How to Overcome Multihomed ISP Link Challenges - Page 2

Multihoming ISP links

Today, problems associated with ISP link availability continue to cause organizations to lose millions of dollars each year. However, deploying a solution that is cost effective and operationally efficient can also be a challenge. The following are four alternatives on how to facilitate multihomed networks.

1. Border Gateway Protocol

Typically, larger organizations multihome their sites with two links from two separate ISPs, using Border Gateway Protocol (BGP) to route across the links. While BGP can provide link availability in the case of a failure, it is a slow and complex routing protocol. It is costly to deploy because it requires special Autonomous System (AS) numbers from the ISPs and it requires router upgrades to be installed.

BGP is also not well-suited to provide multihoming and intelligent link load balancing. In the case of a failure, ISP cooperation is often required for link recovery. In general, BGP causes long and unpredictable failover times, which will not meet high availability requirements.

2. WAN link load balancing

Also known as multihoming, WAN link load balancing is a session-based process of directing Internet traffic among multiple and varied network connections. It requires a single WAN link controller located at the main site between the gateway modems/routers and the internal network. It intelligently load balances and provides failover for both inbound and outbound traffic among the network connections. Assuming there are two ISP connections, both network connections can be used at the same time. The benefit here is that you don't pay for bandwidth that is only used as a backup for when an outage occurs.

For example, traffic will go through network connection number one. If the WAN link controller detects that connection number one is overtaxed or failed, it will direct users across the second ISP connection. Intelligent WAN link controllers will continuously spread the traffic across the network connections based on the available resources. For example, with two T1s, it will not wait until the first T1 is overutilized before sending traffic out the second WAN; it will make use of both lines evenly.

Having two 1.5Mbps network connections does not mean that users have 3Mbps of bandwidth available to them. You would have 3Mbps available, but not for a single session. In other words, you would have 3Mbps of available bandwidth, but only 1.5 of throughput could be dedicated to any individual session. A single session will still only have 1.5Mbps of throughput; as with WAN link load balancing, each user can use only one ISP connection at a time.

3. Site-to-site channel bonding

Site-to-site channel bonding is a form of WAN link load balancing with a different approach that can increase the total combined network bandwidth of multiple network connections between two locations. This approach requires a WAN link controller at the main site and also at the remote site. Unlike WAN link load balancing, site-to-site channel bonding conducts continuous health checks (up and down status) of the network connections in use, and uses packet-based load balancing to distribute traffic across all network connections. However, with site-to-site channel bonding, two 1.5Mbps network connections will equal approximately 3Mbps, providing all traffic with the combined throughput from the multiple network connections.

4. Multiple ISPs

Organizations can multihome their sites with two WAN links from the same ISP. While implementing this solution may be a lower cost to deploy, it is not a very efficient solution, as an outage at the ISP will still cause a network failure, or at least create a bottleneck when both links are unavailable or oversubscribed. For greater WAN redundancy, it is best to have two or more different ISPs and load balance and provide failover for traffic among them.