Provisioning Quality-Aware Social Compute Units in the Cloud

Abstract

To date, on-demand provisioning models of human-based services in the cloud are mainly used to deal with simple human tasks solvable by individual compute units (ICU). In this paper, we propose a framework allowing the provisioning of a group of people as an execution service unit, a so-called Social Compute Unit (SCU), by utilizing clouds of ICUs. Our model allows service consumers to specify quality requirements, which contain constraints and objectives with respect to skills, connectedness, response time, and cost. We propose a solution model for tackling the problem in quality-aware SCUs provisioning and employ some metaheuristic techniques to solve the problem. A prototype of the framework is implemented, and experiments using data from simulated clouds and consumers are conducted to evaluate the model.

A Collaboration Model for Community-Based Software Development with Social Machines

“…Concretely, we build upon the particular SCU model presented in [11] and use it in the context of the encompassing social machine's coordination model. Whenever a development primitive (from Section II-A) needs to be executed, a request with input parameters is sent to the SCU provisioning engine to form a team of developers/experts suitable for that particular primitive (task).…”

“…The full list of available input parameters and descriptions of the team formation algorithms are available in [11]. In this paper, however, we vary only the parameter named job description set (J), while assuming default values for the remaining parameters.…”

“…Therefore, we need a set of metrics to express the requested and measure the obtained quality of this service. In this paper, we adopt the metrics already provided by the SCU [11] to formulate requested QoS.…”

“…2) A feature tree representing the current state of the software, as defined in Section III, following the example in Figure 4; 3) A simplified, idealised SCU model, where teams are formed in response to quality constraints, and perform tasks on the feature tree. In this simplified model, we assume that one worker is returned per task, with a skill set that exactly matches the quality constraints [11]. Workers have scores for four skills: implementation, evaluation, prioritisation, design, each of which ranges from 0..1.…”

Programming Model Elements for Hybrid Collaborative Adaptive Systems

“…Discovery can be either performed by actively picking members [8], or by coordinating the process of self-formation of the collective as integral part of the composition and negotiation phases. In the latter case, the OM uses a static decision tree for scheduling the messages of subscription, agreement and withdrawal to the proposed plans originating from human peers [9].…”