Virtualized Data Center for CLOUD Based Network Architecture and Environment

Transcription

1 Virtualized Data Center for CLOUD Based Network Architecture and Environment Mr. Aniruddha P. Tekade, Department of Computer Technology, VJTI, Mumbai Prof. Varshapriya J.N., Department of Computer Technology, VJTI, Mumbai ABSTRACT Today none of the operational domains in the world work without automated machines or computers. The dependency over the computer systems developed dramatically. The desire for the very perfect, highly speedy and less error-prone calculations of humans, made the computer systems famous and deployable in almost every sector of the business and its intelligence. Today we can observe the long distant but a reliable connectivity between different computer systems has been faithfully accepted. Human interventions have been reduced. A single data storage source has made the remote and distributed businesses connected. But somewhere down the line, each of these operational scenarios has to be monitored and controlled by a ultra-mega-super System a collection of multiple server systems gathered at a common place for a special purpose A DATA CENTER. Data Center is critical for corporal operations, monitoring networks and keeping control over all business intelligence activities. Businesses thrive over the operations of data center today. Data Center systems are like building or a portion of building whose primary function is to house a computer system and its support areas. This paper aims to introduce the concept of Virtualized data centers that can play a vital role in today s very challenging domain of Cloud Computing. The sole purpose of this paper is to make proposal of an architecture that has the power to work as a data center that functions virtually for CLOUD technology and no longer physically. We need no investment on multiple hardware resources to build a business quality data center. The implementations of common storage technologies like RAID and LVM can also be deployed with this architecture. All the instances can be loadedunloaded virtually. In the next sections II and III of this papers, we have attempted to prove how a data center can be simulated in a virtual way so as to work for a real time network. This paper also contains statistical comparison between a datacenter full of giant-costly hardware and our green virtual data center over the factors such as cost, temperature, electricity estimations, maintenance, operational efficiency and some other nontechnical issues. Keywords: Architecture, Green, Virtualization, instances, VDC, Shell Script, Bash, SOA, LVM, RAID, TIA, TR/ER. Introduction Data Center is a facility used to house computer systems and is associated with the components, such as telecommunications & storage systems. Cloud computing is a method of delivering computing services from a large, highly virtualized data center to many independent end users, using shared applications and pooled resources. While there are many different definitions for cloud computing, it is typically distinguished by the following attributes: on-demand self-service, broad network access, resource pooling, rapid and elastic resource provisioning, and metered service at various quality levels. Implementation of these attributes as part of a large enterprise-class cloud computing service that provides continuous availability to a large number of users typically requires significantly more server, networking, and storage resources than conventional data centers. This is only achievable through extensive use of virtualization. This is achievable only through extensive use of virtualization. While server virtualization has existed since the 1960s, when it was first implemented on IBM mainframes, it has only become widely available on affordable commodity x86 servers within the last decade or so. In recent years, many equipment vendors have contributed to the hardware and software infrastructure, which has made enterprise-class virtualization widely available. This, in turn, enables new designs for cloud computing, including hosting multiple independent tenants on a shared infrastructure, rapid and dynamic provisioning of new features, and implementing advanced load balancing, security, and business continuity functions (including multisite transaction mirroring). This has brought about profound changes in many aspects of data center design, including new requirements for the data center network. Data Center generally includes redundant or backup power supplies, redundant data communications connections, environmental controls (e.g., air conditioning, fire suppression) and security devices. Also known as the server farm or the computer room, and is deployed Blue Ocean Research Journals 4

