Virtualisation as a lever on consumption
Why virtualising cuts consumption: utilisation, idle power, consolidation, ghost servers, pooling, PUE, and the limits of virtualisation (sprawl, oversizing, rebound).
- 28 cards
- 5 topics
- Free
The outline, topic by topic
Each topic is weighted the way the exam weighs it.
- Concepts
- Questions
- Sort and link
- The numbers
- Scenarios
A few cards from the deck
Server utilisation: what is it?
Show the answer
The share of its computing capacity that a server actually uses, on average, for the work asked of it.
A lightly loaded server still consumes a lot: energy spent is far from proportional to the work done, hence the point of loading it more.
A server draws 100 W at idle and 200 W at full load. How much does it draw at 10% load if power rises linearly with load?
- 110 W
- 20 W
- 10 W
- 100 W
Show the answer
110 W
Power = idle + (full − idle) × load = 100 + 100 × 0.1 = 110 W. At 10% load the server still draws more than half of its maximum power.
A lightly loaded server already draws a large share of its maximum power: grouping the load makes it possible to switch off the … machines.
Show the answer
A lightly loaded server already draws a large share of its maximum power: grouping the load makes it possible to switch off the unneeded machines.
Idle power is the fixed part, independent of work. That is what is saved by switching off, more than the variable part.
Server consolidation: what is it?
Show the answer
Gathering several workloads onto fewer physical machines, so that the rest can be switched off or never bought.
The gain comes from idle power avoided: each server switched off removes its idle consumption and its future manufacture.