Test instances for paper: "Adaptive large neighborhood search heuristics for multi-tier service deployment problems in clouds"

Published: 14-03-2017| Version 1 | DOI: 10.17632/hnf26mxdvk.1
Contributor:
Anders Gullhav

Description

INTRODUCTION Each test instance are built based on the data in two separate .txt files: - one file specifying the major part of the instance, denoted the "data file", and - one file specifying the replication patterns for all the services, denoted the "reppat file". NOTE: ALL DATA ARE ARTIFICIAL. NAMING CONVENTIONS The data files are named according to the following convention: D_sXtY_nZ_c30ai_1reppat.txt, where X refers to the number of services, and the Y refers to the different seeds (for the random number generator) used to construct the file. If no Y is given, Y = 1. The different test instances with an equal number of services (i.e., equal X) are in the paper distinguished by a letter 'a', 'b', 'c', 'd' or 'e'. These letters correspond to Y = 1, 2, 3, 4 or 5, respectively. The files with Y > 5 are used in the tuning of the ALNS. The Z are used to indicate the number of nodes available for service deployment. For the hybrid cloud cases the Z are not directly a number, but either '075xBB' or '09xBB', which means that the number of nodes are set equal to 0.75 or 0.9 multiplied with the best bound of the corresponding private cloud case (which minimizes the number of nodes used in the objective function). The reppat files are named according to the following convention: sXtY_reppat.txt, where X and Y have the same interpretation as above. DATA FILE CONTENTS The data file specifies the: - number of services (nServices, |S|) - number of components per service (|Q_i|) - number of resources (|G|) - resource demand for each active replica of each component of each service (G^A) - resource demand for each passive replica of each component of each service (G^P) - node resource capacity (B) - maximum number of passive replicas per node (E) - maximum number of different services on a node (D) - maximum number of nodes (|N|) - cost of deploying an active or passive replica of each component of each service in the public cloud (C) In addition, the data file specifies one replication pattern for each service, but this information is not used the computational study. Instead the replication patterns are read in from the reppat data files, as specfied below. The data in the file are organised in matrices with the elements separated by space. The comments in the files explain the organization of the matrices. The service and resource indices start at 0. REPPAT FILE CONTENTS The reppat file specifies the replication patterns as comma-separated lines, one replication pattern per line. The comma-separated line should be interpreted as follows: service index, replication pattern index, availability measure, response time measure. number of active replicas of component 1, number of active replicas of component 2, ..., number of active replicas of component |Q_i|, number of passive replicas of component 1, number of passive replicas of component 2, ..., number of passive replicas of component |Q_i| The service and replication pattern indices start at 1.

Files