Scheduling in data management systems plays a very important role,
particularly in periods of high demand or in the presence of diverging
user preferences, which are often described as Service Level
Agreements (SLAs). In this talk, I will describe our work in two
different areas: (a) scheduling of web transactions in the presence of
deadlines and (b) scheduling in data stream management systems.
With regards to web transaction scheduling, I will present ASETS, a
parameter-free adaptive scheduling algorithm that automatically adapts
to system load and to web transaction characteristics (i.e.,
interdependencies, deadlines, and priorities). ASETS prioritizes the
execution of web transactions with the objective of minimizing
weighted tardiness.
With regards to scheduling in data streams, I will summarize our early
work, which can be used to optimize different Quality of Service
metrics during the processing of multiple heterogeneous Continuous
Queries (CQs). I will then describe our proposal for class-based
scheduling of CQs and present an adaptive, synergistic load-manager
and scheduler.