I think maybe the 100MB internet pipe is a little over kill. Depends on what the business does but I have yet to see any large Corps requesting that large of a internet pipe. If you plan on doing HD quality video and audio teleconferenceing then sure I could see it. But we have used WAN optimization products that allows lower WAN BW.

Your backbone is pretty much the norm for companies now. Please provide more info on what the company is trying to do.

First of all, I appreciate all for time taken to provide input/thoughts.

K12 organization and simply looking to begin planning on preparing the network for voice, data, video convergence. Most of the traffic would be internal. No intent on HD video or audio conferencing to the Internet.

My biggest question would be what would be best for the site-to-site WAN, in terms of BW when thinking about voice, data, and video. I realize that QoS, Optimization, etc. are helpful with traffic/BW management. My slowest link is 10MB and it normally runs at 25& utililzation. So am curious at what your thoughts are on the 100MB WAN link... overkill? My feeling is that DS3 speeds for all WAN links is sufficient for a converged network.

Not sure what others who run a converged network have in regards to WAN bandwidth.

Thanks again.

When working with converged networks that are carrying a voice, video and data, you'll generally need QoS to insure your real-time applications, such as voice and video, obtain the bandwidth they need when they need it. How much bandwidth for those apps depends on the number of concurrent sessions and the bandwidth demand per session. If you can answer those questions, you've addressed half your bandwidth planning question.

Data bandwidth depends much on what your doing with data applications. For interactive applications, figure bandwidth allocation between 2x to 3x their average demand (this to avoid queuing delays). For non-interactive applications, bandwidth can be computed based on your timeframe for moving a fixed amount of data.

Most important is understanding applications that need a certain amount of bandwidth to optimally function, and those that are very elastic in their requirements, and treating applications differently using QoS.

Often I hear, just buy bandwidth until there's no congestion (rather than use QoS). It's true if you can avoid congestion you don't need QoS, but it can be a very expensive approach, especially when dealing with WAN bandwidth.

Real case:

New client backup application roll-out. Monday morning 300 hosts, all running with local 100 Mbps connection, attempt to push multiple gigabyte backup to backup servers at HQ site across pair of T1s.

If planning for bandwidth, alone, to avoid congestion, all that was needed was 30 Gbps of bandwidth. Should we obtain 30 Gbps of WAN bandwidth? What happens when clients, as planned, are upgraded to gig connections, should we then obtain 300 Gbps of WAN bandwidth?

Do you begin to see the difficulty of planning for bandwidth without understanding your application bandwidth needs?

Needless to say, T1s were fully saturated, but they were still running VoIP and more important apps across the links, which continued to function well because of QoS and bandwidth planning that accounted for the bandwidth required for the apps to function as desired.

What is your topology? Do you have a hub-and-spoke configuration with a Data Center somewhere, or are all services still decentralized? That, and your plans with regards to centralization, are important considerations when it comes to WAN BW.

- Thomas

Two data centers, with most critical services centralized between them. Hub-and-spoke topology at each data center, as I have additional VLANs at each location. Two internet gateways, so http traffic is already distributed. VoIP at DataCenter2 only, but do integrate with PBX at other non-VoIP sites for 4-digit dialing. No current video traffic. I do have three remote sites, which are decentralized for services like, file and print, non-critical apps. I would want to eventually centralized all. I do not have any current BW issues.

I have attached somewhat of a drawing depicting my WAN layout. Please excuse my drawing skills. All WAN links are currently microwave technology.

Thanks.

I would not think that 100MB are too much for your Data Centers when you look at a 3-5 year time frame, taking into account traffic increase and further centralization. For your remote sites, 100MB may be too much, and I would rather add addtional 10MB links or DS-3 links, and load-balance. This would give you additional bandwidth, and increase your redundancy. For example from RS2 to DC1.