Hardware Reference
In-Depth Information
from a 100-wat PSU), 75% when the load is between 20% and 40%, and 65% when the load
is below 20%.
a. [15] Assume a power-proportional server whose actual power is proportional to CPU
utilization, with a utilization curve as shown in Figure 6.3 . What is the average PSU
efficiency
b. [15] Suppose the server employs 2 N redundancy for PSUs (i.e., doubles the number of
PSUs) to ensure stable power when one PSU fails. What is the average PSU efficiency
c. [20] Blade server vendors use a shared pool of PSUs not only to provide redundancy
but also to dynamically match the number of PSUs to the server's actual power con-
sumption. The HP c7000 enclosure uses up to six PSUs for a total of 16 servers. In this
case, what is the average PSU efficiency for the enclosure of server with the same util-
ization curve?
6.25 [5/Discussion/10/15/Discussion/Discussion/Discussion] <6.4> Power stranding is a term
used to refer to power capacity that is provisioned but not used in a datacenter. Consider
the data presented in Figure 6.25 [Fan, Weber, and Barroso 2007] for different groups of
machines. (Note that what this paper calls a “cluster” is what we have referred to as an
“array” in this chapter.)
a. [5] What is the stranded power at (1) the rack level, (2) the power distribution until
level, and (3) the array (cluster) level? What are the trends with oversubscription of
power capacity at larger groups of machines?
b. [Discussion] What do you think causes the differences between power stranding at
diferent groups of machines?
c. [10] Consider an array-level collection of machines where the total machines never use
more than 72% of the aggregate power (this is sometimes also referred to as the ratio
between the peak-of-sum and sum-of-peaks usage). Using the cost model in the case
study, compute the cost savings from comparing a datacenter provisioned for peak
capacity and one provisioned for actual use.
d. [15] Assume that the datacenter designer chose to include additional servers at the ar-
ray level to take advantage of the stranded power. Using the example coniguration
and assumptions in part (a), compute how many more servers can now be included in
the warehouse-scale computer for the same total power provisioning.
e. [Discussion] What is needed to make the optimization of part (d) work in a real-world
deployment? ( Hint : Think about what needs to happen to cap power in the rare case
when all the servers in the array are used at peak power.)
if [Discussion] Two kinds of policies can be envisioned to manage power caps [Rangan-
athan et al. 2006]: (1) preemptive policies where power budgets are predetermined
(“don't assume you can use more power; ask before you do!”) or (2) reactive policies
where power budgets are throtled in the event of a power budget violation (“use as
much power as needed until told you can't!”). Discuss the trade-offs between these
approaches and when you would use each type.
g. [Discussion] What happens to the total stranded power if systems become more en-
ergy proportional (assume workloads similar to that of Figure 6.4 ) ?
Search WWH ::




Custom Search