Id uniform rvs in 0 1 10 for all the companies

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: cription may be unknown) can be described by their average rather precisely. M. Chen (IE@CUHK) ENGG2430C lecture 10 17 / 29 Insights from the Law of Large Number It explains the basic intuition principle behind cloud computing. Example: 100 companies, each needs to satisfy its own random demand (X1 to X100 ). Start simple to bring out the intuition: assume they are i.i.d. uniform r.v.s in {0, 1, . . . , 10}. For all the companies to satisfy their demand with 90% probability, each company needs to purchase 9 servers. Thus 900 servers in total. If a cloud company to satisfy the demands (via virtualization) with the same guarantee, we need a bit more than 500 servers. Intuition: not all demands peak at the same time. The same intuition still applies when X1 to X100 follow (different) general distributions and not fully independent. The gap between the server numbers in the two scenarios is the “multiplexing gain”. Thus all companies can pay less but still satisfy the demands. The cloud computing service provider can explore the multiplexing gain to ob...
View Full Document

Ask a homework question - tutors are online