Network Architectures for Data Centers - Term Paper Example

?NETWORK ARCHITECTURES FOR DATA CENTERS Network Architectures for Data Centers Affiliation Table of Contents Introduction 3 DataCenter Role 5 Challenges and Requirements for Efficient Data Center Technology 6 Data Center Resources Fragmentation 8 Server to Server Connectivity 8 Proprietary hardware that scales up, not out 9 Placement anywhere 9 Server to server bandwidth 10 Support 100,000 servers 10 Convergence Application in Next Generation Datacenter Technology 10 Cisco SONA: New Data Center Model 12 14 References 16 Abstract A few years ago a new technology based trend emerged, which quickly gained the attention of industries. This technology based trend is known as a centralized data center. As the central and major provisioning point of communication network, data storage, and compute support services, the corporate data center was established in the middle of the mission critical operational processes and functional areas for one basic purpose: The expenditures of offering huge scale, centralized technology based services to thousands of clients were extremely high that it could simply be managed and controlled by practically collecting and connecting that technology based resources in a common place. On the other hand, at the present technology has changed. Additionally, there is an increase in demands for operational sustainable, efficiencies corporate processes, dynamic operational service diversification, and cost efficiency. In addition, these ever-increasing demands put the data center in the crosshairs of business analysis as its significance has increased more than ever before (LSI Corporation, 2012). This paper presents a detailed analysis of network architectures for data centers. This research will offer a comprehensive analysis of some of the major aspects required for the development and application of data center architecture application and their possible implication for better corporate management. Introduction At the present, data centers serve as the central command center for the majority of business organizations in spite of size and industry. Without a doubt, it offers an excellent support for managing huge amounts of information and data, and ensures effective execution of business operations 24 hours a day and 7 days a week. In this scenario, network based data centers are designed, developed and maintained to store, process, and exchange information and data for economic services corporations, companies, transport business, administration, utilities, educational institutions, network carriers and internet service providers (ISPs). In fact, investments in the development of data centers are growing at a rate of more than 20 percent per year. In view of the fact that the business organizations completely depend on latest information technologies to take actions thus the current data centers have become less efficient along the lines of competence, effortlessness and agility of maintenance (H3C Technologies Co., Limited, 2012; Juniper Networks, Inc., 2012; Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012). In addition, in the field of data centers, several major trends are happening in parallel, with each representing a fundamental change in terms of how it is managed. Data centers currently include both networking and server arrangements. However, the server part of this technology based infrastructure is at the present far down the road of technology based commoditization, in this scenario high end corporate class servers have been replaced by large amounts of low cost technology based servers. Additionally, new technology based improvement in distributed technology based computing and administration systems have supported the irregularity of individual network servers to be masked by the collective communication reliability of the data center system as a whole (H3C Technologies Co., Limited, 2012; Juniper Networks, Inc., 2012; Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012). Moreover, the quality and performance of network architectures for data centers are improving day by day. This research will present a detailed analysis of network architectures for data centers and potential developments in this regard. This paper will present an analysis of possible technology improvements for data centers. Data Center Role As the functionality of the data center remains much similar as it has for all the time been, the way in which it accomplishes this functionality is developing speedily. Additionally, to deal with current evolving demands, a dichotomy of sorts appears in the area of data center technology based architecture and operating viewpoint. Alternatively, there is a junction happening that is guiding to the emergence of what many acknowledge as the “mega-scale data center,” a center that offers storage, computing and communications arrangements on a large-scale level. In the same way, we see the emergence of remote and distributed data center functioning in response to the popularity and effectiveness of the cloud model. On the other hand, these situations cause a wide variety of issues and challenges of putting together a remote cloud technology based services in the corporate data center, that at the present requires initiating looking more similar to a local or internal cloud service supplier, and consequently incorporates a complete new array of issues in the company of offering data center technology services (LSI Corporation, 2012). Driving Forces for Data Center Technology Evolution The main force behind this technology based evolution is more often than not surprising, however tremendously quick expansion of unstructured data and an associated demand for rich media, both of which put wonderful pressure on computing and storage technology based resources. As a result of the permanent pressure implemented on information technology businesses to regulate their functioning financial plans to imitate more efficient financial systems of operations and scale competencies in response to data center evolution. In addition, another factor is increasing accessibility of service Ethernet switching hardware, fundamental to the flawless arrangement of the next generation data center (LSI Corporation, 2012). Challenges and Requirements for Efficient Data Center Technology A wide variety of software tools and applications execute inside a single data center, normally with every application is hosted on its individual set of (potentially virtual) communication server systems. In this scenario, every system is related to one or more visibly noticeable and routable IP communication addresses to those clients in the Internet transmission their working requests and from that they get answered. Additionally, inside a data center, requests are sent to a pool of front-end servers that