2 where the majority of an enterprise servers and storage are located, operated and managed. For Example, The National Climatic Data Center (NCDC) is a public data center that maintains the world's largest archive of weather information. A private data center may exist within an organization's facilities or may be maintained as a specialized facility. Every organization has a data center, although it might be referred to as a server room or even a computer closet. Functionality - a restricted access area containing automated systems that constantly monitor server activity, Web traffic, and network performance, etc. To cope up with above mentioned problems, taking consideration of dedicated virtualizations and its inter-related technology to design high end systems like Data Center is the only sensible solution to the problem. Every time we cannot compete with the hardware, in fact whatever the advancement in the softwares be made, we just cannot match the pace of dedicated hardware units which are designed for a special purpose, ex; RAID (Redundant Array of Independent Disks). But still, today even the desktop home PC processors have been developed and advanced enough to serve this motto of ours. They are capable to produce multithreaded programs and run them, so we can certainly opt for this solution of Virtualized Data Center. Requirnments of Virtual Data Center Basic motives that we must have a data center in today s era of computing: Today each and every business thrives on data and internet and IT operations are a crucial aspect of most organizational operations. When we have many servers to take care of, bringing them together and keeping them at a centralized location so that their maintenance and monitoring becomes easier is the only option. i.e. The purpose of data center Important for any organization that owns a number of servers, since datacenters bring everything at one place. They can be proved very effective against initial errors such as manual intervention with data, software or hardware failures. When there is centralized location of all critical hardware of a business group. A data plan of an enterprise should be able to work against all the odds such as manual errors, application failures, natural and man-made disasters such as fire, theft, floods etc. The data center stores not only servers but also routers and switches that help the servers to communicate with one another at one place. It offers high availability and business continuity. When one server fails, the others take over, providing you a seamless experience. High scalability. It accommodates all the changes as a business grows. You can maximize the usage of your own networking assets to improve company internal network operations and many other technical & non-technical benefits. Functional Requirement of a Data Center There are very few factors enlisted in the following lines which collectively provide the high end simplicity, scalability, flexibility and modularity of data center operations. 1. Location It is a place to locate computer systems, storage and networking devices. All these systems should be in a common place so as to be able to monitor, system record, daily output and network traffic logs, etc. 2. Power Power needed to maintain the devices and provide maximum uptime and high availability of 99.9%. All the specifications and figures need to be run with standardized input values only. i. HAVC Need of creating Temperature controlling environment so as to reduce heat generation through processors of data centers. ii. Structured cabling Connectivity provided to other devices both inside and outside. Issues Facing Data Centers As the functional requirements ships with the data center thought process there are some issues that a data centers has to go through while installing the whole setup. 1. Reliability and Availability The factors that are responsible for the reliability of data centers are : Fig 1.1 Reliability and Availability Blue Ocean Research Journals 5

3 2. Power The power factor holds coolings, lightnings, critical loads, Battery charging etc factors in the consideration: 1. Rack Fundamentals 2. Cooling Fundamentals 3. Power Distribution Fundamentals 4. Availability and Reliability Fundamentals 5. Physical Infrastructure Management Fundamentals 6. Fire Protection Methods 7. General Design Data Center Standards Following are the existing standards of data centers 1. TIA 942 Telecommunications Infrastructure standard for Data Centers, April EN Information Technology data centers Europe 3. ANSI-BICSI-002 Data Center Standard complementary to TIA ISO/IEC NP Information Technology Generic cabling for Data Centers 3. Fig 1.2 Power Factors 4. Rack Space A data center management server needs total of 5 basic utilities: Power, Data, Control/KVM and Cooling & Racking; i.e It has been surveyed that approx. 18% of Rack Space goes unused due to power consumption and cooling. 5. Site Location While choosing the location for the site of data center we need to think over Seismic problems, ground subsidence problems, security/criminal problems, availability of electricity, water, sewage and telecoms. Other than these, proximity to railway lines, airports, chemical storage and military, EMC problems like mobile masts and radar transmitters. 6. Heat & Cooling Heat actually accounts for 55% of electronic failure. There must be 1kw of air conditioning for 1kw of heat inside your rack. Heat normally conducted in three ways transferring through metal (Conduction), transferring through liquid or gas (convection), and transferring heat through electronic waves (Radiation). It s important not to over fill your rack. For every server racks, you should have one communication and one storage rack. Heat generation occurs at the chip, server, rack and room levels. A 4kw rack is ideal i.e. Average server is 370W. Critical components of a typical data center The data center expertise says, there are 7 factors that are very critical for the building of data center. Some other standards in the world of data center design are ANSI/TIA-942 Telecommunication Infrastructure Standards for Data Center consist of: 1. TIA/EIA 568 Copper & Fiber Cabling, 2. TIA/EIA 569 Pathways & Spaces, 3. TIA/EIA 606 Administration, 4. TIA/EIA 607 Grounding and Bonding, 5. TIA/EIA 768-An Outside Plant. There are 2 special kinds of rooms TR Telecommunication Room and ER Equipment Room. A TE is an enclosed architectural space for housing telecommunications equipment, cable terminations and cross connection cablings. The other ER is an environmentally controlled centralized space for telecom equipment. Difference between TR and ER s are that ERs serve a building or campus while TRs serve a floor in building. Requirement of a Data Centers A data center is always think of and given the most utmost importance as each single operation of the business it is running is very critical. The single set of hardware that collectively builds the data center ships with a tabulated list of lots of requirements. The core requirements are tabulated in following manner Blue Ocean Research Journals 6

