Hi Guys...
I would not recommend to do per-packet load balancing. Per-packet load
balancing will give a near perfect split of traffic across the available
paths, but has the disadvantage of causing packets in the same session
to following different routes through the network. This can cause
packet reordering and non-predictive latency for that session. All
packets in a session have the same source and destination IP addresses.
Per-destination load balancing addresses the disadvantage of per-packet
(packet reordering and latency), while attempting to provide the same
accurage split of traffic across the available paths. Per-destination
is perhaps a misnomer; it should be called per-session load balancing
because CEF algorithm is using a hash based on the destination AND the
source IP addresses...so not only on the destination. This has the
advantage of keeping packets in a session in the same order and
following the same route across the network, which helps with
maintaining the latency for each packet. And to be even more precise,
the CEF algorithm is using a third value called routerID. This value is
used to eliminate the traffic polarisation.
Now as far as the 60/40% packet split mentionned earlier... The
per-packet is "near perfect split" because the CEF algorithm is in
reality using per-packet load balancing for a given destination prefix.
It is not a pure packet-by-packet round robin. CEF keeps a record
for every destination prefix in order to know which path is next in the
round robin. Also which command are you using to get 60/40 ? You
should use CEF command related to see the split. Ex.: _show ip cef
internal_ You should see the packet split count for every
bucket.
Hope this help !
Frank S.
[EMAIL PROTECTED] wrote:
> I am doing this exact thing right now for a customer and it works like a
> charm here is a moc config for you without ip addresses of course.
> See the config below and I made everything bold in reference to what
> everyone is speaking about with cef. One thing I have noticed it is
> supposed to be per packet but I am actually getting about 60/40% load
> balancing over the links according to theinterface counters and the
packets
> going over each t1. That is one question that I have do you truly get
> 50/50 load balancing or is everyone else seeing this type of result with
> load balancing also. I am using ospf on both routers.
>
> mocrouter#sh conf
> Using 2582 out of 29688 bytes
> !
> version 12.1
> service timestamps debug uptime
> service timestamps log uptime
> service password-encryption
> !
> hostname mocrouter
> !
> no logging console
> enable secret 5 XXXXXXXXXXXXXXXXXXXXXXXXX
> !
> !
> !
> !
> !
> clock timezone EDT -5
> clock summer-time EDT recurring
> ip subnet-zero
> ip cef
> no ip finger
> ip domain-name XXXXXXX
> ip name-server XXXXXXXX
> ip name-server XXXXXXXX
> !
> !
> !
> !
> !
> !
> interface Loopback0
> description loopback for mocrouter
> ip address XXXXXXXXXXXXXXXXXXXX
> !
> interface FastEthernet0/0
> description mocrouter
> ip address XXXXXXXXXXXXXXXXXXXXX
> ip broadcast-address XXXXXXXXXXXX
> speed 100
> full-duplex
> !
> interface Serial0/0
> description Point to Point T1 to other end mocrouter
> ip address XXXXXXXXXXXXXXXXXX
> ip load-sharing per-packet
> no ip mroute-cache
> no fair-queue
> !
> interface FastEthernet0/1
> no ip address
> shutdown
> duplex auto
> speed auto
> !
> interface Serial0/1
> description Point to Point T1 to Other Moc Router
> ip address XXXXXXXXXXXXXXXXXX
> ip load-sharing per-packet
> no ip mroute-cache
> no fair-queue
> !
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=34166&t=33599
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