process the requests. In addition, this process is normally performed by a specialized load balancer. In this scenario, with predictable load balancer network terminology, the internet protocol address to that responses are transmitted is known a VIP (virtual IP) address and the internet protocol addresses of the network servers over that the communication requests are sent are recognized as Direct IP Addresses (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012). Figure 1 Network architecture for data centers Image Source: http://research.microsoft.com/pubs/79348/presto27-greenberg.pdf Figure1 demonstrates the conventional data center architecture. In this arrangement requests coming from the Internet are IP (network layer 3) routed by access and border routers to a communication layer 2 domain foundational on the target VIP communication address. Additionally, the virtual internet protocol is configured onto the 2 load balancer linked connection on the top network switches, and complex working systems are employed to make sure that if one load balancer stops working and the other backs up arrangement picks up the network data traffic load. In addition, for every VIP, the network communication load balancers are effectively configured through a record of DIPs, which are the internal and private addresses of network physical communication servers in the racks below the load balancers. Moreover, this catalog of DIPs encompasses the pool of network data center servers those are able to manage computing requests to that VIP, and the load balancer spreads demands all through DIPs in the pool (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012). Furthermore, in the conventional approach of corporate network and resource management we can face a number of issues. I have outlined below some of the useful techniques to deal with such issues: (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012) Data Center Resources Fragmentation Majority of load balancing methods for example destination NAT and direct server return, need that all the DIPs in a VIP’s pool are in the same network communication layer 2 area. Hence, this restriction demonstrates that if a software system expands and needs more network servers, it is not able to make use of accessible network servers in additional layer 2 domains: eventually result in underutilization and fragmentation of network technology based resources. In this scenario, network load balancing through Source NAT (or full-NAT) does permit servers to be extended all through network layer 2 communication domains, however then the servers never observe the user IP that is frequently improper for the reason that servers make use of the client IP for all the activities such as from response customization and data mining to authoritarian fulfillment (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012). Server to Server Connectivity The hierarchical environment of the communication network outlines that for network servers in different layer 2 domains to correspond, data traffic must move to the layer 3 part of the network. Additionally, layer 3 communication ports are a great deal more costly than layer 2 ports. Hence, these communications are normally oversubscribed (for example capability of the links among access border and routers is considerably limited as compared to the sum of the production capability of the servers linked to the network access routers). In this scenario, the result is that the network communication bandwidth accessible among servers in various components of the data center is quite inadequate or ineffective, which causes a severe worldwide optimization issue thus for all the servers associated to all applications have to be placed with a great care to make sure the amount of their data and information traffic does not flood some of the network communication lines or links. However, attaining this point of management among applications is flawed in practice (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012). Proprietary hardware that scales up, not out The network load balancers in the customary architecture are employed in couples in a 1+1 resiliency pattern. When the network load turns out to be too high for the load balancers, operators put back the present load balancers with a new pair having the extra capability; however it is a costly and un-scalable policy. In addition, in place of traditional network communication architecture, we need a structural design that can help us resolve below given issues: (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012) Placement anywhere The data center communication architecture should allow a server anywhere in the data center to be a part of the pool of communication servers following some VIP, consequently that server pools can be dynamically expanded or shrunk. Server to server bandwidth In a number of cases, the amount of data and information traffic between servers within the data center the quantity of traffic is swapped with clients outside the data center (for example a number of systems necessitate common computation spread over a wide variety of servers in an attempt to produce an output in a short time period). Thus, it entails that the arrangement should be capable of offering as much bandwidth as possible between each pair of network servers in the data center, without taking into consideration their positions. Support 100,000 servers Large data centers currently have 100,000 or more servers for process large volumes of data, although the communication network stands in the way of these communication servers in random pools. However, larger size data centers should facilitate exhaustive inner communications among all network servers (Greenberg, Lahiri, Maltz, Patel, & Sengupta, 2012; Juniper Networks, Inc., 2012). Convergence Application in Next Generation Datacenter Technology Data center networks are at the present offer a wide variety of functionalities which are separated into various categories such as data network, High Performance Computing (HPC) and storage area network (SAN) cluster network. In addition, the structural design is typically characterized through: (H3C Technologies Co., Limited, 2012) A very high speed data network, typically developed through Ethernet based technologies, working as an interface to the data center  A storage network, typically also Network Attached Storage (NAS) with Ethernet or Fiber Channel SAN, operational below the hood Parallel network working and computing servers linked with Infini-Band or Gigabit Ethernet links Servers executing with the heterogeneous OSs and systems, and communicating with the diverse ports through diverse protocols. Network servers developed through a wide variety of network adapters interconnecting various networks Figure 2 Data Center Architecture Image Source: