Tuning of database audits to improve scheduled maintenance in communication systemsStefano Porcarelli1, Felicita Di Giandomenico2, Amine Chohra2*, Andrea Bondavalli3 1Univ. of Pisa, Computer Engineering Dep., Via Diotisalvi 2,I-56126, Pisa, Italystefano.porcarelli@guest.cnuce.cnr.it2 IEI/CNR, Via Moruzzi 1, I-56100, Pisa, Italy{digiandomenico, chohra}@iei.pi.cnr.it3 Univ. of Firenze, Dip. Sistemi e Informatica, V. Lombroso 6/17, I-50134, Firenze, Italyandrea.bondavalli@cnuce.cnr.itAbstract.To ensure the consistency of database subsystems involved in com-munication systems (e.g., telephone systems), appropriate scheduled mainte-nance policies are necessary. Audit operations, consisting in periodic checksand recovery actions, are typically employed in databases to cope with run timefaults which may affect the dependability and quality of service of the overallsystem. This paper aims at investigating on appropriate tuning of audit opera-tions, so as to find optimal balances between contrasting requirements, namelysatisfactory database availability and low overhead due to audits. For this pur-pose, a methodology to analyse the behaviour of the database under scheduledmaintenance is here suggested. Analytical models, essentially based on Deter-ministic and Stochastic Petri Nets (DSPN), are defined and analysed, in termsof dependability indicators. A sensitivity analysis wrt to the most affecting in-ternal and external parameters is also performed on a case study.1 IntroductionThe problem of protecting data used by applications during their execution, against run-time corruption, has long been recognised to be a critical aspect highly impacting on the reliability/availability of systems relying on such internal database. Communi-cation systems, such as telephone systems, are today-typical systems suffering from this problem, especially when a wireless environment is involved, which makes the data more prone to corruption. Indeed, these systems need to keep trace of resource usage status and of users data for correctly setting up and managing user calls. For this purpose, a database is included, where data are organised in such a way to capture the relationships existing among them. Data corruption may result in the delivery of a wrong service or in the unavailability of the service, with (possibly heavy) conse-quences on the quality of service perceived by users. Effective mechanisms to detect and recover from data corruption are then necessary; typically, audit operations are used, to perform periodic maintenance actions. Audits check and make the appropri-*On leave at IEI-CNR, supported by ERCIM (European Research Consortium for Informatics and Mathematics) under post-doctoral training program (Contract Nr: 99-04).