Model-based Performance Evaluation of Large-Scale Smart Metering Architectures Johannes Kroß, Andreas Brunnert, Christian Prehofer fortiss GmbH Guerickestr. 25 80805 Munich, Germany {kross,brunnert,prehofer} @fortiss.org Thomas A. Runkler Siemens AG Corporate Technology Otto-Hahn-Ring 6 81739 Munich, Germany thomas.runkler@siemens.com Helmut Krcmar Chair for Information Systems Technische Universität München Boltzmannstr. 3 85748 Garching, Germany krcmar@in.tum.de ABSTRACT Smart meter devices are used to monitor and control energy consumption and are interlinked with smart grids. Their growing use leads to an extensive amount of available data to be processed and causes smart grids to evolve to large-scale systems of systems. Guaranteeing appropriate scalability and performance characteristics is a tremendous challenge. In this paper, we focus on the provisioning of sufficient com- puting capacity to efficiently analyze the produced data in such a distributed system. For this purpose, we show the use of performance models to plan and simulate this distributed computation in smart grid systems. It demonstrates how different system architectures can be evaluated and required capacities can be estimated to cope with the occurring data volume. We analyze response times for time-critical tasks and assess the scalability of smart grid systems. Categories and Subject Descriptors C.4 [Performance of Systems]: Modeling techniques Keywords Smart Meter, Smart Grid, Advanced Metering Infrastructure, Performance, Evaluation 1. INTRODUCTION Smart meter devices are replacing conventional energy meters in several countries and form the basis to manage and monitor energy consumption [10]. These devices are in- terlinked as part of smart grids and allow two-way communi- cation via interfaces so data can be automatically exchanged between smart meters and energy management operators. A smart grid connects smart energy devices such as smart me- ters to a distributed energy delivery network that allows for automatic communication and management of devices [10]. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full cita- tion on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or re- publish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org. LT’15, February 1, 2015, Austin, Texas, USA. Copyright c  2015 ACM 978-1-4503-3337-5/15/02 ...$15.00. http://dx.doi.org/10.1145/2693182.2693184. In order to built up a smart grid, an advanced metering infrastructure (AMI) is necessary to manage smart meter devices. It is responsible for connecting distributed smart meters and storing their data [11]. In realistic scenarios, often more than hundred thousand devices are managed by a few central systems and data is continuously exchanged. Hence, the produced data volume is enormous and can easily cause performance issues. The responsibility of these systems (called smart grids) is to store the data produced by smart meter devices and manage them. An additional time-critical task is to calculate optimized en- ergy plans based on the collected data. Therefore, they must combine analytic capabilities with real-time processing [3]. As we consider here large distributed systems, the distribu- tion of this processing is important and has not been consid- ered before. Since the introduction of smart meter devices is growing in many countries, these systems must also be able to scale-up to continuously reach and comply with their per- formance goals. Performance models provide a common way to mirror sys- tems and simulate their behavior to guarantee such non- functional requirements [2]. They allow for predicting and measuring performance metrics such as throughput, response time and resource utilization. Performance models can be used for capacity planning as well as to answer sizing ques- tions. They also enable developers to examine design al- ternatives of architectures and find optimized system con- figurations. By being able to simulate different workloads on such models, they also support to evaluate a system’s scalability. This paper shows how performance models can be used to model large distributed smart grid systems and simu- late hundreds of thousands connected smart meters. While prior work on smart grid performance has mainly focused on the networking aspects [7, 5, 9], we focus here on the com- putation required for the analysis of the data in large dis- tributed systems. We develop two prototype models present- ing two different use cases. For each model, we implement two infrastructure approaches and simulate them with vary- ing amount of smart meter devices. Although both models are kept as simple as possible, they involve the specifica- tion of multiple parameters and allow us to already address common problems in the smart grid context. Therefore, we prioritize to analyze the performance metrics utilization, throughput and scalability in this paper. 9