Nick Whalen <[email protected]> writes: > I'd prefer facilities that don't meter > power, as the VM hosts and SAN that I plan on dropping in aren't, > shall we say, green :).
Hm. My experience has been that power is what really costs money, much more than rackspace or bandwidth, I know I spend more on power than anything else. Most co-location centers are of the mixed-air type and don't have particularly modern or efficient cooling; every watt your servers burn is going to take two or three watts to cool, and that cooling cost is usually built in to your power costs. Of course, most of my experience is somewhat to the south of you, in California, but I bet it holds true up there, too. To give you an example, I can get 15a of 120v power (that's 11.25a that I can actually use) and a full rack here at a lower-end data center for $400/month (3 year contract) - and I can get 100Mbps bandwidth for another $100/month. At a data center in the next town over, I pay $1000 a month for a full rack with 2x20a 120v circuits. (that's 30a usable) - so the cost per watt is really pretty close, even when I more than double my density. In places outside of silicon valley, I would expect power and rackspace to be cheaper and bandwidth to be more expensive, but I'd still expect power costs to dominate rack space costs. Now, it seems that most co-lo providers are doing all they can to make the market less efficient by requiring a lot of negotiation before you discover the true prices, which seems especially silly on the 1/2-1/4 rack level, as that's lots of transactional friction in terms of negotiation and price discovery for relatively small amounts of revenue, but it seems to me that if you aren't specifying the power draw, or if you are specifying the power draw in a vague manner, it is possible that the wide variance you are getting in pricing is due to differences in the sales people's guesses as to the amount of power you will be using. As far as I am concerned, metering power is the primary challenge to selling sub-circuit sized co-location. Power draw is variable, and /everyone/ is trying to push it just a little bit... and if one person pushes the circuit a little too far? bam. everyone sharing that circuit is down. My personal opinion that the only sane way to do it is to put everyone on metering switching PDUs, and have an automated system that automatically warns people when they near the edge, and then automatically shuts down that customer before they blow the whole circuit. Of course, this opinion is as yet untested. I have yet to commoditize my own co-location offerings. _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
