Fault diversity among off-the-shelf SQL database servers

Abstract

Abstract

Reliability Modeling of a 1-Out-Of-2 System: Research with Diverse Off-The-Shelf SQL Database Servers

“…Also we have observed in earlier research [4], [3] that AB-faults can fail in different ways in the two DBMS products, and hence can be detected (and potentially corrected [16]). As a result, the estimates that we get using our models for the reliability benefits of diversity will most probably be underestimates: the true benefits may be higher.…”

“…We have fully described these studies and provided analysis of the results in [4] and [3]. We will be utilizing the results of those studies in this paper as empirical evidence with one of the models, as well as for verification of assumptions.…”

“…With regard to the dependability of a fault tolerant DBMS, we have reported previously on a study with the publicly available fault reports of four OTS DBMS products (both open-source and closed development) [4] and later releases of two of them [3]. We found that a high number of these faults would not be tolerated (or even detected) by the existing non-diverse faulttolerant schemes but did not cause failures in any two diverse DBMS products.…”

“…In this section we will use the results of our previous studies with the bugs [4], [3] (which we summarized in section 2.1) to derive empirical estimates of β. The results of the first study from running the DBMS product faults that could be run on each pair and the failures that they cause are given in Table 1: -n A are faults reported in product A -n AB are product A reported faults that also affect B -n B are faults reported in product B -n BA are product B reported faults that also affect A.…”

“…we assume the fault will always cause a failure in the DBMS product for which it was reported (but not, unless it does cause a failure, for other products) even 4 Terms introduced by Gray [8], defining two types of bugs:…”