A Hybrid Approach for Optimizing Carbon Footprint in Intercloud Environment
Abstract: This paper focuses on the problem of workload placement in an Intercloud with the view of minimizing the carbon footprint of such a computing environment. In order to reduce the ecological impact of the data center Greenhouse Gas (GhG) emissions, this paper addresses the problem as a whole, by proposing a global mathematical formulation, based on the joint optimization of the Virtual Machine (VM) placement and their related traffics, along with a workload consolidation method and a cooling maximization technique that considers the dynamic behavior of the cooling fans. As the Virtual Machine Placement Problem (VMPP) is classified as an NP-hard Problem, with the addition of the traffic embedding, the problem becomes more complex and stays NP-hard. Therefore, we propose a Hybrid approach, for solving such problem and find good feasible solutions in a polynomial time. The results obtained from comparing With the exact method and other reference approaches help in assessing the efficiency of the proposed algorithm, as the carbon Footprint costs are relatively close to the lower bound, with an average gap of about 3%, and found within a reasonable amount of time.
Existing system: The aforesaid challenges have been tackled in our previous works, where carbon footprint optimization has been performed in an Intercloud setup while considering a workload defined by standalone VMs. However, with the rising number of applications spanning multiple VMs and making use of the switches and links to route the inter-VMs traffic demands, the network power consumption and their associated carbon footprint can no longer be neglected. As the latter accounts for about 25% of the equipment power consumption, efforts have been deployed in order to minimize the network consumption. However, optimizing the VM placement or the traffic routing alone is too restrictive, as VM consolidation may introduce network congestion, and traffic routing may result in servers’ resource wastage. Having a control on both entities will help efficiently optimize data centers’ resources and reduce IT total power consumption. Proposed system: Security or redundancy issues, which may impact the Virtual Machine (VM) placement, have been ignored. Also, due to the rapid growth in the number and use of large data centers and the increasing energy cost, various workload-aware consolidation techniques have been proposed in order to tackle the servers’ power minimization problem, under defined performance constraints. Approaches for improving cooling cost have also been developed in order to further reduce data centers’ power consumption. However, the dynamic behavior of IT equipment fans which influences the cooling efficiency at high temperatures has often been neglected. Furthermore, clouds’ high energy consumption does not only affect providers’ profits, but highly impacts climate change, as data centers’ carbon footprint represents 2% of the total Greenhouse Gas (GhG) emissions , and is Increasing. Advantages: The same instances are also tested under three references models: VMOPT (B model), NETOPT (C model) and EFFOPT (D model). VMOPT optimizes the VM placement in order to reduce their carbon footprint; NETOPT minimizes the network consumption, while EFFOPT optimizes the data centers’ power usage effectiveness.
Our purpose is to demonstrate the advantages of jointly optimizing VM placement and traffic routing, in the view of minimizing data centers’ power consumption and equivalent carbon footprint. The total (TO), the chassis and servers (VM) and the network (NE) carbon footprint costs (in tons). Disadvantages: These results help assessing the efficiency of the proposed model, as it considers a joint optimization of the computing nodes power consumption, including their dynamic behavior, the network resources dissipation and the cooling efficiency. This model provides a more accurate evaluation of the carbon footprint cost, allowing one to perform an efficient VM placement, however at the expense of a longer average execution time, as illustrated in Fig.6, due to the high complexity of the problem. Modules: GCBF: ITS GCBF and Lower Bound The proposed ITS GCBF heuristic is executed on 20 small instances and the results are compared with the optimal solution, OPT GCBF , computed using a linear program implemented in AMPL/CPLEX. The resulted lower bound enables to assess the performance of the proposed resolution method. The results are presented in Table 3, where column 2 and 3 give respectively the problem sizes and the carbon footprint cost calculated with the exact method. Regarding ITS GCBF , the average carbon footprint and the cost gaps (minimal, mean and maximal) with respect to the lower bound values are shown in columns 4 to 7. The total number of iterations as well as the iteration at which the near optimal solution is found are also presented in the last two columns. Table 3 demonstrates that, in general, the carbon footprint costs obtained from the implementation of ITS GCBF are very close to the lower bound values. Cloud computing: CLOUD Computing has recently emerged as the new trend in the Information and Communication Technology (ICT) sector, allowing end users to run their applications, encapsulated in Virtual Machines (VMs), in a secure computing
environment, while ensuring proper performance isolation among co-located VMs. However, security or redundancy issues, which may impact the Virtual Machine (VM) placement, have been ignored. Also, due to the rapid growth in the number and use of large data centers and the increasing energy cost, various workloadaware consolidation techniques have been proposed in order to tackle the servers’ power minimization problem, under defined performance constraints. Approaches for improving cooling cost have also been developed in order to further reduce data centers’ power consumption. However, the dynamic behavior of IT equipment fans which influences the cooling efficiency at high temperatures has often been neglected. Virtual Machine Placement Problem: Therefore, in the context of a more general workload placement, which includes complex applications involving inter-VMs traffic demands, reducing the carbon footprint of an Intercloud environment becomes even more challenging as it remains unclear how to jointly consider techniques such as data center selection, smart VM consolidation and cooling efficiency optimization methods, introduced in , with intelligent traffic demand routing approaches, which select the best power-efficient switches while preventing network congestion. In this paper, we are therefore interested in a global approach to the Virtual Machine Placement Problem (VMPP). Also, as many aspects of the VMPP refer to well-known NPhard problems, such as the Knapsack or an enhanced version of the Bin-Packing Problem (BPP) , the XDimensional Vector Packing Problem (X-DVPP) , exact algorithms are unsuitable to solve moderate and large instances of the problem, as in common optimization solvers. Client and workload representation: The applications, submitted by the clients, are encapsulated in VMs, characterized by a multidimensional resource vector consisting of the following components: [CPU, disk, memory]. The workload can be of two types: stand-alone VMs or applications spanning multiple VMs and with inter- VMs traffic demands. Due to technical constraints, for redundancy or security purpose, interference constraints require that certain VMs are not co-located and affinity coefficients are derived accordingly. Clients may also impose location constraints, in terms of potential
data centers to host the applications. Moreover, the VM placement constraints related to performance degradation, which were introduced in and adapted in, are also considered in this paper.