This question appears in Mumbai University > Parallel And Distributed System

Load sharing:

• Load sharing means one can split the traffic from a network to be transported by different routers (paths).
• That's exactly what Cisco does with MHSRP. The document on Configuring Multichassis Multilink PPP states that when it tells it to configure half of the hosts with one default gateway and the second half with the other.
• Load sharing is inherent to the forwarding process of a router to share the forwarding of traffic, if the routing table has multiple paths to a destination.
• If equal paths, the forwarding process will decide the manner of forwarding and forward packets based on the load-sharing algorithm used.
• This still bears the possibility of unbalanced forwarding.
• If unequal paths, the traffic is distributed inversely proportionally to the cost of the routes.
• That is, paths with lower costs (metrics) are assigned more traffic, and paths with higher costs are assigned less traffic.
• Load sharing is a term used when attempting to share some of the traffic across multiple links.
• A good example of load sharing is when having two devices connect using two links of different speed. Let’s say link one is 9Mbit/s, and the other is 3Mbit/s.
• For every three packets we send through the 9Mbit link, we would want to send one packet down the 3Mbit/s link.
• The result is that the 9Mbit/s link would send a higher proportion of traffic than the 3Mbit/s link.

Load Balancing:

• Load balancing is a concept that aims to make a network more efficient.
• Load balancing distributes of traffic load evenly across a network with multiple-paths, in order to get optimal resource utilization, maximize throughput and minimize response time.
• Thus load-balancing will split the traffic down the configured paths equally towards the destination.
• E.g., with two 768 kpbs links and 800 kpbs traffic at any point, conceptually with load-balancing each path should have 400 kpbs worth of traffic.
• Load balancing is an attempt to process traffic evenly across a network with multiple links.
• The reason this term is less preferred than load sharing is because it is difficult to achieve perfect load balancing.
• With load balancing, if I looked at two traffic graphs, I’d expect to see two identical amounts of bandwidth being used on each path to the destination.
• Because of the different ways I can achieve load balancing it can be difficult to achieve true load balancing across each of the paths.
• Load balancing means distributing the traffic evenly and dynamically among different paths to avoid link congestion and saturation.
• This can be done in a packet-by-packet basis or per destination in a round-robin fashion.
• The packets sent by a host follow different paths to the same destination. All paths belong to all hosts.