Load-Optimization in Reconfigurable Networks

2021 ◽  
Vol 48 (3) ◽  
pp. 39-44 ◽  
Author(s):  
Wenkai Dai ◽  
Klaus-Tycho Foerster ◽  
David Fuchssteiner ◽  
Stefan Schmid
Keyword(s):  
Data Centers ◽  
Algorithmic Problem ◽  
Reconfigurable Networks ◽  
Load Optimization ◽  
Clos Networks ◽  
Optical Circuit ◽  
Tree Topologies ◽  
Packet Switch ◽  
Routing Policies

Emerging reconfigurable data centers introduce the unprecedented flexibility in how the physical layer can be programmed to adapt to current traffic demands. These reconfigurable topologies are commonly hybrid, consisting of static and reconfigurable links, enabled by e.g. an Optical Circuit Switch (OCS) connected to top-of-rack switches in Clos networks. Even though prior work has showcased the practical benefits of hybrid networks, several crucial performance aspects are not well understood. In this paper, we study the algorithmic problem of how to jointly optimize topology and routing in reconfigurable data centers with a known traffic matrix, in order to optimize a most fundamental metric, maximum link load. We chart the corresponding algorithmic landscape by investigating both un-/splittable flows and (non-)segregated routing policies. We moreover prove that the problem is not submodular for all these routing policies, even in multi-layer trees, where a topological complexity classification of the problem reveals that already trees of depth two are intractable. However, networks that can be abstracted by a single packet switch (e.g., nonblocking Fat-Tree topologies) can be optimized efficiently, and we present optimal polynomialtime algorithms accordingly. We complement our theoretical results with trace-driven simulation studies, where our algorithms can significantly improve the network load in comparison to the state of the art.

Hybrid buffer‐based optical packet switch with negative acknowledgment for multilevel data centers

2020 ◽  
Vol 0 (0) ◽  
Author(s):  
Sumit Chandra ◽  
Shahnaz Fatima ◽  
Raghuraj Singh Suryavanshi
Keyword(s):  
Data Centers ◽  
Web Applications ◽  
Optical Switch ◽  
Contention Resolution ◽  
Multilevel Data ◽  
All Optical ◽  
Optical Buffering ◽  
Different Types ◽  
Packet Switch ◽  
Novel Design

AbstractIn the present scenario, data centers serve many functionalities like storage, transfer of data, supporting web applications, etc. In data centers, various levels of hierarchy different types of switches are required; therefore, multifunctional data centers are desired. This paper discusses a novel design for optical switch which can be placed at various levels of hierarchy. In the proposed design, multifunctionality contention resolution schemes which consider electronic and optical buffering and all-optical negative acknowledgment (AO-NACK) are considered. In buffering technologies, contending packets are stored in either in electronic RAM or in fiber delay lines. In case of the AO-NACK scheme, contending packets are blocked, and a negative acknowledgment is sent back to the transmitting node and blocked packets are retransmitted. For various considered schemes, Monte Carlo simulation have been performed, results in terms of packet loss probability are presented, and it has been found that the performance of optical buffering is much superior to electronic buffering and AO-NACK schemes. It is found that, in the AO-NACK scheme, the numbers of retransmitted packets due to contention blocking are 33,304 which can be reduced to 7, by using a small amount of buffer at each node.

Cloud Load Balancing Algorithms: A Comparative Assessment

2019 ◽  
Vol 16 (9) ◽  
pp. 3989-3994
Author(s):  
Jaspreet Singh ◽  
Deepali Gupta ◽  
Neha Sharma
Keyword(s):  
Cloud Computing ◽  
Load Balancing ◽  
Data Centers ◽  
Research Area ◽  
Comparative Assessment ◽  
Cloud Services ◽  
Cloud Data ◽  
Cloud Data Centers ◽  
Cloud Users

Nowadays, Cloud computing is developing quickly and customers are requesting more administrations and superior outcomes. In the cloud domain, load balancing has turned into an extremely intriguing and crucial research area. Numbers of algorithms were recommended to give proficient mechanism for distributing the cloud user’s requests for accessing pool cloud resources. Also load balancing in cloud should provide notable functional benefits to cloud users and at the same time should prove out to be eminent for cloud services providers. In this paper, the pre-existing load balancing techniques are explored. The paper intends to provide landscape for classification of distinct load balancing algorithms based upon the several parameters and also address performance assessment bound to various load balancing algorithms. The comparative assessment of various load balancing algorithms will helps in proposing a competent load balancing technique for intensify the performance of cloud data centers.

scholarly journals Numerical and experimental study of a high port-density WDM optical packet switch architecture for data centers

2013 ◽  
Vol 21 (1) ◽  
pp. 263 ◽  
Author(s):  
S. Di Lucente ◽  
J. Luo ◽  
R. Pueyo Centelles ◽  
A. Rohit ◽  
S. Zou ◽  
...  
Keyword(s):  
Experimental Study ◽  
Data Centers ◽  
Switch Architecture ◽  
Optical Packet Switch ◽  
Packet Switch

scholarly journals Scalable topological forwarding and routing policies in RINA-enabled programmable data centers

