User acceptance test plan for 1000-branch WAN
My vendor has asked me to prepare a user acceptance test plan for a WAN of 1000 branches, which have leased line and ISDN connectivity. What are the specific parameters I should look for at each branch connection? The vendor has just specified ping and SH command. Can you give me a standard test plan for testing WAN/router at a user site?
Acceptance test plans are the underpinning for your service-level agreement. Whatever you accept may be the best you will ever get with renegotiating or even changing providers. At 1000 branches we are talking 7-figure contracts (U.S. dollars). While the vendor has just specified ping and SH(ell), I would want latency, if at all possible, over the longest time period they will allow. Latency measures transit times more effectively than ping alone and tests at morning (7 a.m. to 9 a.m. local), noon (lunch time) and night (5 p.m. to 7 p.m. local). It gives you a better idea of what the network performance will be. If you have a standard application you run over the network (i.e., client-server database), put it on a laptop and try it at each location. If your application won't work then great ping times won't save your job.
A standard test at each site should be:
- Attach laptop. Do you get an IP address in a reasonable length of time?
- Connect to your login server. Can you actually log in? Do NSlookup.
- Do the names of your servers resolve? Tracert the path. Any surprises?
- Does your application work? Is the data entry person going to yell when the system can't keep up with their typing?
Other questions to ask:
- Who owns the network?
- Does this side own ALL segments?
- If like most deals, it goes through various companies. Who do you call? You want one person on the hot seat for repair, AND you want to know who to call if the failure is at some third party on a Friday night.
This was first published in October 2005