The Océano project

The Océano project is designing and developing a pilot prototype of a
scaleable, manageable infrastructure for a large scale "computing utility
powerplant" that enables multi-customer hosting on a virtualized
collection of hardware resources. A computing utility infrastructure
consists of a "farm" of massively parallel, densely-packaged servers
interconnected by high-speed, switched LANs. This project aims to address
many of the open technical issues in these powerplant environments. Hosted
customers increasingly require support for peak loads that are orders of
magnitude larger than what they experience in their normal steady state.
Thus, a hosting environment needs a faster turnaround time in adjusting the
resources (bandwidth, servers, and storage), assigned to each customer to
the dynamically fluctuating workload. The "colocation" hosting model uses
dedicated infrastructure and servers for each customer, typically in
physical cages. In this model, enabling peak-load scale on demand requires
large investments in standby, non-shared resources, which would be mostly
underutilized and would occupy large amounts of physical space. Clearly,
the colocation model is not suited to efficiently mitigate the differences
between average and peak load. Océano introduces high levels of automation
to dynamically adjust web sites to actual traffic demands over a massively
parallel array of shared and distributed Linux servers

The objectives of the Océano project include:

• Implement an infrastructure that enables large numbers of hosted
  customers over Linux servers
• Reduce the costs of setting up and operating the hosting farms by automation
• Dynamically assign resources to accommodate planned and unplanned fluctuation of workloads
• Offer a wide variety of services levels to customers
• Secure sharing of resources across multiple customers
• Provide adequate reliability through massive redundancy, and automated re-provisioning