: 1,928 Adverts \| Friends	Thanks guys. I think (hope) all your answers are a little too detailed for my question as I haven't really come across anything like that in the course notes. I just went with the fact that there would be more data if the packets were smaller as there would be header data for each packet. Does that sound like something that might get me a few marks? Thanks.