Capacity Management in IaaS Cloud

David Breitgand, IBM Research Haifa


Presentation slides

One of the promises of elastic cloud computing is relieving its customers from capacity planning by adding just the right amount of resources just in time when elastic applications need them. While realizing this vision might indeed exempt the customer from the complex and effort consuming capacity management task, the cloud provider still needs to execute on capacity planning to strike the right balance between SLA commitments and cost efficiency. The cost efficiency is intimately related to statistical multiplexing of workloads in the cloud, allowing over-committing cloud resources. Naturally, over-committing implies risk of resource congestion.

Therefore, there is a tradeoff between improving resource utilization by increasing an over-commit ratio and exposing the infrastructure provider and customers to the risk of resource congestion. In this talk I am going to explore a number of approaches to managing this trade-off.