The SLEDS project is about storage performance management—allowing customers to
make the best use of the performance available in their storage systems. Basic performance
management tracks the use of system resources. SLEDS goes further by allowing a
system administrator to measure and control the Quality of Service (QoS) offered by a storage system to each application host and to each application. With the technology developed in SLEDS, an administrator can do the following:
Guarantee and document that I/O performance meets service level agreements (SLAs)in a service provider environment
Ensure that heavy use by one host application does not swamp the system
resources and degrade the performance of other applications
Apply differentiated service for latency-sensitive and bandwidth-intensive applications
Identify any bottleneck resources that are limiting storage performance
SLEDS applies instrumentation to the I/O-path connections between applications and their
storage, typically in a Storage Area Network (SAN). The instrumentation is used for
monitoring and control. A central application communicates with the instrumentation, analyzes
the data and applies user-supplied policies. This central application is a single point for
administering QoS for the entire storage installation. The research phase of SLEDS has been completed.
As was
presented
at the
Symposium on Reliable Distributed Systems
in 2003 and referenced to the right,
the SLEDS project created a robust prototype and evaluated its effectiveness in mediating
performance resource use between competing workloads.
In addition, an extensive web-based management application has been created that integrates
SLEDS monitoring and control into a facility for managing system I/O performance
| 
