gmail outages? what gmail outages?

I think Dan nailed it with this post

Wow! There’s been plenty of coverage of the recent massive (yeah, right) 2 hour GMail outage. The end of cloud computing? Erm… I don’t think so.

…

Go figure how much all this would cost you: email server, email software, o/s licenses, backup software, hosting, ongoing management and maintenance, staff costs, etc… Yes. This all adds up. Compare the costs against GMail…. Erm.. That’s a no-brainer.

Hey – ask yourself this as well. For those in corporate environments, how many times is your email service down? Probably more than GMail I would say.

…

A two hour email outage is not the end of the world, nor for cloud computing.

Amen, brother :) Until Juha commented about it in the Friday Fryup, I hadn’t even realised there WAS an outage!

For a while, and to anyone who’ll listen, I’ve been a very vocal user of google apps – both our domains email are hosted there (fastchicken and verdandi), our email archive (about 3GB / 6 years of email), we use calendar for pretty much everything, google docs for a few things, sites, reader, my iphone is hooked up to it, etc. While I’m not 100% comfortable putting all my eggs in one basket, it’s proven to be a very robust basket. And as Dan points out, correctly:

Email ain’t everything. Get over it.

And I think – and this surprises me about Juha’s reaction, because I KNOW he knows this better than I do! – that most people using email (I’d say 99.999% of people using email) have no idea how robust and redundant the email protocol (SMTP – Simple Mail Transfer Protocol) is!

For example, my domain has 7 mail servers that can receive email for it:

fastchicken.co.nz       MX preference = 20, mail exchanger = aspmx5.googlemail.com
fastchicken.co.nz       MX preference = 20, mail exchanger = aspmx3.googlemail.com
fastchicken.co.nz       MX preference = 20, mail exchanger = aspmx4.googlemail.com
fastchicken.co.nz       MX preference = 20, mail exchanger = aspmx2.googlemail.com
fastchicken.co.nz       MX preference = 10, mail exchanger = alt2.aspmx.l.google.com
fastchicken.co.nz       MX preference = 10, mail exchanger = alt1.aspmx.l.google.com
fastchicken.co.nz       MX preference = 5, mail exchanger = aspmx.l.google.com

These are in reverse order, so the preference = 5 one is tried first, then either one of the 10′s etc, until either none of them work (unlikly) or a working one is found.

If, on the off chance, seven of google’s server clusters are down (um, yeah….), the sending server will just store the email, and wait a period of time, and try again. Usually it’ll retry again after 5,10,30 then 60 mins, then it goes up into the hours, and after about three days, it gives up. All of this is configurable. And after the longer ones, it notifies the sender that it’s still trying, but there is a problem.

So, the chances of an email being lost from an outage of less than 72 hours is close to zero. The chances of it being delivered on the first try are extremely high. And if the 72 hour issue is a risk factor, you can put another MX (Mail Exchange) record in at a separate provider (I used to use dyndns.org) which has a much longer retention policy. The statistical probability of google AND dyndns.org going down at the same time is pretty damn low. Using one MX record (pointed at a server at the end of my DSL/cable connection) and ONE backup MX (dyndns.org), I managed outages of up to 30 days (moving house) without a single lost email.

Access to your inbox is another matter. If it’s that big an issue, then offline your email using IMAP or POP3. Personally, I like the flexibility of any-where access, and I’m willing to take the chance of soemthing going offline for a period of time. Even tho I pay $50/box/year for it, I think the downtime in the past 18 months has been in the seconds.

And, after all, it’s just email. If it’s that urgent, pick up the phone. Or go outside and read a book or something.