As the business world continues to become more reliant on Internet access to do business, with more and more of what we do every day in business sits in “the cloud” it is becoming critical for even small businesses to have highly redundant Internet access. Most network providers and large businesses maintain large connections to different Internet backbones and use complex technologies to make the network operate through a failure of one of those paths. Small to mid-size businesses are generally reliant on a single connection or a single and a back-up without any automated failover. In the server world of small business redundancy in server hardware is common, and High Availability clusters are becoming more common, where redundant systems communicate with each other and automatically fail over when the other one fails. We are in the business of bringing the high availability (or “HA”) concept to network connectivity and are introducing our HA Internet Access product for small business that bring the small business high availability over inexpensive circuits, and we handle the complex technology to make it fail over.
Most voice and data outages to a customer premise are due to “last mile” issues, referring to the connection between the service provider’s edge router and the customer’s office. The “last mile” may actually be several miles and is typically a single physical wire pair, coaxial cable, or fiber-optic connection through city streets on poles or in underground conduits. The last mile may also be fixed wireless in some areas.
There are several options for mitigating last-mile failures, each with advantages and disadvantages both in terms of protection provided and cost.
One option is to order two similar circuits and either share bandwidth between them or use one as a backup. This helps to some extent but doesn’t prevent many failures from causing an outage. If both circuits run through the same manhole that fills with water or on the same pole that is hit by a car, they will both fail.
Dual cable modems provide very little redundancy. A single wire goes to a splitter in your office. Both modems take the same path on the same wire to the splitter.
Another choice is to use two dissimilar circuits such as T-1 and DSL. We have been doing this at Impulse for some time. This is better than two similar circuits in that the switching infrastructure is diverse, but it still uses wiring in the same physical conduit or pole line.
Yet another possibility is to use both a cable modem and DSL or T-1. This offers a great deal of path redundancy. While some of the outside plant may share the same poles, the cable distribution points are in different locations than telephone offices and the topology is diverse enough to offer quite a bit of protection. It’s very unlikely that both will fail simultaneously in day-to-day operation barring a major disaster.
Having both cable and telco physical facilities from two providers is a good option for reduction of last-mile issues causing an outage, but it comes at a cost. Each provider will provide different IP addresses. This means that servers such as Exchange, Citrix, and the like will need to be re-addressed should the primary path fail. This scenario can also cause problems with VPNs, VoIP phones, and other applications. There are appliances that can alleviate some of these issues as well as DNS tricks, but in the long run there will always be issues caused by renumbering IP addresses as well as downtime while things reconverge. At best, these appliances are a compromise. Troubleshooting can be aggravating. Is the problem with the telco provider, the cable provider, the appliance, or configuration?
Ideally, the best scenario would be diverse last-mile paths, both cable and telco, without the need to renumber should either link fail. Even better would be to have this managed by a single local company that monitors the connections and deals with the last-mile providers. No special hardware, no DNS tricks, no reachability issues should you lose a link. In other words, an extra last-mile path between you and the world’s voice and data networks without any of the hassles of multiple carriers, multiple contracts, multiple trouble reporting contacts.
Impulse is now in a position to offer this to our customers in the southern Santa Barbara County area. For the last several months we have been developing a diverse network using both Cox Cable’s HFC cable plant and conventional T-1 and DSL via Verizon copper. This means that our customers can have the advantages of bulk downloads over cable as well as low latency of T-1 and/or DSL. Because our connection to both last-mile transport providers is at layer 2, our customers’ IP addresses don’t change should wither medium fail. VPNs continue to work, mail and file servers are reachable over either path, IP phones don’t need to re-register with a new address.
Also because this is at layer 2, our customers can enjoy all of the Impulse services with path redundancy including voice, MPLS private networks, and IPv6 connectivity.
How does it work?
Each customer cable modem terminates at Impulse as a separate VLAN from Cox Cable. These VLANs can be further tagged to separate data for MPLS, voice, or other services.
Similarly, customer T1 lines can be configured with multiple PVCs to deliver both private and public data networks, voice services, etc.
Under normal circumstances both networks are used. Voice traffic utilizes the T-1 circuits for low-latency and low jitter. Bulk file transfers and general Internet browsing take place over the faster cable connection. Private data networks such as MPLS can be configured to use either transport depending on need.
In the event of a failure within the cable plant, bulk data traffic is automatically rerouted over the T-1 or DSL circuit within 30 seconds and an alert is generated to the Impulse NOC advising of the failure. Quality-of-service configuration within Impulse assures that voice traffic isn’t adversely affected by the increased usage. The customer’s IP addresses do not change and other than a decrease in available data bandwidth there is no customer impact.
In the event of a failure within the telco copper plant, voice traffic is rerouted to the cable network within a period of 30 seconds and an alert to the Impulse NOC will be generated. Phone calls in progress at the instant of the failure will experience degradation during the transition but phones will not need to re-register and IP addresses will not change.
MPLS traffic similarly will transition from the affected to the unaffected network.
When the affected network is restored, traffic flow will revert to normal.
Think of this deployment as having two diverse last-mile connections, each of which is best suited to certain tasks but capable of sustaining your operation should the other become damaged or go out of service.