Krisztián Fekete

Analyzing computation offloading energy-efficiency measurements

No one disputes that mobile phones have become part of everyday life. Besides phone calling we use them for browsing, messaging, playing games and many other things which were possible only on a desktop computer several years ago. Although these devices are even more “smart”, but due to the growing resource requirements their batteries are discharging in a very short period. This is a very important phenomenon, which forces to find alternative ways to reduce the energy consumption of the smartphones. One of them is the method of “computation offloading” where a part of the processes are executed on a remote device (e.g. in the cloud). The creation of an energy-efficiency model consists of several steps. One of the hardest part is to understand how a smartphone behaves in different circumstances. During the execution of an application every step has its own energy cost. To make the execution energy-efficiency, we need first to measure and analyse these costs. In this paper we are going to present a measurement system which is used to analyse the energy consumption of smartphones. The measurements were made in different scenarios and the goal was to save energy through offloading some tasks. The offloading process is based on scheduling theory.

Automatic NoSQL Schema Development: A Case Study

Cloud service providers offer a huge variety of schema-less NoSQL data storage solutions. The flexibility of these data stores offer greater freedom in structuring the data than relational databases. However, it would be desirable to make use of the strong mathematical background of relational data structures. In this paper, we introduce an automatic NoSQL schema optimization that uses a normalized data schema as starting point. We analyze the predefined set of queries, and compile the schema that can serve the queries with minimal cost at a certain query load. The introduced process is performed on a conceptual model of the database, and the queries are defined in Object Constraint Language to simplify the analysis. The optimization algorithm is introduced through a case study.

Energy-efficient computation offloading model for mobile phone environment

The popularity of smartphones is growing every day. Thanks to the more powerful hardware the applications can run more tasks and use broadband network connection, however there are several known issues. For example, under typical usage (messaging, browsing, and gaming) a smartphone can be discharged in one day. This makes the battery life one of the biggest problems of the mobile devices. That is a good motivation to find energy-efficient solutions. One of the possible methods is the “computation offloading” mechanism, which means that some of the tasks are uploaded to the cloud. In this paper we are going to present a new energy-efficient job scheduling model and a measurement infrastructure which is used to analyze the energy consumption of smartphones. Our results are going to be demonstrated through some scenarios where the goal is to save energy. The offloading task is based on LP and scheduling problems.

Graph partitioning algorithm for model transformation frameworks

Software modeling has become an everyday practice. Modeling extra-large models have enormous constraints: both memory and computational capacity of a single computer might be insufficient for handling model transformations. One solution to overcome this barrier is to extend the infrastructure. Cloud computing provides feasible realizations for these needs. However, existing algorithms have to be extended/modified to support cloud computing and use its advantages most efficiently. Generally, models can be easily mapped to graphs. This paper provides an algorithm for partitioning graphs representing models. Models can be mapped onto several computational instances and processed on these instances in a distributed fashion. Our algorithm is based on the heuristic Kernighan-Lin method, but we allow manually altering the number of partitions dynamically based on the actual needs. Moreover, we do not build on knowing the entire model when creating the partitions, since it would not fit into the memory of a single instance. Instead, model nodes are received and processed one by one. Our algorithm is fine tuned to these special conditions. The efficiency of the algorithm is illustrated by a case study.

Energy efficient code optimization in mobile environment

Nowadays the mobile devices energy consumption has become a very serious issue. Due to the fast growing mobile industry the current devices usually contain wireless (Wi-Fi, 3G, 4G and Bluetooth), GPS and other heavy network sensitive technologies. The cell battery manufacturers usually cannot keep the pace with this fast altering environment and demands, hence the devices get inappropriate battery. Recently many researchers deal with this topic trying to find out a reasonable solution for energy optimization without compromising the functionality of the devices. A possible approach to deal with the problem today and achieve the desired result is the code side optimization. The market based mobile application distribution model makes this challenge harder. Sometimes the code quality of the uploaded applications are poor and the market owners are not able to force the developers to write clear and energy efficient code, they can only give them recommendations. The question is coming up: how could we still encourage the individual developers and companies to write optimized and a bit more standardized code? The answer could be delivering them tools to facilitate the refactoring and code quality managing processes. One way we can deal with energy problem is to reorganize the heavy computational tasks out of the device to the cloud or to just another machine. This technique is called “offloading”. In this paper we are going to introduce a new code generation extension for software engineers which is aiming to automate the offload to web services. Those services can then be easily deployed to any desired location to ease the mobile devices computational task. We think this tool would be really helpful to the community and the industrial users also.

Route Separation Strategies for Human Movement Datasets

Learning patterns of human movement is a complex and hard task, including several computationally expensive algorithms. This issue has even higher emphasis in mobile environment, since handheld devices contain significantly less memory and computing power than a usual PC does. In this paper we are going to compare novel, mobile-optimized methods for separating trajectories in a human routine recognition framework.