The trend to e-business and business on demand is enabled
by new distributed and highly flexible IT infrastructure approaches, such
as
| • |
Grid computing, |
| • |
storage area networks, |
| • |
peer-to-peer file sharing and |
| • |
autonomous computing. |
A key challenge with these new approaches is to control the provisioning and
consumption of resources such as bandwidth, storage and processor cycles in
an optimal manner. Critical overload situations as well as over-provisioning
should be avoided (see Figure 1).
The IBM Zurich Research Laboratory has developed solutions for resource
usage profiling and prediction for application in the areas of distributed
storage, network management and control, Grid computing, server farms
and database optimization (see Figure 2).The profiling solutions are based
on a number of research results developed at the IBM Zurich Research Laboratory.
They range from advanced sampling techniques to new analysis methods for
time series data. The benefits of the new approach for resource usage
profiling and prediction comprises the optimization of data migration
and replication, load balancing, proactive traffic engineering, admission
and fault control as well as accounting and pricing.
In a specific network profiling project, a network profiling engine (Aurora
engine) was developed. The Aurora engine is able to generate various bandwidth
usage reports regarding traffic on high-speed WAN links with a large number
of involved flows. In various trials, the Aurora engine proved to help
identify and resolve real network problems (see Figure 3).
|
|
