Load Balance using IP Address

Hi Guys,

I have a question with regad to load balancing using Ether Channel.  For example if I have 2 100mb links configured as Ether channel and I am using ip addressess as a load balance method in DC and if the end host generates 150mb traffic, will the traffic just use the first 100 mb link or it will also use the 2nd 100mb link ?

 

Tks.

Comments

  • If it was one flow generating 150 Mbit/s then you would only get 100 Mbit/s throughput because no matter what hashing method you use if you have only one flow it would select one of the two links. If you have more than one flow then the hashing method would be important and depending on that you may use more than 100 Mbit/s but not in one flow of course.

  • Ok I am bit confused here. I know with different ports you can have different flows but how about with ip addressess. If ip address is used then will it be one flow or more ?

  • peetypeety ✭✭✭

    Ok I am bit confused here. I know with different ports you can have different flows but how about with ip addressess. If ip address is used then will it be one flow or more ?


    Insufficient information given.  Are you using source IP address as the hash, destination IP address as the hash, or source XOR destination IP address as the hash?  Once you've answered that question, I suspect the answer to your original question should become clearer. 

  • Source ip address I.e. the machine is talking to different destinations. If you can explain how many flows will be there if using source ip / destination ip. Tks

  • A flow is a unique combination of SRC and DST IP and SRC and DST ports. So you might be downloading a file via a web browser on port 80 (HTTP) from IP x.x.x.x. At the same time you are also downloading a file via port 20 ((FTP) 21 is to open session). Those would be two different flows. In this case the hash does not matter as it would use the same link unless you hash on layer 4 information (port).

    If you are downloading via FTP from two different destinations and hash on DST IP then those two would use different links. If you hash on SRC IP then the same link would always be used. That is why you have to know your traffic patterns when choosing hashing algorithm but generally it's better to hash based on both layer 3 and layer 4 information if possible.

  • Thank you very much Daniel.

     

    Tks everyone.

Sign In or Register to comment.