LightBlog
Contact at mumbai.academics@gmail.com or 8097636691/9323040215
Responsive Ads Here

Saturday, 3 February 2018

Scalable and Cost-Effective Interconnection of Data-Center Servers Using Dual Server Ports(2011)



Scalable and Cost-Effective Interconnection 

of Data-Center Servers Using Dual Server

 Ports(2011)

ABSTRACT:            
The goal of data-center networking is to interconnect a large number of server machines with low equipment cost while providing high network capacity and high bisection width. It is well understood that the current practice where servers are connected by a tree hierarchy of network switches cannot meet these requirements. In this paper, we explore a new server-interconnection structure. We observe that the commodity server machines used in today’s data centers usually come with two built-in Ethernet ports, one for network connection and the other left for backup purposes. We believe that if both ports are actively used in network connections, we can build a scalable, cost-effective interconnection structure without either the expensive higher-level large switches or any additional hardware on servers. We have proven that FiConn is highly scalable to encompass hundreds of thousands of servers with low diameter and high bisection width. We have developed a low-overhead traffic-aware routing mechanism to improve effective link utilization based on dynamic traffic state. We have also proposed how to incrementally deploy FiConn.
PROJECT PURPOSE:
In this paper, we explore a new server-interconnection structure. We observe that the commodity server machines used in today’s data centers usually come with two built-in Ethernet ports, one for network connection and the other left for backup purposes.
PROJECT SCOPE: 
The goal of data-center networking is to interconnect a large number of server machines with low equipment cost while providing high network capacity and high bisection width. It is well understood that the current practice where servers are connected by a tree hierarchy of network switches cannot meet these requirements.
In this project, we explore a new server-interconnection structure. We observe that the commodity server machines used in today’s data centers usually come with two built-in Ethernet ports, one for network connection and the other left for backup purposes.
INTRODUCTION:
DATA-CENTER networking designs both the network structure and associated protocols to interconnect thousands of or even hundreds of thousands of servers at a data center, with low equipment cost, high and balanced network capacity, and robustness to link/sever faults. Its operation is essential to offering both numerous online applications, e.g., search, gaming, Web mail, and infrastructure services. It is well understood that tree-base solution in current practice cannot meet the requirements.
In this paper, First, it costs less to build a data-center network. We do not need high-end, expensive switches, which are widely used today. Standard, off shelf servers with two ports (one for operation in network connection, the other for backup) are also readily available. Second, the wiring becomes relatively easy since only two server ports are used for interconnection. We do not need to add additional hardware or wires on a server except the two NIC ports. Third, it may spawn more academic research into data centers. New problems and solutions in data-center networking, systems, and applications can be found, implemented, and assessed through an easy-to-build test bed at a university or institution. Today, data-center infrastructure may only be afforded by a few cash rich companies such as Microsoft, Google, and Yahoo.
Neither current practice nor recent proposals can solve our problem. The tree-based solution requires expensive, high-end switches at the top level of the tree in order to alleviate the bandwidth bottleneck. The scaling of the Fat-Tree solution is limited to the number of ports at a switch, and it also needs more switches. The fundamental problem is that, we need to design a new network structure that works for servers with node degree of only 2 in order to scale. We propose FiConn, a scalable solution that works with servers with two ports only and low-cost commodity switches. FiConn defines a recursive network structure in levels. A high-level FiConn is constructed by many low-level FiConns. When constructing a higher-level FiConn, the lower-level FiConns use half of their available backup ports for interconnections and form a mesh. This way, the number of servers in FiConn, , grows double-exponentially with FiConn levels.
PROBLEM DEFINITION:
Data-Center networking designs both the network structure and associated protocols to interconnect thousands of or even hundreds of thousands of servers at a data center, with low equipment cost, high and balanced network capacity, and robustness to link/sever faults.
EXISTING SYSTEM:
Existing network architecture typically consists of a tree of routing and switching elements with progressively more specialized and expensive equipment moving up the network hierarchy. Unfortunately, even when deploying the highest-end IP switches/routers, resulting topologies may only support 50% of the aggregate bandwidth available at the edge of the network, while still incurring tremendous cost. Non-uniform bandwidth among data center nodes complicates application design and limits overall system performance.
LIMITATIONS OF EXISTING SYSTEM:
Ethernet switches to support the full aggregate bandwidth of clusters consisting of tens of thousands of elements. Similar to how clusters of commodity computers have largely replaced more specialized SMPs and MPPs, we argue that appropriately architected and interconnected commodity switches may deliver more performance at less cost than available.
PROPOSED SYSTEM:
v  In this paper, we explore a new server-interconnection structure. We observe that the commodity server machines used in today’s data centers usually come with two built-in Ethernet ports, one for network connection and the other left for backup purposes.
