Having open books doesn't change much. We host with Exodus, and they very suddenly
had massive financial problems that caused a number of IT people around
here to run around looking stressed for several weeks. Very few people saw it coming,
unless you count the Slashdotters who said (in reply to the "Exodus goes bankrupt" story)
"Yeah, I saw it coming, I was there a few weeks ago and the data centre was half empty."
We have a new colocation set up with Q9 networks
(Canadian) and they seem to be doing fine.
It's difficult, by looking at a hosting company, or even by talking to them,
whether they are any good, largely because it's so easy to fudge their numbers and you have to really press them
to learn what their methodology is. Do they call uptime "being able to get ICMP replies to pings",
or that all their services are up and taking orders?
Nearly any top-tier hosting company will be able to get power and packets to your computers
reliably with many 9's in the number. The real source of problems is, as you might guess,
the carbon-based, fallible life forms that attend to the devices.
To address the human falliability problem, you either need 3 sysadmins (if you give your
sole sysadmin the pager all the time, he will go nuts, become cranky, and quit within 6 months). The best
arrangement is one sysadmin god to set policy, 2 to act as backups and generally knowledgable,
and one minion to run around installing Windows, fixing dead desktops, running network cable,
and dealing with vendors.
If you can't afford 3 sysadmins, managed hosting seems the best way to go.
I would be inclined to find a professional company who will take care of your colocated equipment. I would be inclined
to seek out a company where they are big enough to provide a person to pick up the phone 24x7x365,
but small enough that you can have a dedicated, identifiable team assigned to you rather than just getting
a different person each time you call.
I would also ask for a demonstration of their monitoring system. No brochureware. Show me the system in action,
and what happens when someone trips over a power cord or unplugs one of the drives in a SCSI RAID or
accidentally types 'halt' rather than 'reboot'.
