CCNP_LOAD_BALANCING......balance me now!!

rameeshmeshadrameeshmeshad Member Posts: 15 ■□□□□□□□□□
...........messed up with!!!!

if eigrp using the equal cost load balancing to reach destination, one packet is sent and then the next packet through other link...

1) is both the packets are sending at the same time?

if yes...how should the destination router acts(i mean both the packets are heading same time...is it possible)

if no .......then how can the router save the time ?( it can efficiently tune the load using two links....but the time is same....na? )
.....with luv rameeee

Comments

  • networker050184networker050184 Mod Posts: 11,962 Mod
    The load balancing method used isn't determined by the routing protocol. The routing protocol just gives multiple paths that can be used by the switching method for load balancing. With CEF the default load balancing mode is per-destination. So if both of those packets were for the same destination they would use the same link.
    An expert is a man who has made all the mistakes which can be made.
  • kryollakryolla Member Posts: 785
    load balancing via equal or unequal cost links involve 2 things

    1 is the routing protocol having the same metric to the same destination which puts the links in the routing table

    2 is the switching method either CEF or process. Process is per packet so 1 packet is sent across 1 link and the other is sent across the other link. CEF is per source/destination or per packet

    Load balancing isn't about saving time its using multiple links efficiently what saves time is CEF instead of process switching.

    Rack1R6#sh ip cef 155.1.23.0
    155.1.23.0/24, version 155, epoch 0, per-destination sharing
    0 packets, 0 bytes
    via 155.1.146.4, FastEthernet0/0.146, 0 dependencies
    traffic share 1
    next hop 155.1.146.4, FastEthernet0/0.146
    valid adjacency
    via 155.1.146.1, FastEthernet0/0.146, 0 dependencies
    traffic share 1
    next hop 155.1.146.1, FastEthernet0/0.146
    valid adjacency
    0 packets, 0 bytes switched through the prefix
    tmstats: external 0 packets, 0 bytes
    internal 0 packets, 0 bytes

    Rack1R6#sh ip route 155.1.23.0
    Routing entry for 155.1.23.0/24
    Known via "ospf 1", distance 110, metric 129, type intra area
    Redistributing via eigrp 100
    Advertised by eigrp 100 metric 1 1 1 1 1
    Last update from 155.1.146.4 on FastEthernet0/0.146, 1d12h ago
    Routing Descriptor Blocks:
    155.1.146.4, from 150.1.2.2, 1d12h ago, via FastEthernet0/0.146
    Route metric is 129, traffic share count is 1
    * 155.1.146.1, from 150.1.2.2, 1d12h ago, via FastEthernet0/0.146
    Route metric is 129, traffic share count is 1

    Rack1R6(config-subif)#ip load-sharing ?
    per-destination Deterministic distribution
    per-packet Random distribution

    Rack1R6# sh ip cef 155.1.23.0
    155.1.23.0/24, version 155, epoch 0, per-packet sharing
    0 packets, 0 bytes
    via 155.1.146.4, FastEthernet0/0.146, 0 dependencies
    traffic share 1, current path
    next hop 155.1.146.4, FastEthernet0/0.146
    valid adjacency
    via 155.1.146.1, FastEthernet0/0.146, 0 dependencies
    traffic share 1
    next hop 155.1.146.1, FastEthernet0/0.146
    valid adjacency
    0 packets, 0 bytes switched through the prefix
    tmstats: external 0 packets, 0 bytes
    internal 0 packets, 0 bytes
    Studying for CCIE and drinking Home Brew
  • adam-badam-b Member Posts: 36 ■■□□□□□□□□
    kryolla wrote: »
    load balancing via equal or unequal cost links involve 2 things

    1 is the routing protocol having the same metric to the same destination which puts the links in the routing table

    2 is the switching method either CEF or process. Process is per packet so 1 packet is sent across 1 link and the other is sent across the other link. CEF is per source/destination or per packet

    Load balancing isn't about saving time its using multiple links efficiently what saves time is CEF instead of process switching.

