Each load balancing technique comes with its own set of pros and cons. Some of the common load balancing solutions includes DNS load balancing, bridged load balancing and routed load balancing.
DNS load balancing
DNS load balancing is achieved with a single name that resolves to multiple names or IP addresses.
Easy to configure and understand.
DNS based cluster nodes don’t require multiple network interface cards (NICs). Each machine can have a single NIC with a unique IP address.
Multiple IP addresses can be assigned to the host record. The DNS server can rotate these addresses in a round-robin manner and workload gets divided equally among the members of the Exchange Server cluster.
Load balancing pools for various geographic regions are established. The administrator can take advantage of infrastructure dispersed geographically and improve performance by reducing the distance between the receivers and data centers.
No native failure detection or fault tolerance and no dynamic load re-balancing.
No capability other than round-robin.
No way to ensure connection to the same server twice, if required.
DNS cannot tell if a server has become unavailable.
Cannot take into account the unknown percentage of users who have DNS data cached, with varying amounts of Time to Live (TTL) left. So, when TTL times out, visitors may still be directed to the ‘wrong’ server.
Load may not be evenly shared as DNS cannot tell how much load is present on the servers.
Each server requires a public IP address.
Bridged load balancing
Bridged load balancing uses a virtual IP address created in the same IP network as the real server. Packets designated for the virtual IP addresses are redirected to the real servers.
Can be embedded into an existing network with no additional IP networks required.
Could be easier to understand for simple networks.
Usually cheaper than a routed model.
Usually limited to a single local network.
Layer-2 issues including loops and spanning-tree problems can appear if balancing solution is not designed carefully.
Can be more difficult to understand for people used to layer-3 environments.
Routed load balancing
Routed load balancing is load balancing at layer three, in which the virtual IP address exists on one network with the real servers existing on one or more other networks.
Allows real servers to be geographically diverse providing expandability.
Easier to comprehend for people used to layer-3 environments.
No spanning-tree issues.
Layer-3 load balancing can be costly.
Requires additional IP address and network design to implement.
Network load balancing (NLB)
Network load balancing allows you to create a cluster of between 2 and 32 web servers, with each having its own computer name and static IP address.
NLB provides fault tolerance at the network layer ensuring that connections are not directed to a server that is down.
Good for scalability as it supports up to 32 servers per segment.
NLB is easily configurable.
No special hardware is required. Two network adapters can be used to mitigate a point of failure.
Unable to detect if a server is unavailable and can direct a user to a system that can’t provide the requested service.
There is no shared data.
NLB does not work with Layer three switches or Token Ring adapters.
All servers in a cluster must be in the same subnet.
While each of them is balanced out in terms of pros and cons, it comes down to what kind of traffic you expect, your budgetary constraints and the overall simplicity of maintenance.