Welcome to Your Weekly Paradigm Shift: Very Large "Free" Web E-Mail Accounts

This week I asked J'e St Sauver to give us his take on the implications of Google's Gmail for higher education institutions. He's done an excellent job of summarizing the pros and cons and suggesting some interesting twists to the implications.

----------------------------------------------------
J'e St Sauver, Ph.D. (j'e@uoregon.edu)

The folks at Google have turned their eye to free Web e-mail accounts, and in typical Google fashion, they're making quite a splash: each Gmail user gets a 1,000MB worth of disk space. To put that in perspective, if you're at a typical college or university, your university e-mail account probably has a quota of 50 to 100MB.

Under some circumstances, a "mere" 50 to 100MB quota may be more than enough. For example, if you routinely download your e-mail to a PC using POP3, 50 to 100MB will probably be more than enough space to act as a buffer in the interval between downloads-you can do the bulk of your long term e-mail storage on your PC's large hard drive.

Unfortunately, most people don't read their e-mail using POP3. They use Web e-mail (and leave their mail on the server), or IMAP (and leave their mail on the server), or they may even a command line e-mail client such as pine (which also leaves the mail on the server). When that's the case, 50 to 100MB to hold all your e-mail may feel like a Dixie cup of water at high noon in the middle of the Sonoran desert in August.

Google, on the other hand, will give you virtually all you'd want to drink, with plenty left over for a shower, for watering some needy cactus, and for making mud pies, too.

Moreover, Google gives users not just raw disk space, cool as that may be in and of itself, Google also gives you private indexing of your e-mail-the ability to privately "Google" a gigabyte of stored e-mail is obviously quite helpful, and not something that non-Google folks will be able to deliver (by definition).

Notwithstanding that, I predict that others *will* try to keep up with Google. For example, Yahoo! Mail Plus will be offering 2,000MB of storage, POP access, mail forwarding, and other features for a nominal yearly fee, and Hotmail Plus will be doing much the same thing.

For this audience, however, the interesting question is "what will colleges and universities do?"

Will colleges and universities join the keep-up-with-Google club, or will they try to just muddle along with their good old 50-to-100MB quotas?

There's risk either way. Let's begin by considering what happens if colleges and universities try to "keep up with the Googles."

Going from 50-to-100MB quotas to 1,000-to-2,000MB quotas means that suddenly I/O throughput will become key, and unfortunately many university e-mail systems skate right on the thin edge of I/O congestion at the best of times, largely due to our friend Mr. Spam. Increasing the size of your faculty/staff/student e-mail disk quota puts you at risk of saturating your disk farm, either in terms of operations/second or delivered throughput.

At least for some POP3 and IMAP based architectures, the key equation quickly becomes:

(number of users)*(inbox size in MB/user)*(access frequency)

Looking at that equation, and assuming everything else remains constant, a 10 or 20 times bigger disk quota implies a proportionately larger I/O load. Is your disk farm ready to handle 10 or 20 times its current I/O load?

Other factors to keep in mind:

  • With larger disk quotas, users want and expect larger per message quotas (at 1GB, 10MB/message is no longer a big deal)
  • Choice of mailbox format and file system may also become quite important, if you think about performance issues associated with UFS-derived file systems attempting to cope with inserting and deleting inodes in a directory with lots and lots of individual files.You'll also want to think about how you back up and handle restoration of files in large disk array.
  • There are network implications to accessing that volume of storage (deployed disk space is an excellent proxy for/correlates strongly with bandwidth requirements-are you ready for the rise of private e-mail music trading, for example?).

But let's assume that you decide that you *do* want to stand up to the "Google challenge," and offer your users not just 1,000MB or 2,000MB, but 5,000MB.

In terms of funding something like a 5GB quota for 20K hypothetical users, the raw disk (assuming you use high-density serial ATA drives) would potentially be quite doable:

(20,000 users)*(5GB/user)
------------------------------------- * 1.25 (RAID overhead, spares, etc.) ==>
(250 GB/drive)*(8:1 oversubscription)

That's roughly 63 drives @ ~$185/drive ==> $11,562 for just the raw disk itself.

Of course, if you're buying five dozen drives at once, you can expect to see a somewhat better price than the typical onesie-twosie price mentioned; that 8:1 oversubscription may be unduly pessimistic, too.