    Rack1R6#sh ip cef 155.1.23.0
    155.1.23.0/24, version 155, epoch 0, per-destination sharing
    0 packets, 0 bytes
    via 155.1.146.4, FastEthernet0/0.146, 0 dependencies
    traffic share 1
    next hop 155.1.146.4, FastEthernet0/0.146
    valid adjacency
    via 155.1.146.1, FastEthernet0/0.146, 0 dependencies
    traffic share 1
    next hop 155.1.146.1, FastEthernet0/0.146
    valid adjacency
    0 packets, 0 bytes switched through the prefix
    tmstats: external 0 packets, 0 bytes
    internal 0 packets, 0 bytes

    Rack1R6#sh ip route 155.1.23.0
    Routing entry for 155.1.23.0/24
    Known via "ospf 1", distance 110, metric 129, type intra area
    Redistributing via eigrp 100
    Advertised by eigrp 100 metric 1 1 1 1 1
    Last update from 155.1.146.4 on FastEthernet0/0.146, 1d12h ago
    Routing Descriptor Blocks:
    155.1.146.4, from 150.1.2.2, 1d12h ago, via FastEthernet0/0.146
    Route metric is 129, traffic share count is 1
    * 155.1.146.1, from 150.1.2.2, 1d12h ago, via FastEthernet0/0.146
    Route metric is 129, traffic share count is 1

    Rack1R6(config-subif)#ip load-sharing ?
    per-destination Deterministic distribution
    per-packet Random distribution

    Rack1R6# sh ip cef 155.1.23.0
    155.1.23.0/24, version 155, epoch 0, per-packet sharing
    0 packets, 0 bytes
    via 155.1.146.4, FastEthernet0/0.146, 0 dependencies
    traffic share 1, current path
    next hop 155.1.146.4, FastEthernet0/0.146
    valid adjacency
    via 155.1.146.1, FastEthernet0/0.146, 0 dependencies
    traffic share 1
    next hop 155.1.146.1, FastEthernet0/0.146
    valid adjacency
    0 packets, 0 bytes switched through the prefix
    tmstats: external 0 packets, 0 bytes
    internal 0 packets, 0 bytes


    Many good points here and I do respectfully disagree with some of them. There are routing protocols that will inject multiple routes to the same destination even if the metrics vary; ie. EIGRP unequal cost load balancing (the variance command)

    Also, i think that load balancing CAN be about time, in that you are effectively increasing the bandwidth to a load balanced destination when doing so; so therefore when the links are maxed out you are moving more data quicker than if they were not load balanced. Correct me if im wrong on any of the above points. Thanks!
  • kryollakryolla Member Posts: 785
    adam-b wrote: »
    Many good points here and I do respectfully disagree with some of them. There are routing protocols that will inject multiple routes to the same destination even if the metrics vary; ie. EIGRP unequal cost load balancing (the variance command)

    Also, i think that load balancing CAN be about time, in that you are effectively increasing the bandwidth to a load balanced destination when doing so; so therefore when the links are maxed out you are moving more data quicker than if they were not load balanced. Correct me if im wrong on any of the above points. Thanks!

    I think I said this

    load balancing via equal or unequal cost links involve 2 things


    and sure if you are experiencing congestion but that delay is not processing delay which is experience during process switching but network delay. So I guess it depends on what type of delay.
    Studying for CCIE and drinking Home Brew
  • adam-badam-b Member Posts: 36 ■■□□□□□□□□
    kryolla wrote: »
    I think I said this

    load balancing via equal or unequal cost links involve 2 things


    and sure if you are experiencing congestion but that delay is not processing delay which is experience during process switching but network delay. So I guess it depends on what type of delay.

    I guess i was misunderstanding your point number one. Thanks for clarifying this for me.
  • kryollakryolla Member Posts: 785
    point number 1 was made to explain that it is the routing protocol that puts multiple links in the routing table for that destination, I should of put unequal metric as well, my bad. BGP also does unequal cost load balancing :)
    Studying for CCIE and drinking Home Brew
  • e24ohme24ohm Member Posts: 151
    kryolla wrote: »
    point number 1 was made to explain that it is the routing protocol that puts multiple links in the routing table for that destination, I should of put unequal metric as well, my bad. BGP also does unequal cost load balancing :)
    Not to jack your thread, but when implementing BGP in a production environment, what is it called when the ISP(s) bond a single IP address to your trunks so load-balancing takes place both ways, inbound and outbound? I am looking at doing this, but I forget the steps, what the vocabulary is called.

    thank you.
    Utini!
  • networker050184networker050184 Mod Posts: 11,962 Mod
    e24ohm wrote: »
    Not to jack your thread, but when implementing BGP in a production environment, what is it called when the ISP(s) bond a single IP address to your trunks so load-balancing takes place both ways, inbound and outbound? I am looking at doing this, but I forget the steps, what the vocabulary is called.

    thank you.

    What kind of trunks are you talking about? Multilple T-1s? If so you can use a multilink group.
    An expert is a man who has made all the mistakes which can be made.
Sign In or Register to comment.