We’ve been happy customers of Peer1 in New York since 2003 when we first moved in (a process that started, in part, from this hearty recommendation). Mike, the facility manager, has been great to work with, and the two or three times over 4 years that I’ve had to haul someone in on the weekend to do a hard reboot they’ve always been smart and helpful. The bandwidth has been rock solid, and with the exception of one “unplanned power event,” the facility has been all we could have hoped for. If you’re looking for colo, I have no hesitation in sending you their way.
When it came time to refresh Yankee’s hardware this fall, we looked at simply purchasing another fleet of Dell PowerEdge servers — they’ve served us well for almost 10 years — but also decided to look at “managed hosting” as an option.
When we started the process, the whole “managed” vs. “self-managed” vs. “colocation” was a mystery to me; I assumed that “managed” meant a return to the old PEINet days, with a server controlled by overlord go-betweens.
As it turns out, that sort of regime is called “shared hosting.”
Managed hosting involves, in essence, leased hardware (which they take care of, repair when needed, etc.) over which you have complete control, supplemented by a various a la carte services like managed OS package updates, backup, and service monitoring. It’s more like “renting part of a sysadmin” than it is like “installing a troll to guard the bridge.”
In other words it’s a good fit for a situation like ours where on one hand we want to be able to have complete control over the iron but, on the other, could benefit from an extra pair of hands.
And so we dove in.
So far the process has gone very well: tech support responds quickly to tickets and answer the phone on the first ring. The servers they provisioned for us — 2 Intel 64-bit machines and 3 32-bit ones, all running RHEL5 — were ready 3 days after we placed the order.
In 2003 the migration to colocation in New York meant buying servers, getting them delivered to Dublin, NH, flying to Boston and driving up to Dublin, spending 4 days installing RHEL on the servers and then migrating over our data and content, then loading the servers into a car and driving to lower Manhattan to install them.
In other words, a lot of physical work, supplemented by considerable pizza.
This time there was no physical work at all, just a lot of scp from server to server over the last month.
Not content to simply do a straight migration, I decided that it would be a good opportunity to move our development process into the 20th century, so most of my work of late has been as much about installing Subversion and Trac and moving our 44,000-odd pieces of code and related objects into the repository. Among other things this has been akin to an archaeological dig through coding-past for me, as this is a project that, in one way or another, I’ve been involved with for 11 years now.
Today was the final “flip the switch” day. This meant coming into the office early this morning, taking the old servers offline, copying about 18GB worth of MySQL data over the wire, changing the DNS to point to the new servers, and then bringing each of 7 websites back only in step, dealing with any niggling migration issues as I went along (somewhere in there I learned how to make a built-in PHP module into a loadable PHP module so as to add dBASE support — dBASE support! — to our PHP).
The switch is now fully flipped. We’ve got a few more days worth of power and bandwidth in NYC to let me pull over anything I missed in the process, but now all the action is in Miami.
Time to go home and get some sleep…