We own a couple servers and RAID units in a colocation facility. I am trying to wrap my head around pricing various services that my firm is hosting in this private cloud and selling to clients. Examples are: Crashplan backup hosting, colocated servers, colocated NAS, virtualized servers.
The challenge I'm running up against is figuring out actual costs for each type of hosting application we provide, so I can then determine pricing structures.
For example, if we colocate the client's server or NAS in our rack, we have the following costs:
a. amps used by their equipment
b. rack space used by their equipment
c. bandwidth used, 95th percentile billing
But if we are charging the client for Crashplan backups stored on our equipment, we have:
a. amps used by OUR equipment
b. rack space used by OUR equipment
c. bandwidth used by their data (again, 95th percentile)
d. storage space used on our RAID
Costs A and B are quasi-fixed; they only increase once we hit certain thresholds (i.e. the rack fills up, or we exceed our contracted power usage). But we're far from exceeding either of those. Storage space, too, is so cheap that it's quasi-fixed, though less so than costs A or B. Do I consider the costs of power and space as operating costs that must be paid for via a break-even calculation on the more direct costs like bandwidth? Or do I figure out a per-unit cost for them and factor that into the direct costs for each client's services?
Speaking of bandwidth: 95th percentile is really weird: "324GB per Mbit/sec figure is a roughly accurate [amount of total data transfer] for a 31 day calendar month of continuous 1 Mbit/sec utilization. If you are billed based on 95th percentile usage, as are most dedicated and co-location products, those numbers like 324GB are totally irrelavent [sic.] to you. Typical hosting traffic patterns result in approximately 190GB of data transferred per 1 Mbit/sec billed on a 95th percentile model."
So, do I make my estimates based on typical traffic? If we're providing multiple types of hosting services, each one will have its own traffic patterns that only real-world measurements will reveal. (This is getting beyond my networking fu to measure, too.)
My first thought is: if we are paying $80 per month per Mbps to our colo, and each Mbps allows a client to transfer 190GB in a month under typical traffic patterns, then each GB transferred costs us 42 cents. Thus I should base my pricing on the expected data transfer for that client; if we estimate that the client will transfer 95GB in a month (50% of 190GB), my cost for that client's bandwidth is $40 (50% of $80 per Mbps), so if I want a 50% margin I should bill them $80 for bandwidth. But something's not sitting right about that; maybe it's because I'm still trying to convert 95th percentile billing into a metered-bandwidth measurement, and it is not that simple.
Brain shutting down now. I knew there was a reason I avoided accounting classes.