4 Components Computer Room/Server Room Control Room General Office Area Entrance Facility Fire Gas Suppression Store Electrical Switch Room UPS and Battery Room Generator Room Oil Store Storage and Build Area Delivery and Loading area Planning and Meeting Room Internal staff facilities Electrical Substation Air Conditioning Condensers External staff facilities Function House Computer racks and communication equipment All control and monitoring functions are concentrated Office are where IT staff can work Where all the external communication cabling enter the building Storage space depends on which gas is used Insert gaze or Halocarbons External power cables enter the building and forms a demarcation point. Loads of 100kVA is recommended to have a separate UPS room. House a standby Diesel Generator. House Diesel fuel for generator To store and unpack equipment to build Adjacent area to allow heavy equipment to be shipped into building To hold meetings Male/Female/Disabled toilets and showers Due to power needs a separate substation may be required A secure area is required depending on the air conditioning requirements Parking space, bicycle and smoke shelter. Table No. 1.1 Requirements of a Traditional Data Center Data Center Global Market Survey The survey made for the need of Data Center in the businesses proves all above data. The statistics we received 1. $21 billion market in $30.12 billion expected in %+ spending of IT and Facilities Managers budgets will go to this area. 4. Data Center expect growth for 2013/2014 will be 30-40% 5. Need of virtualization 76% Yes Other than this, there are few of the primary functions that a standardized business data center is expected to perform 1. Monitoring and maintaining properly functioning systems 24x7 2. Scheduling routine maintenance and inspections of all systems to ensure continued proper functioning 3. Scheduling periodic tests of emergency systems to certify proper functioning and assess capacity 4. Reviewing regularly authorizations for physical access to Data Centers 5. Planning space, including power and cooling assessments, for the installation of new equipment or the relocation of existing equipment 6. Investigating and assessing of new data center technologies (for example, aisle containment systems, power distribution) 7. Establishing standards and procedures for the use of Data Centers. On the other side, there are some supporting functions - 1. Environmental systems, such as HVAC units controlling temperature and humidity 2. Power systems, such as Uninterruptable Power Systems (UPS) for power stability and generators for backup power 3. Fire suppression systems 4. Physical security, such as the Authorized Entry via Card Access System 5. Physical space, including both integrity and use planning and control Now instead of going for all these expensive factors, adopting VIRTUALIZATION is the most sensible and affordable option iff our processors are capable enough to run the combination of both heavy-loaded and lightweight processes successfully. Architecture of Virtual Data Centers At present the technology era strives for Tier-III architecture when it come to the moment of designing a high performance, scalable and dependable data center; which indeed has become a critical issue because of the increasing use of the Internet in supporting various Web-based services. As the network bandwidth continues to increase faster than the server capacity, data centers are anticipated to be bottleneck in hosting network based services. It has been observed that the data centers contribute, to approximately 40% of the overall Blue Ocean Research Journals 7

5 delay, and this delay is likely to grow with the increasing use of dynamic Web Contents. The Tier- III architecture depicted as below Figure shows a typical three-tier architecture of data center. This architecture is a SOA i.e. Service Oriented Architecture. In this architecture, the frontend Web server is the closest to the edge of a data center and is responsible for handling the static requests. The middle tier, called the application server, is responsible for handling the dynamic Web contents, while the back-end database server is dedicated for complex database contents, while the back-end database server is dedicated for complex database transactions. Transactions.is modular, it also allows you to build redundancies into the configuration for an assumed Almost all clusterbased web servers/data centers use TCP/IP over Ethernet for supporting inter/intra cluster communication. Very few of them have exploited the advantage of switching to a user-level communication so as to enhance the communication performance. So this actually suits more to cloud. In the area of high-performance scientific computing employing clusters, it has been demonstrated that user-level communication mechanisms such as the Virtual Interface Architecture (VIA) can significantly reduce the communication overhead by minimizing the kernel involvement in message transfer. Basically this architecture is just a part of cores of architecture, as depicted by figure following Fig A Generic 3 Tier Architecture Data Center Fig3.2. : Industrial Different Data Center Tiers Blue Ocean Research Journals 8

