Project Overview:
Predictable Real-Time Transaction Processing


Predictable transaction processing in real-time database systems requires different data models and correctness criteria. Most of real-time database scheduling algorithms have been developed and evaluated under almost the same workload and operating environment model as used in conventional database systems. While this model is suited to some real-time database applications with soft or firm deadlines, typical hard real-time database applications do not fit into this model, since they have different characteristics and semantics on their data and transactions. Hard real-time constraints must have a guarantee that their deadlines will always be met.

To make such a strong guarantee, we cannot simply use the best-effort scheduling protocols. We must have scheduling protocols which can control the behavior of real-time database systems to guarantee the timeliness. To do this, real-time database systems must have advance knowledge of the resource and data requirements of transactions. Important research issues include how to eliminate sources of unpredictability, and how to handle interactions between hard and soft real-time transactions.





Other Projects:

[Starbase] [Security] [Predict] [Schedule] [Multimedia] [Dependability]


Last modified: Mon Oct 2 18:34:05 EDT 1995 [SK]