3ware and other vendors offer 8 and 12 port SATA controllers
(see for example http://www.3ware.com/products/serial_ata9000.asp), with street prices well under $1K/controller. You'd obviously also need systems to hold and drive the disks, plus a backup solution, but my point is that the raw ingredients aren't that expensive any more. (And if you don't want to screw around buying and integrating hardware, doing just a cursory search, I'm seeing 4TB SATA NAS building blocks with 16 drives from commercial vendors for $13K or so; multiply as required to get the capacity desired).

Before anyone mentions it, yes, there's no doubt that these boxes aren't the same as the top-of-the-line fiber channel-based boxes that folks would prefer if money was not an issue, but money is an issue at most schools. I think you might be surprised at the capacity and performance you can get from SATA-based storage solutions.

What if you decide not to try to compete with Google?

If you decide not to try to compete with Google, you should expect your users to migrate from your local e-mail accounts to one of the new huge Google accounts, some probably forwarding their institutional accounts to their new accounts in the process, some probably not much caring that they have an institutional e-mail address at all.

That has some interesting implications.

If students move to Google and don't forward their mail, you sure better hope that they remember to check their old, sad, sorry, puny institutional account, since that's probably where the university is going to be mailing Important Official Notices That They Should Read.

If students move to Google and do forward their mail, it will be a miracle if you don't end up having your school's e-mail blocked. Remember that when you forward e-mail, from the point of view of the host that's receiving the forwarded e-mail, the e-mail appears to be "coming from" the forwarding host. Thus, when a spammer sends e-mail to your students' institutional e-mail account, and the students in turn automatically forward it to their new Google accounts, if Google happens to looks at that spam, it is going to "look like" it "came from" *your* institutional mail server. We all know what happens to spammy looking hosts, right? They get blocked, yeperoo.

There's also the issue of wide area bandwidth. For the first time students have a serious remote storage resource available to them. Are you ready for the Internet traffic that will be associated with all that storage?

But don't despair: Gmail isn't perfect (even though it sure seems pretty dang good based on the Gmail account I've been testing). Specifically:

  • Google Gmail accounts are still in test deployment, and are available by invitation from an existing Gmail user only (then again, if you ask around, you can probably find a friend or acquaintance who has a Gmail account, and more likely than not they'll probably have a spare invitation which they can share with you).
  • When you pick a Gmail username, it has to be at least six characters long. J'e, Tom, Sue and Sally simply aren't available as Gmail usernames, sorry.
  • Gmail d'esn't work with all Web browsers. As of July 13th, Gmail supports IE 5.5 and later, Netscape 7.1 and later, Mozilla 1.4 and later, Mozilla Firefox 0.8 and later, and Safari 1.2.1 and later. Opera users, sorry, you're not on the officially supported list.
  • You need to have Javascript and cookies enabled, which can be rather suboptimal from a security point of view.
  • When you use Gmail, Google d'es show you advertising, and the advertising is automatically targeted based on the content of your mail messages. Thus, if you're talking about going to Hawaii with a correspondent, don't be surprised if you see ads for airline airfare specials to Hawaii, or Hawaiian hotel advertising. (On the other hand, the ads are low key and unobtrusive, rather than obnoxious popups, popupunders, or banner ads).
  • Some folks worry about Google being able tie their identity to their Web searching and browsing habits via the cookies it uses. (But Gmail d'es have a privacy policy, which you can review to see if it is sufficiently strong to address privacy concerns you may have).

Bottom line, one way or the other, Gmail is here and here to stay. You probably are going to want to start thinking about how you and your institution will cope with it, if you haven't already done so.

Wow. What an excellent summary of Gmail and its implications for higher education institutions.

My own opinion . . . Gmail is what we should be doing for our students, faculty, staff, and alumni. My bet is that this is going to evolve into users having a "magic purse" on the Internet that they can put everything digital that ever might need into and access it any time from anywhere. Wh'ever provides that magic purse has guaranteed eyeball time from the user, and this could be a way for institutions to lock up their academic communities through entire user lifetimes. Maybe an institution could license its own "brand" of Gmail from Google?

P.S. If anyone has a spare Gmail invitation, I'm looking to try it out myself: terry.calhoun@scup.org. I've got 300,000 e-mail messages stored on my hard drive and the thought of being able to Google that equivalent in Gmail-once more have accumulated-is very attractive.

comments powered by Disqus

Campus Technology News

Sign up for our newsletter.

I agree to this site's Privacy Policy.