Concurrency Control in Multilevel Relational Databases - Multilevel Security for Relational Databases - page 146

7.6 Summary
In this chapter, the secure multiversion concurrency control model
based on the Rajwinder Singh model was implemented to solve
the problem of covert channel, retrieval anomaly, and starvation of
high security level transactions by maintaining multiple data ver-
sions. The model divided W −set T j into two parts— W −set done T j and
W −set remaining T j —and used an infinite timestamp instead of using
t-lock to prevent the creation of retrieval anomaly problems and to
ensure serializability of transaction. These modifications reduced the
conflict area between high- and low-level transactions and reduced
the blocking time of high-level transactions, resulting in improve-
ment of their response time. Also, this chapter implements a working
multilevel secure database prototype in a Microsoft SQL server data-
base to measure the performance experiments that were instrumented
using the system. Additionally, this chapter investigated the perfor-
mance of the proposed secure multiversion concurrency control model
by varying the transaction arrival rate. Finally, the correctness of a
secure multiversion concurrency control model by using the criteria of
security, serializability, fairness, selection of correct data version, and
fast version selection was proven.