Thanks for your input Joel...Yes, it's a lot of bandwidth, today. In fact, our smallest customer is 10 meg/sec. Our biggest is 10 Gig/second. Here in Silicon Valley California most companies are outsourcing everything except the circuit they need to access it..it's the new portability initiative! I recall 13 years ago when I said I was going to start a Fiber Only ISP...everyone including my previous VCs, Hedge Funds and business partners in my first ISP laughed at me. It was the Dot Bomb period. Today, everyone here asks for fiber to do all this crazy things they now can't live without. It's all about Bigger, Faster, Cheaper and mostly Store it someplace else that has lead to these big pipes. Thank You Bob Evans CTO
My suggested rule of thumb if you can't actually measure the traffic in advance for your population: count the number of workstation devices that will be your network, figure at least 0.5 Megabit of WAN for each typical business user workstation or laptop.
I can't help but laugh (laughing with, not laughing at--all due respect to the NA part of NANOG) at this.
I've been spending the last 4 years working on various UN networks where getting 0.5Mb of bandwidth to a site can be a challenge, and 4 Mbit/second for an office of 8 users is an unaffordable luxury. And these are sites where the end users want to move to Office 365.
We've done a bit of testing, and one of the issues with O365 is that O365 is a BIG thing and you have to decide which slice of O365 you are calling "O365" at a particular site.
For some people, that's just "outsourced Exchange" (in which case we would allocate 30K-50Kbps per office user downstream bandwidth, and drop in a WAN Opt box plus do some shenanigans to break into the HTTPS through proxy).
For other people, O365 is the whole "nothing is on my hard disk (but cache)" thing, plus Lync (not just voice, but voice+video). Those folks really are going to require major bandwidth; this is where numbers like 512K/simultaneous user make more sense.
You can excuse (or at least explain) Microsoft's lack of benchmarks and guidance because of the complexity of O365 and also because they have the sort of North American viewpoint that makes it hard for them to understand high latency/low bandwidth pipes.
They try hard, but often just don't get it because of the amazing resources and richness available to a company of that size. I had a great conversation with them about 3 years ago about Exchange and AD forest design where they were strongly advocating centralizing everything in data centers, rather than pushing anything like a DC or mailbox server out to a branch office. When I asked about the bandwidth required, they said that it was "not much." Pressed for details, they said "we do it ourselves, and it hardly impacts the bandwidth on our most poorly-connected offices." Pressed even further, it turns out that a T3/E3 is the lowest link they would consider acceptable for an office. (My total upstream bandwidth budget at one agency for 100 offices and 9,000 users in 24 timezones is less than a single T3... Thanks Microsoft!)
Anyway, not adding much to this conversation since it's clear that Bob is asking in the context of "bandwidth is cheap, fast, and inexpensive," but I couldn't help but giggle at the kinds of numbers you guys are throwing around here for people to read email and work on spreadsheets.
jms
-- Joel M Snyder, 1404 East Lind Road, Tucson, AZ, 85719 Senior Partner, Opus One Phone: +1 520 324 0494 jms@Opus1.COM http://www.opus1.com/jms