Discover the key to revamping a cross-carrier WAN to make real-time applications run well
In our last blog we explained how inadequate QoS standardization across carriers degrades real-time application performance to unacceptable levels in multi-carrier corporate WANs that grow higgledy piggledy. Getting QoS to work along the long chain of different administrative domains we described requires a herculean effort. This week we again draw on our colleague John Bartlett’s first-hand experience, this time to describe how to revamp a corporate WAN to make real-time applications run well.
In the previous post we talked about using network monitoring tools to watch the behavior of an extended network like that shown below. But if you can restructure the network, you can prevent many of the performance problems from occurring in the first place–so this time we’ll tell you how to fix the situation at a design level.
The most straightforward approach is to have a single service provider connecting all international sites as shown here.
If your enterprise has offices in major international cities this solution may be quite plausible. Having a single service provider means there is only one QoS standard, and there are far fewer administrative domains to manage. Traffic can be marked to a single standard and monitored at the carrier edges. You can then use tools to quickly assess if the carrier is meeting its SLAs.
But some enterprises, by nature, cannot use a single service provider. If your company has grown by acquisition, you have to manage with the facilities and locations you are dealt. Your business locations may need to be near the populations you serve, the minerals you mine, or the resources you need for your manufacturing processes (e.g., materials, energy, water). In these cases, John suggests the architecture shown below. This is similar to our first figure, but adds a direct path between the regional carrier and the international carrier.
Session Border Controllers (SBCs) terminate the VPN within a regional carrier and pass only the voice or video traffic through to the VPN within the International carrier. There is no routing across this path. The SBC can manage address translation if required, and can also re-mark the DSCP markings of packets as they pass through. So if there is a different QoS standard deployed in a region and in the international carrier, the SBC can handle the translation.
If you deploy your voice or video infrastructure components–especially bridges–in a co-location facility connected to the International carrier, you get what John calls a “Communications Center.” This is different from a data center because it sits in the WAN cloud and aggregates traffic at the right location. The communications center has a different security model than the data center–one based on SBCs instead of firewalls.
John sees enterprises running this architecture today with the help of third parties who provide the connection and SBC support, and he predicts we will see global companies adopting this architecture as they realize its benefits for global communications.
