Renting CPUs at the CAC
If you need exclusive access to one or more of the compute nodes at the CAC, we offer a program that allows you to rent time on the CAC systems. By renting time, we allocate either a set of CPUs (cores) in the general pool for you or we set aside the CPUs (cores) you need into a partition that can be accessed only by the people you specify. When you rent CPUs, the amount of time you wait in the queue is dictated only by the jobs you have submitted, you are not waiting on anyone else.
Nodes must be rented by the core (or CPU) in integer core-months. That is, you can rent one core for a month, or 2 cores for 2 weeks, or 4 cores for one week.
If you need very long times of exclusive access to nodes (a year or longer), you may find it worth while to purchase nodes instead of renting them. Additionally, some funding sources discourage the renting of computer time and prefer that the project buy nodes. In either of those cases, the Center for Advanced Computing can still help you: please see the information on buying nodes.
The rates for renting time at the CAC are below. Because of a subsidy from the College of Engineering, research groups within the College are exempt from paying the staff component of the rate.
| CAC CPU Rental Rates 2005 |
|
| College of Engineering | |
| Ethernet connected core | $38 / month |
| Myrinet/Infiniband connected core | $43 / month |
| Large memory node | $167 / 90 hours |
| Outside of the College of Engineering | |
| Ethernet connected core | $60 / month |
| Myrinet/Infiniband connected core | $65 / month |
| Large memory node | $242 / 90 hours |
Please contact the CAC via email at cac-support@umich.edu or call Andrew Caird at 7-5273 for more information or to set up a rental of dedicated nodes.
The CAC will complete the CAC Node Subscription Form and email it to you to complete and sign.
The CAC rate is approved by the University of Michigan Office of Financial Analysis and is reviewed and adjusted periodically. The Center for Advanced Computing is forbidden from making a profit; these rates reflect our costs for maintaining and running these clusters.