v  We believe that if both ports are actively used in network connections, we can build a scalable, cost-effective interconnection structure without either the expensive higher-level large switches or any additional hardware on servers.
v  We have proven that FiConn is highly scalable to encompass hundreds of thousands of servers with low diameter and high bisection width. We have developed a low-overhead traffic-aware routing mechanism to improve effective link utilization based on dynamic traffic state.
v  We have also proposed how to incrementally deploy FiConn. FiConn a novel server-interconnection network structure that utilizes the dual-port configuration existing in most commodity data-center server machines it is a highly scalable structure because the total number of servers it can support is not limited by the number of server ports or switch ports. It is cost-effective because it requires less number of switches and links than other recently proposed structures for data centers.
ADVANTAGES OF PROPOSED SYSTEM:
We compare FiConn with other recently proposed data-center networking structures in detail, disclosing that FiConn is advantageous over all other in terms of deploying cost, but holds high scalability with constant number of server ports and switch ports. we make three main contributions in FiConn. First, FiConn offers a novel network structure that is highly scalable with off-the-shelf servers of node degree 2 and low-end commodity switches while having low diameter and high bisection width. FiConn uses traffic-aware routing that exploits the available link capacities based on traffic dynamics and balances the usage of different links to improve the overall network throughput.
MODULES DESCRIPTION:
NETWORKING MODULE:
Client-server computing or networking is a distributed application architecture that partitions tasks or workloads between service providers (servers) and service requesters, called clients. Often clients and servers operate over a computer network on separate hardware. A server machine is a high-performance host that is running one or more server programs which share its resources with clients. A client also shares any of its resources; Clients therefore initiate communication sessions with servers which await (listen to) incoming requests.
DATA CENTER:
v  First, it costs less to build a data-center network. We do not need high-end, expensive switches, which are widely used today. Standard, off shelf servers with two ports (one for operation in network connection, the other for backup) are also readily available.
v  Second, the wiring becomes relatively easy since only two server ports are used for interconnection. We do not need to add additional hardware or wires on a server except the two NIC ports.
v  Third, it may spawn more academic research into data centers. New problems and solutions in data-center networking, systems, and applications can be found, implemented, and assessed through an easy-to-build test bed at a university or institution.  Today, data-center infrastructure may only be afforded by a few cash rich companies such as Microsoft, Google, and Yahoo.
INTERCONNECTION STRUCTURE:
Interconnection structures proposed for data centers, the current practice of the tree-based structure, and two recent proposals of Fat-Tree.
TREE:
In current practice, servers are connected by a tree hierarchy of network switches, with commodity switches at the first level and increasingly larger and more expensive switches at the higher levels. It is well known that this kind of tree structure has many limitations. The top-level switches are the bandwidth bottleneck, and high-end high-speed switches have to be used.
Moreover, a high-level switch shows as a single-point failure spot for its sub tree branch. Using redundant switches does not fundamentally solve the problem, but incurs even higher cost.
FAT-TREE:
FiConn differs from Fat-Tree in several aspects. First, FiConn puts the interconnection intelligence on servers, rather than on switches as in Fat-Tree. Second, there are three levels of switches in Fat-Tree, but only one lowest level in FiConn.
Hence, the number of used switches is much smaller in FiConn. Consider the total number of servers as and -port switches being used the number of servers Fat-Tree supports is restricted by the number of switch ports, given the three layers of switches. FiConn does not have this limitation and extends to a very large number of servers.
DEPLOYMENT OF FiCONN:
We design a two basic routing algorithm in FiConn that leverages the level-based characteristic of FiConn.
TRAFFIC-AWARE ROUTING (TAR):
TOR balances the use of different levels of FiConn links and serves as the basis for FiConn routing. However, it has two limitations. First, a pair of servers cannot leverage the two ports on each to improve their end-to-end throughput in TOR. Second, TOR cannot further utilize the available link capacities according to dynamic traffic states to improve the networking throughput. To overcome these limitations, we design TAR in FiConn.
TRAFFIC-OBLIVIOUS ROUTING (TOR):
We find that the aggregate throughput of TOR is only 1 Gb/s, resulting from the bottleneck level-2 link that connects the two selected FiConn s. However, by exploiting the links beyond the two FiConn ’s and the bottleneck level-2 link, TAR achieves an aggregate throughput of 99.5 Gb/s, which shows a tremendous improvement over TOR.
HARDWARE AND SOFTWARE REQUIREMENTS:
HARDWARE REQUIREMENTS:
•         System                        :           Pentium IV 2.4 GHz.
•         Hard Disk                   :           40 GB.
•         Floppy Drive   :           1.44 Mb.
•         Monitor           :           15 VGA Colour.
•         Mouse             :           Logitech.
•         Ram                 :           512 Mb.
SOFTWARE REQUIREMENTS:
•         Operating system        :  Windows XP.
•         Coding Language       :  JDK 1.6
•         Tools                           :  Eclipse 3.3

No comments:

Post a Comment