Project Overview:
Scheduling Algorithms for Real-Time Systems
Scheduling real-time transactions is far more complex than traditional
real-time scheduling in the sense that
1) worst case execution times are typically hard to estimate, and
2) certain aspects of concurrency control may not integrate well with
real-time scheduling.
To improve the timeliness,
we have started to investigate scheduling and concurrency control methods
that allow database systems to meet timing constraints
without reducing the concurrency level of the system.
We observed that by using a priority-dependent synchronization protocol,
the serialization order
of active transactions can be adjusted dynamically.
This makes it possible for transactions
with higher priorities to be executed first, so that high priority
transactions are not blocked by uncommitted low priority transactions,
while lower priority transactions may not have to be aborted even in face
of conflicting operations.
The adjustment of the serialization order can be viewed as a mechanism
to support time-critical scheduling.
Our effort has resulted in a number of hybrid protocols.
Preliminary performance results show that these protocols
can be very effective
in improving timeliness of real-time database systems.
Currently, we are extending the notion of dynamic adjustment of
serialization order to apply to distributed scheduling and
semantic-based scheduling algorithms.
We believe that the extension would produce many useful contributions,
since it processes transactions in an optimistic fashion and does not
need to support the two-phaseness of locking.
Other Projects:
[Starbase]
[Security]
[Predict]
[Schedule]
[Multimedia]
[Dependability]
Last modified: Mon Oct 2 18:34:05 EDT 1995 [SK]