Article Preview
Top1. Introduction
Scheduled outages occurs because of some planned maintenance activities on various services in a Service-Based Systems (SBS)(Tripathy & Naik, 2014). Planning scheduled maintenance on a regular basis is one of the way of avoiding or minimizing unplanned outages, as very often maintenance is done before it’s actually needed. During scheduled outages, the restoration plan is clear. That means, if maintenance is scheduled on some very important service in a SBS the switchover or rollback to backup/redundant service is decided beforehand(Dumitraş & Narasimhan, 2009). Although, the outages are scheduled and outage period during which some specific services will be unavailable is also planned(May et al., 2009). Safety-critical systems and business-critical systems cannot afford unavailability whether it is planned or unplanned. Until now, developers have had to use schedule outages manually for maintenance purposes. While this process is important to minimize unplanned breakdowns, it’s a rigorous and partial solution(Zhang et al., 2018). It cannot be able to completely accommodate and consider all possible future scenarios when it comes to scheduling outages in SBS. Therefore, some predictive expert Machine learning (ML) algorithm need to be designed.
Fault-Tolerance (FT) is the ability of system to continue its working uninterruptedly despite the failure of one or more services within SBS. During outage, fault tolerance can be used to provide continuity of services by transferring the control from down service to backup service instantly(Mansour & Dillon, 2011). FT is useful in giving the recovery mechanism for handling the scheduled outages. The scheduled outages within SBS can occur because of many complex factors such as connectivity loss between services, unavailable dependent or composable web services etc., Machine learning will make it possible to keep vital or critical services running efficiently and avoiding downtime that result in costly unavailability of services, loss of revenues, reputation damage and unsatisfied customers(Anees & Zeilinger, 2013). In this paper, Explanation Based Learning (EBL) has been used for learning -possible cases of scheduled outages and recovery mechanism based on service critical of web-services in SBS. The reason behind selecting EBL is that it is knowledge-intensive and knowledge driven technique. Also, EBL is capable of formulating a generalization by analyzing only a single example. Due to the limitation of suitable database for this purpose, rules has been made for scheduled outages and FT mechanisms.
The major contribution of this paper is predicting and providing early recovery mechanism for scheduled outages (SO). The proposed work, in this paper, is used as a preventive and predictive measure to avoid or minimize scheduled unavailability during scheduled maintenance. The proposed methodology gives an advanced indication that if scheduled outage occurs because of specific reason, it can be recovered through following particular FT mechanism. Rather than scheduling outages that will cause disruptive breakdowns and wreak havoc among the users during outage period, planners can handle almost all possible future scenarios of scheduled maintenance through generalized achieved from the analytical learning. Therefore, the major contribution of this research paper is as follows:
- 1.
Firstly, possible scheduled outages are identified in the context of Service-Based Systems (SBS).
- 2.
Secondly, “A Learning Model for Fault Tolerant Scheduled Outages (LMFTSO)” has been proposed. This model is further comprised of other two proposed models- “Scheduled outage learning model (SOLM)”, this is used in predicting SO and “Fault Tolerance Learning Model (FTLM)”, this is used in providing early recovery mechanism for SO.
- 3.
Lastly, the proposed learning model has been implemented with PROLOG and demonstrated with the case-study of a SOA based E-commerce system.
The remainder of the paper is organized as follows: section 2 presents the related work. Section3 and section 4 discusses SOs and its impact and SOs in the context of SBS respectively. Section 5 presents reasons for SOs to be FT. Section 6 presents the proposed LMFTSO using EBL. Section 7 extends the discussions and section 8 demonstrates the proposed methodology on Case study of E-commerce. Finally, the topic in concludes in Section 9.