Indirect cube: A power-efficient topology for compute clusters
نویسندگان
چکیده
Interconnection networks arranged as k-ary n-trees or spines are widely used to build highperformance computing clusters. Current blade-based technology allows the integration of the first level of the network together with the compute elements. The remaining network stages require dedicated rack space. In most systems one or several racks house the upper network stages, separated from the compute elements. This incurs significant additional costs, especially if a rack containing only a few switches has to be added. In this paper we propose and evaluate the performance and power-consumption of an alternative arrangement that connects elements in a cube-like topology. Building an indirect cube topology requires only the use of the switches that are integrated within the compute elements and also simplifies deployment. We explore a wide variety of system scales, ranging from 120 to 7680 compute nodes, in order to find out to which size the proposed topology can scale while keeping adequate performance levels and low power demands. An additional advantage of our proposal is that the same equipment can be reused to form a tree-based topology if a performance boost is needed. Keywords—Direct Topologies, High Performance Computing Clusters, Indirect Topologies, Interconnection Networks, Performance Evaluation, Power-Efficiency, Simulation.
منابع مشابه
Novel Techniques for Data Warehousing and Online Analytical Processing in Emerging Applications
A data warehouse is a collection of data for supporting of decision making process. Data cubes and on-line analytical processing(OLAP) have become very popular techniques to help users analyze data in a warehouse. Even though previous studies on a data warehouse and data cube have been proposed and developed, as new applications emerging, there are still technical challenges which have not been...
متن کاملA Hierarchy Topology Design Using a Hybrid Evolutionary Algorithm in Wireless Sensor Networks
Wireless sensor network a powerful network contains many wireless sensors with limited power resource, data processing, and transmission abilities. Wireless sensor capabilities including computational capacity, radio power, and memory capabilities are much limited. Moreover, to design a hierarchy topology, in addition to energy optimization, find an optimum clusters number and best location of ...
متن کاملNovel Correlation Parallel Network using Exchanged Cross Cube
In communication networks, Exchanged cross cube network is a new type of data network that makes use of cutting edge technology from several research areas to solve some problems current networks are faced with. The interconnection network plays an important role in large-scale parallel systems, because there is usually a need for one processor to communicate with other processors when a collec...
متن کاملImprovement Performances of Active and Reactive Power Control Applied to DFIG for Variable Speed Wind Turbine Using Sliding Mode Control and FOC
This paper deals with the Active and Reactive Power control of double-fed induction generator (DFIG) for variable speed wind turbine. For controlling separately the active and the reactive power generated by a DFIG, field oriented control (FOC) and indirect sliding mode control (ISMC) are presented. These non linear controls are compared on the basis of topology, cost, efficiency. The main cont...
متن کاملan Efficient Drive Circuit for Switched Reluctance Motor
The purpose of this paper is to present an efficient converter circuit used to drive switched reluctance motors. It uses the C-dump topology in which the trapped energy in the phase windings is returned to the supply capacitor to be used in the following strokes. In addition to the C-dump topology, it uses two extra feedbacks. One comes from a governor mounted on the shaft of motor which contro...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
- Optical Switching and Networking
دوره 8 شماره
صفحات -
تاریخ انتشار 2011