2017 ◽  
Vol 28 (12) ◽  
pp. e3256 ◽  
Author(s):  
Sergio Leon Gaixas ◽  
Jordi Perelló ◽  
Davide Careglio ◽  
Eduard Grasa ◽  
Diego R. López ◽  
...  
Keyword(s):  
Data Centers ◽  
Routing Policies

Optical Switching in Next-Generation Data Centers

2019 ◽  
pp. 164-193 ◽  
Author(s):  
Vaibhav Shukla ◽  
Rajiv Srivastava ◽  
Dilip Kumar Choubey
Keyword(s):  
Data Center ◽  
Optical Switching ◽  
Large Scale ◽  
Data Centers ◽  
Current Data ◽  
Limited Bandwidth ◽  
Switch Architectures ◽  
Pros And Cons ◽  
High Power Consumption ◽  
Packet Switch

The leading content provider companies like Google, Yahoo, and Amazon installed mega-data centers that contain hundreds of thousands of servers in very large scale. The current data center systems are organized in the form of the hierarchal tree structure based on bandwidth-limited electronic switches. Modern data center systems face a number of issues like high power consumption, limited bandwidth availability, server connectivity, energy and cost efficiency, traffic complexity, etc. One of the most feasible solution of these issues is the use of optical switching technologies in the core of data center systems. In this chapter a brief description about the modern data center system is presented, and some prominent optical packet switch architectures are also presented in this chapter with their pros and cons.

Load Optimization in a Grid Structure for Parallel Simulations of the Throughput of a Packet Switch Node

2014 ◽  
Vol 12 (2) ◽  
pp. 23-30
Author(s):  
Tasho Tashev ◽  
Vladimir Monov ◽  
Radostina Tasheva
Keyword(s):  
Computer Simulations ◽  
Parallel Computer ◽  
Grid Structure ◽  
Load Optimization ◽  
Parallel Simulations ◽  
Crossbar Switch ◽  
Packet Switch ◽  
Simulation Results

Abstract In the present paper we employ the grid structure of IICT-BAS for parallel computer simulations of the throughput of a crossbar switch node. In our simulations we use PIM-algorithm for non-conflict scheduling in a crossbar node with hotspot load traffic. The obtained simulation results enable us to propose a procedure for optimizing the load of a grid structure in order to minimize the overall time of performance.

scholarly journals Understanding the Performance Guarantee of Physical Topology Design for Optical Circuit Switched Data Centers

2021 ◽  
Vol 5 (3) ◽  
pp. 1-24
Author(s):  
Shizhen Zhao ◽  
Peirui Cao ◽  
Xinbing Wang
Keyword(s):  
Data Centers ◽  
Topology Design ◽  
Performance Guarantee ◽  
Performance Limit ◽  
Coupling Technique ◽  
Performance Guarantees ◽  
Performance Metric ◽  
Optical Circuit ◽  
And Performance ◽  
The Ideal

As a first step of designing O ptical-circuit-switched D ata C enters (ODC), physical topology design is critical as it determines the scalability and the performance limit of the entire ODC. However, prior works on ODC have not yet paid much attention to physical topology design, and the adopted physical topologies either scale poorly, or lack performance guarantee. We offer a mathematical foundation for the design and performance analysis of ODC physical topologies in this paper. We introduce a new performance metric β(G ) to evaluate the gap between a physical topology G and the ideal physical topology. We develop a coupling technique that bypasses a significant amount of computational complexity of calculating β(G). Using β(G ) and the coupling technique, we study four physical topologies that are representative of those in literature, analyze their scalabilities and prove their performance guarantees. Our analysis may provide new guidance for network operators to design better physical topologies for their ODCs.

scholarly journals Multitask Learning by Multiwave Optical Diffractive Network

2020 ◽  
Vol 2020 ◽  
pp. 1-7
Author(s):  
Jing Su ◽  
Yafei Yuan ◽  
Chunmin Liu ◽  
Jing Li
Keyword(s):  
Neural Networks ◽  
Hardware Implementation ◽  
Electrical Networks ◽  
Accuracy Improvement ◽  
Optical Neural Networks ◽  
Optical Circuit ◽  
Optical Grating ◽  
Comparable Performance ◽  
The Cost

Recently, there has been tremendous research studies in optical neural networks that could complete comparatively complex computation by optical characteristic with much more fewer dissipation than electrical networks. Existed neural networks based on the optical circuit are structured with an optical grating platform with different diffractive phases at different diffractive points (Chen and Zhu, 2019 and Mo et al., 2018). In this study, it proposed a multiwave deep diffractive network with approximately 106 synapses, and it is easy to make hardware implementation of neuromorphic networks. In the optical architecture, it can utilize optical diffractive characteristic and different wavelengths to perform different tasks. Different wavelengths and different tasks inputs are independent of each other. Moreover, we can utilize the characteristic of them to inference several tasks, simultaneously. The results of experiments were demonstrated that the network could get a comparable performance to single-wavelength single-task. Compared to the multinetwork, single network can save the cost of fabrication with lithography. We train the network on MNIST and MNIST-FASHION which are two different datasets to perform classification of 32∗32 inputs with 10 classes. Our method achieves competitive results across both of them. In particular, on the complex task MNIST_FASION, our framework obtains an excellent accuracy improvement with 3.2%. In the meanwhile, MNSIT also has the improvement with 1.15%.

Sign in / Sign up

Export Citation Format

Share Document