Late last year the TSSG Emerging Networks Laboratory (ENL) was invited by HEAnet to carry out some short-term research on the FP7 Mantychore project. As part of the Mantychore project, HEAnet, in conjunction with their partner institutions (I2CAT, NORDUnet, UNI-C, U. Essex, TID, TCD) are undertaking the development of a Network as a Service platform, which has been released under an open source license as OpenNaaS (http://www.opennass.org). The aim of this platform is to allow virtual connections be created dynamically across multiple infrastructure providers. At the same time, the project was very interested in how energy and sustainability considerations could be factored into their platform. Hence HEAnet contacted ENL in relation to some of the research undertaken within the group in the area of Green Data-Centres.
ENL’s contribution to the project consisted of a number of specific work-packages. First, we investigated potential energy efficiency/green metrics that could be incorporated directly into the OpenNaaS resource model. The aim is that these metrics could be used for monitoring and reporting of the network power usage, efficiency and sustainability. Also, the addition of these metrics will allow the intelligent creation of dynamic circuits across multiple infrastructures, through matching of users green requirements to the energy properties of the individual infrastructures. Effectively, this would allow the user to specify the importance of energy considerations when creating a virtual pipe, which would then be used to negotiate with individual providers to determine the best path for the creation of the pipe. Further to this, there is the potential that, in the future, routing decisions could be made using this data in order to determine the greenest paths through the network.
In line with this, we developed a Green Information Model for capturing the energy metrics relevant to networks. As Mantychore hopes to look at a wider IaaS solution in the future, the model was designed in an abstract way, such that it could be applied to a range of different areas. We investigated how best to capture the energy usage of network equipment in a flexible manner, given the variance of physical configurations that infrastructure providers may wish to cater for. For instance, a device’s energy can be measured in a number of ways (e.g. by the device itself or by other power delivery equipment), and it is important that the model is capable of capturing these differing cases. The model considered not just metrics for measuring energy usage, but also measures of sustainability such as renewable energy sources. This key relationship is captured in the association of power consuming (e.g routers), power delivery (e.g. PDUs) and power supply (e.g. wind generators) resources. Based on this model, we performed an analysis of the existing OpenNaaS architecture and an integration plan was developed to incorporate the relevant green metrics into the OpenNaaS platform. Furthermore, the ENL work in the project also developed an SNMP-based PDU driver and simple test implementation based on the designed model, facilitating the capture of a devices energy usage and it’s energy source statistics.
Another important aspect of this work was to use OpenNaaS to facilitate greener data-centres through the dynamic migration of virtual machines between data-centres. A GA-based (genetic algorithm) optimization approach was used to determine the optimal placement of VMs, considering a number of parameters, such as renewable energy levels available in the data-centres, as well as hosting costs and migration distance. The evaluation was carried out through a set of simulated experiments that demonstrated, using real energy data from Mantychore nodes, that increased renewable energy usage could be achieved.