http://www.h3c.com/portal/Home/Index_News_Event/Solutions/201202/739416_40_0.htm In addition, this kind of network topology necessitates latest server equipped with 3 different network adaptors, making difficult development and increasing capital expense. Additionally, this arrangement also encompasses a Converged Network Adapter that maintains a widespread structure for storage and networking decreases CapEx for switches, adapters, cables, and leads to the junction of heterogeneous communication networks. It can effectively deal with high speed data and information transfers for both SAN and LAN. Also, the Converged Network Adapter increases the instant performance and convinces the output requirements of high performance network and data center computing. In view of the fact that networking business vendors are leading in the direction of an all Ethernet solution to end the varied network arrangement, thus Ethernet communicating adaptors are attaining a great deal more than FC-based storage based network adapters (H3C Technologies Co., Limited, 2012). Moreover, Converged Enhanced Ethernet and FC-over-Ethernet adapters’ technology permits FC and Ethernet to share the similar network arrangement minimizes the overall cost of ownership for the data center. In fact, it as well produces a more elastic network arrangement that is capable to become accustomed to changes in our company (H3C Technologies Co., Limited, 2012). Cisco SONA: New Data Center Model Cisco’s SONA technology based structure denotes how centralized corporations are able to develop an intelligent information network that is able to optimize systems, business procedures and technology based resources. Additionally, with Cisco’s SONA central CIOs are able to noticeably augment efficiency and competence by increasing business dependence, minimizing prices and supporting mission precedence if they make the accurate investment in the data center communication network (Cisco, 2012; H3C Technologies Co., Limited, 2012). In addition, Central Network Architecture, foundational on Cisco SONA, offers a system level structure to tackle instant data center requirements for business continuance and consolidation. Additionally, Cisco combines application networking services and network infrastructure services to facilitate growing service oriented architectures, on-demand computing and Virtualization tools and systems. Moreover, Cisco's new technology based architecture and methods provide central information technology decision makers with an excellent support for organizing computing resources, storage and software technologies that effectively support their mission requirements. Furthermore, the next generation data center architecture technology includes 3 layers: (Cisco, 2012; H3C Technologies Co., Limited, 2012) Network Compute Storage As shown in figure 3, every layer presents a distinct function inside the Cisco Data Center Network Architecture, connecting enhanced and well organized interactive technology based systems and services to critical company process flows. Figure 3 Three-Layer Data Center Model Image Source: http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/net_implementation_white_paper0900aecd805fbdfd.html The network layer offers secure and reliable access to the network compute layer. It makes use of Layer 2/Layer 3 network communication switches to establish connection between clients and data center technology based resources. In this scenario, service and appliance modules ensure protected access and optimize resource use and performance. Additionally, the compute layer offers the computing technology based resources that execute software executed by end users. In the same way, the server switches communicate the computing technology based resources and offer access to the data and information storage layer. Conventionally in the compute layer, parallel systems execute on supercomputers that are prohibitively classy for a wide variety of businesses to get and perform. In addition, the storage layer of the overall data center architecture stores the data employed by applications on storage subsystems. Moreover, the storage layer makes use of storage switches and optical transport arrangements to connect and offer access to disk resources inside as well as among data centers (Cisco, 2012; H3C Technologies Co., Limited, 2012). Consolidation and End-to-End Virtualization Implementation of silos cannot be successfully scaled to convene the increasing requirements of these days’ data centers. In this scenario, the next generation of new technology based data centers merge the arrangement and virtualizes technology based resources throughout a smaller amount of facilities and components, and saves valuable kilowatt hours. In addition, data center technology based resources are no longer devoted to precise applications however are logically allocated to systems as required. As shown in the below given figure 4: (Cisco, 2012; H3C Technologies Co., Limited, 2012) Figure 4 Next-Generation Data Centers-Consolidation and Virtualization, Image Source: http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/net_implementation_white_paper0900aecd805fbdfd.html Conclusion Data centers serve as the central command center for the majority of business organizations in spite of size and industry. At the present, many firms establish data centers to deal with their huge amount of data. There is a need for an effective architecture to make best use of this network. This paper has presented a detailed analysis of some of the major aspects of network architectures for data centers. In this scenario, I have discussed some important aspects such as a basic need of the network data centers and arrangements. This research has also offered a comprehensive analysis of new data center technology infrastructures. References Cisco. (2012). Next-Generation Federal Data Center Architecture. Retrieved November 07, 2012, from http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/net_implementation_white_paper0900aecd805fbdfd.html Greenberg, A., Lahiri, P., Maltz, D. A., Patel, P., & Sengupta, S. (2012). Towards a Next Generation Data Center Architecture: Scalability and Commoditization. Retrieved November 08, 2012, from http://research.microsoft.com/pubs/79348/presto27-greenberg.pdf H3C Technologies Co., Limited. (2012). H3C next generation solution for datacenter. Retrieved November 04, 2012, from http://www.h3c.com/portal/Home/Index_News_Event/Solutions/201202/739416_40_0.htm Juniper Networks, Inc. (2012). White Paper - Security in the Next-Generation Data Center. Retrieved November 07, 2012, from http://www.juniper.net/us/en/local/pdf/whitepapers/2000374-en.pdf LSI Corporation. (2012). The Next-Generation Data Center. Retrieved November 06, 2012, from http://www.lsi.com/invention/Pages/Articles/next-generation-data-center.aspx Read More