+ Reply to Thread
Results 1 to 2 of 2
  1. Senior Member
    Join Date
    Aug 2005
    Location
    Scotland
    Posts
    213

    Certifications
    MCP (70-270), (70-290), (70-291), (70-294), (70-293) WIP: Exchange
    #1

    Default Ethernet Utilization

    Hi, im taking the 70-293 course and am struggling to understand what the recommended ethernet utilization guidlines means!!I seem like i almost understand it but not quite I.e. below is the guidlines for Shared 10Mbps Ethernet

    Wire Speed - 10 Mbps
    Average Utilization limit - 30%
    Raw Data Throughput limit - 3 Mbps
    Peak Utilization limit - 80%
    Peak Data Throughput limit - 8 Mbps

    Wire speed i obviously understand, that is the max capacity that the wire can theoretically handle.
    However i dont understand the other ones. I know utilization is the percentage of time that the wire is occupied with both succesful and unsuccesful frame collisions so how does 30% average utilization limit relate to the raw data throughput limit of 3Mbps?
    Can anyone put me out my misery??
    Reply With Quote Quote  

  2. Senior Member
    Join Date
    Jun 2006
    Location
    Tampa Bay
    Posts
    1,266

    Certifications
    MCSA 2000, MCSE 2003, Exchange 2000, CCNA, CCNA Security, CNE, A+, Network+, Security+
    #2
    When dealing with CSMA-CD, you have to factor in that the more people trying to talk, the more collisions you get, which have to be backed off. I remember dealing with this in college 20 years ago (wow, was it really that long ago?) and we were shown the statistical math on it, but I don't recall it now. I do remember though that because of the higher probablility of collisions the closer you get to the wire utilization, you get somewhere between 30-60% useful utilization.

    Then, for the frames that successfully get to where they are going, you have to consider the overhead in the frame compared to the real data.

    This is why switches help utilization so much. With switches, you get a virtual circuit where it appears that only two machines are connected. Because of this, collisions are pretty much eliminated, especially in full duplex so the full wire speed can be used to send frames.

    Think of it this way. Ethernet was designed like a telephone party line. A person would pick up the phone, listen, and if the line was clear (dial tone) would start placing a call. In rotary phone days, you couldn't dial a number if another extension was off hook since dialing really was hanging up and picking up really fast. So you start dialing and someone picks up while you dial. Collision.
    If the party line is hardly used, there isn't a problem. But if everyone on the party line wants to talk, you have lots of collisions and not many conversations get through. Telephone switches eliminated this problem.
    I realize this example is not the best since Ethernet works by short autonomous frames where a telephone conversation could hog a line for hours.

    Highway traffic could be another example of lower effective utilization due to high demand.

    I doubt though that this would be tested to this degree. Just understand that 100 machines connected by a hub sending 100KB of data files (10MB total) will take quite a bit longer than a second over a 100Mb line (average throughput) versus if only one machine was to send 10MB file (peak throughput).
    Reply With Quote Quote  

+ Reply to Thread

Social Networking & Bookmarks