I'm still working through the issue of how to connect multiple WAN providers to get the best global connectivity for Telepresence. I described the problem in this post, and then last week discussed one solution in this post. A few weeks ago I had the privilege of spending some quality technical time with Jason Redisch, Principal Architect for Virtela Communications, and he described their solutions for this challenge.
I'm still working through the issue of how to connect multiple WAN providers to get the best global connectivity for Telepresence. I described the problem in this post, and then last week discussed one solution in this post. A few weeks ago I had the privilege of spending some quality technical time with Jason Redisch, Principal Architect for Virtela Communications, and he described their solutions for this challenge.Here is the challenge. We have an enterprise distributed across the globe hoping to run telepresence at key locations. For reasons of coverage, contract or price the enterprise would like to use multiple carriers to get connectivity between all their sites and have created an architecture as shown below.
This design puts strain on the New York office, creates a bandwidth bottleneck, and requires constant reevaluation as the Telepresence deployment increases in use and size. The solution shown last week solves this design by providing a single connection point that has very high bandwidth links to each provider, thus avoiding the New York problem.
Virtela has an interesting solution because they are by definition already connected to many global network providers. Virtela does not own wide area links between their points-of-presence (PoPs) like most network providers. Instead they connect to multiple other providers. Clever software running at each Virtela PoP does dynamic route optimization. This means they constantly watch the performance of each carrier, and then optimize how traffic is sent from PoP to PoP based on the current performance of the WAN provider and the needs of the application. Using this approach allows Virtela to be very resilient to network failures and to provide exceptional SLA specifications in support of video streams.
So providing a solution to the Telepresence problem described above is a natural for Virtela. They are probably already connected to the carriers the enterprise wishes to use, and can provide telepresence-class SLAs. Here is what the diagram would look like for Virtela:
The links between each enterprise site and the Virtela PoP can be dedicated links or can just be connections that Virtela makes to the existing enterprise MPLS VPN in WAN Provider A or B. Traffic then flows between Virtela locations on the best possible path and is handed off to the WAN provider or link supporting the destination site.
Virtela offers a series of connection methods with increasing levels of security to meet the needs of different enterprises. Connection to outside resources such as bridging and gateway providers or the Internet can be accommodated, or traffic can be maintained within VPNs dedicated to the customer.
Stick with me, there are a couple more solutions in the pipeline, I'll write more on this next week. This space is evolving quickly and the vendors are offering solutions because the enterprises want to make this happen.
