Adaptive TrimTree: Green Data Center Networks through Resource Consolidation, Selective Connectedness and Energy Proportional Computing

Abstract

Abstract:A data center is a facility with a group of networked servers used by an organization for storage, management and dissemination of its data. The increase in data center energy consumption over the past several years is staggering, therefore efforts are being initiated to achieve energy efficiency of various components of data centers. One of the main reasons data centers have high energy inefficiency is largely due to the fact that most organizations run their data centers at full capacity 24/7. This results into a number of servers and switches being underutilized or even unutilized, yet working and consuming electricity around the clock. In this paper, we present Adaptive TrimTree; a mechanism that employs a combination of resource consolidation, selective connectedness and energy proportional computing for optimizing energy consumption in a Data Center Network (DCN). Adaptive TrimTree adopts a simple traffic-and-topology-based heuristic to find a minimum power network subset called 'active network subset' that satisfies the existing network traffic conditions while switching off the residual unused network components. A 'passive network subset' is also identified for redundancy which consists of links and switches that can be required in future and this subset is toggled to sleep state. An energy proportional computing technique is applied to the active network subset for adapting link data rates to workload thus maximizing energy optimization. We have compared our proposed mechanism with fat-tree topology and ElasticTree; a scheme based on resource consolidation. Our simulation results show that our mechanism saves 50%-70% more energy as compared to fat-tree and 19.6% as compared to ElasticTree, with minimal impact on packet loss percentage and delay. Additionally, our mechanism copes better with traffic anomalies and surges due to passive network provision.

Energy Conservation with Open Source Ad Blockers

“…Data servers use a substantial and growing amount of energy, and there is considerable effort to reduce that energy use [48][49][50] using both thermal management [51][52][53][54], electrical management [55], configuration [56], and smart systems [57]. The concept of green data centers [58,59] is growing. Facebook open sourced their designs [60], and there is even open hardware that could be used to help monitor and improve them [61].…”

Energy Aware Virtual Machine Scheduling in Data Centers

“…Figure 16 shows the maximum memory access bandwidth for STREAM running inside each VM while running STREAM separately and running STREAM and PrimeSearch concurrently. The shorter STREAM calculation time (calculation time of single Copy, Add, Scale, or Triad are generally not longer than one minute and the total run time is generally 10 min) is significantly less than the PrimeSearch completion time (3)(4)(5)(6)(7)(8)(9)(10)(11)(12) VMs is about 1620 sec, 24 VMs is 2980 sec). Even in the concurrent running of VMs with STREAM and PrimeSearch, the server power consumption remains significantly less than the sum of the power consumption of two servers running separately.…”

“…However, for small-and medium-sized data centers, they are not always in a high-load operating state and thus the servers still consume a lot of energy when they are idle. These small-and medium-sized data centers account for more than 95% percentage of total installed servers [3,8]. In such smaller scale data centers, server utilization is much lower than large scale data centers and their power use effectiveness (PUE) is also lower.…”