Revisiting Heterogeneous Storage Optimization in the Vector-Sum Model
نویسندگان
چکیده
Large-scale data centers often adopt more than one type of storage device, each with different storage capacity, I/O capability, and cost. Optimizing the performance-to-cost efficiency of such heterogeneous storage systems is of great practical importance (CapEx), and it is a classic problem in computer system design. The Vector-Sum Model (VSM) is a mental model widely-used by system administrators for this task, due to its conceptual simplicity. The model encompasses various commonly-used rules-of-thumb, such as the five-minute rule or various Knapsack-based heuristics. In this paper we revisit the vector-sum model and study heterogeneous storage using a new form of optimization diagrams. These diagrams give raise to a near-optimal solution to the problem, which subsumes the existing rules-of-thumb used in practice. Our solution also explains that these heuristics are indeed optimal under their respective assumptions, while they become sub-optimal in more general cases. Specifically, our analysis implies that the recent adoption of SSD in data centers may challenge the quality of these commonly-used heuristics, and that our new optimization approach can sustain data center-scale workloads at lower total purchasing cost. Finally, we show that, although the commonly-used I/O metrics of storage are non-additive, we can use regression techniques to transform the metric into an additive form. Experiments using web search production workloads show that the Vector-Sum Model becomes more accurate after the metric transformation.
منابع مشابه
Backhaul-Aware Decoupled Uplink and Downlink User Association, Subcarrier Allocation, and Power Control in FiWi HetNets
Decoupling the uplink and downlink user association improves the throughput of heterogeneous networks (HetNets) and balances the traffic load of macro- and small- base stations. Recently, fiber-wireless HetNets (FiWi-HetNets) have been considered as viable solutions for access networks. To improve the accuracy of user association and resource allocation algorithms in FiWi-HetNets, the capacity ...
متن کاملA Robust Optimization Approach for a p-Hub Covering Problem with Production Facilities, Time Horizons and Transporter
Hub location-allocation problems are currently a subject of keen interest in the research community. However, when this issue is considered in practice, significant difficulties such as traffic, commodity transportation and telecommunication tend to be overlooked. In this paper, a novel robust mathematical model for a p-hub covering problem, which tackles the intrinsic uncertainty of some param...
متن کاملAn Integrated Model for Storage Location Assignment and Storage/Retrieval Scheduling in AS/RS system
An integrated optimization framework, including location assignment under grouping class-based storage policy and schedule of dual shuttle cranes, is offered by presenting a new optimization programming model. The objective functions, which are considered at this level, are the minimization of total costs and energy consumption. Scheduling of dual shuttle cranes among specified locations, which...
متن کاملDeveloping a Model of Heterogeneity in Driver’s Behavior
Intelligent Driver Model (IDM) is a well-known microscopic model of traffic flow within the traffic engineering societies. While it is a powerful technique for modeling traffic flows, the Intelligent Driver Model lacks the potential of accommodating the notion of drivers’ heterogeneous behavior whenever they are on roads. Concerning the above mentioned, this paper takes the lane to recognize th...
متن کاملSustainable Supplier Selection by a New Hybrid Support Vector-model based on the Cuckoo Optimization Algorithm
For assessing and selecting sustainable suppliers, this study considers a triple-bottom-line approach, including profit, people and planet, and regards business operations, environmental effects along with social responsibilities of the suppliers. Diverse metrics are acquainted with measure execution in these three issues. This study builds up a new hybrid intelligent model, namely COA-LS-SVM, ...
متن کامل