Real-time database systems are designed to handle workloads where tran
sactions have completion deadlines and the goal is to meet these deadl
ines. However, many real-time database environments are characterized
by workloads that are a mix of real-time and standard (non-real-time)
transactions. Unfortunately, the system policies used to meet the perf
ormance goals of real-time transactions often work poorly for standard
transactions. In particular, optimistic concurrency control algorithm
s are recommended for real-time transactions, whereas locking-based pr
otocols are suited for standard transactions. In this paper, we presen
t a new database system architecture in which realtime transactions us
e optimistic concurrency control and, simultaneously, standard transac
tions use locking. We prove that our architecture maintains data integ
rity and show, through a simulation study, that it provides significan
tly improved performance for the standard transactions without diminis
hing the real-time transaction performance. We also show, more general
ly, that the proposed architecture correctly supports the co-existence
of any group of concurrency control algorithms that adhere to a stand
ard interface.