14 Comments

  1. A quick napkin-math calculation suggests storage to allow 5GB/user would cost about $1M per year. This is a non-trivial thing to implement…

    We started with much less, 250mb per user, and have only had a few run up against it so far (of course it’s their first year using the service). There’s also a long tail you have to take into account. The majority won’t use it, or they’ll set up a single blog (~70MB) and use it sporatically. There’s no need to adjust costs based on such large per-user numbers in my eyes. I’m also interested in how integrating cloud storage like S3 could reduce these costs. In my eyes storage is cheap, but that’s also the beauty of being in a small(ish) liberal arts college I suppose.

    • Yeah. The $1M/year for storage is a worst/best case scenario. Our data centre storage costs are ~$6/GB (including backup, server-class drives…) so it adds up quickly. It could easily cost well under that, assuming most users don’t come near a storage cap…

    • at 250MB/user, that’s a more reasonable $60,000/year for datacentre storage. interesting…

      we may have difficulty using S3 on the back end, because it’s on the other side of an imaginary line, and people get all uppity about which side of that line their 1s and 0s get stored on.

      • Thankfully for now we haven’t had much pushback from folks on hosting off-site (although I think in the long run it’s more cost-effective on-site but haven’t done a lot of research on it). Our larger cost is providing domains to all students. We’ll have a rate of around $9.50 per domain so even with if the entire campus takes advantage we’re looking at around $40,000/year for that. It also helps that this is being rolled out in phases so the costs start small and can ramp up as more space and domains is necessary (We’re projecting it still to be less than $100,000 with the entire campus using it, which isn’t terribly expensive but it’s also not nothing)

  2. Don’t create problems before you actually encounter them. Get an ec2/s3 instance going, get fantastico (or theirs a fully OS alternative that escapes me) and learn how to configure it with some additional educationally relevant apps (as well as leaving lots of the default ones there as they work just dandy.) Lauch it with 3-5 profs (and maybe their students, so 100 or so users); pick ones that are more likely to succeed and yet don’t totally stack the deck with early adopters. Less than $10k later you will have a working model that can scale up (and even migrate to Canadian soil if you find you really need to.) By year three it will have a life of its own.

    You know all of this already. It is immensely do-able. Iterate early and often, and do it in a way that is cheap to fail/learn. The institution may need LMS and other “enterprise” systems, but if it wants to grow and change these eventually, it doesn’t need top-down risk-front-loaded enterprise-y projects to replace them, it needs many smaller experiments out of which the next solution emerges. Has all of this LMS evaluation report writing gone to your head? 😉

    • but…. we can’t do ec2/s3 because of the whole US thing. that leaves us with starting on canadian soil. likely some cheaper ways to pilot it, for sure. just spit-balling best case scenario to see what the potential scale of such a beast could be.

      and, yeah… I am drunk with the power of report writing 😉

      • Actually, you can do ec2 – you are in Alberta (cf. U of A google apps contract.) Federal law doesn’t prohibit it, just provincial, in some jurisdications. But if you want to give into the privacy FUD, fine, pick one of the myriad options now popping up in Canada. Campuses have just finally gotten around to “virtualization” (ok, maybe not just, but lots haven’t even); ec2 and the like are a step beyond that (dynamic, self-provisioning) and they are no longer even the bleeding edge (cf Boris’ urging to investigate PaaS.) It’s not just the faculty who need to opportunity to investigate ad innovate but the IT providers too.

        • legally, maybe. but the actual users tend to freak out. we just did the campus LMS engagement stuff, and people got downright religious about things not being on US services. It’s irrational, but there’s not a damned thing I can do about it. Hopefully we can cure that insanity – lots of services are actually being run on US servers already, but somehow without raising the cackles of the zealot crowd…

    • what I was trying to get at, but muddled it, is that the datacentre is basically already in place to do this. we have Enterprise Systems For Doing Big Applications. If that’s set up properly, handling webserver accounts should be trivial. of course, it’s not set up to be that easy. but in a perfect world, this would be a simple matter of assigning resources within a flexible platform that’s already in place, and giving some more control to individuals to manage what gets run on it.

  3. Thanks for the post D’Arcy. You get at the nub of the problem. What you suggest sounds good, and I like Scott’s iterate up approach, but even for that, there are no administrative/IT/financial systems in place in unis that would support it. It’d be a hard case to make, and you’d need a sympathetic budget holder somewhere. So you need a centre, or maybe just a budget code, for ‘Tech innovation’ or something that tests these kinds of models. One of the issues is that even when we try it we mess it up eg I’ve heard lots of people from different unis say they’ve installed WP but then disabled lots of features, ability to add plug-ins because of security or support issues. I’m not sure how we resolve it, but I think it goes to the culture of institutions and IT support. We’ve become very enterprise focused and risk averse.

    • Martin, it’s exactly that kind of Centre for Innovation in Educational Technology (or similarly named, likely acronymized) that I’m hoping to help spin up at UCalgary. Too early to say much more (there aren’t details yet, but a plan is afoot).

Comments are closed.