6 Factor T1 T2 T3 T4 Activity Capacity N N+1 N+2 N+3 Comparison to support IT operation Load Distribution Path active 1 alternate 2 active Concurrency No No Yes Yes Maintainability Fault Tolerance(Single No No No Yes Event) Compartmentalization No No No Yes Continuous Cooling Load Density Dependent Load Density Dependent Load Density Dependent Load Density Dependent Availability 99.67% 99.75% 99.98% 99.99% Downtime per Year 28.8 hrs 22 hrs 1.6 hrs 0.8 hrs Basically in the proposed architecture, there is a main super-server that actually acts as a main management server and all the servers are consolidated inside this. The advantage of consolidating these many servers is that we can get maximum throughput from this and we also have mathematically well-improved percentage of overall performance. For instance if I invest $10000 today on the hardware of the server, the actual work done I am able to generate from it, is only worth $1000. That means exactly $9000 working capacity of the ever is getting wasted or we can say is remaining idle. So the proposed concept is we should virtualize the server. That means try to build maximum number of the servers virtually in a single powerful hardware, but all of them should be virtual. The advantage of virtualized servers is that we can restore them very easily and successfully by the feature called Snapshots of the systems this feature is available with almost every single Hypervisor such as KVM, Virtual Box, Table No. 2.1 Industrial Different Data Center Tiers VMware, etc. In this paper, we will be following diagrams of the abstract components of this architecture. By virtue of this, the throughput of any system can be approached to 80-85% with maximum uptime and availability. Using well defined interfaces facilitates independent subsystem development by both hardware and software design teams. The simplifying abstractions hide lower-level implementation details, thereby reducing the complexity of the design process. Along with these, there also floats one hidden objective and that is to realize Virtual Power. The goals of this virtual power management are : 1. To support the isolated and independent operation by guest virtual machines(vms) running on virtualized platforms. 2. To make it possible to control and globally coordinate the effect of the diverse power management policies applied by these VMs to virtualized resources. Fig. No.3.3 Virtual Data Center Network Architecture Blue Ocean Research Journals 9

7 Fig. No.3.4 Virtual Data Center Centralized RAID 5 Storage Figure 3.2 elaborates how the build of a Data Center should look like. The architecture we are trying to propose contains a super server that manages all the virtual machines and server that are consolidated in your single bare hardware. These contains file server, web server, mail server, application server, database server, etc. Now a Linux machine with a ZEBRA protocol configured on it over 2 NATed NICs can be used as a Gateway + Router. So that solves both the security and networking purpose for the outside world. We also propose to configure the DMZ (demilitarized zones) in our network architecture. I would like to make one point very clear the architecture we are planning to configure and deploy is centralized over the core requirements of educational institutions and colleges. This architecture solves the purpose of those institutions who are unable to invest more in hardware but still meticulously needs the requirements of the data centers in their premises. At the same hand the dedicated hardware-performance can never be achieved by their software alternatives installed on normal bare hardware. The next paper of this series (VDC) is planned to have complete build up steps, core configuration of different servers and maintenance. We look forward to have successful analytical comparisons and readings of the same. Conclusion Efficiency in the production of IT services is crucial in today s competitive markets. Virtualization and server consolidation can lead to increased utilization of hardware, and therefore, to increased production efficiency. Server consolidation poses a number of new and fundamental decision and planning problems in IT service management. Deploying and adopting to Virtualization Technology can server your problem a lot and the most noticeable thing is you are guaranteed to experience 99%+ uptime and very negligible down time. Workloads can change over time. It is important to analyze the workload traces regularly and re-optimize the allocation if necessary. If workloads are less predictable or exhibit significant trend, an automated controller can perform these tasks and move applications automatically from one physical server to another, allowing for adaptive, self-organizing data center management. The models discussed in this Blue Ocean Research Journals 10

8 Journal of Engineering, Computers & Applied Sciences (JEC&AS) ISSN No: paper can serve as the basis for respective controllers. We also believe that the increased proximity of the switch and the server can enable other new optimizations and are interested in looking at these further. Overall, as virtualization gets more widely adopted and networking and I/O performance become ever more important, we believe that approaches like ours that examine blurring of the boundaries between the computing and networking infrastructure and their tighter integration, are likely to be a key part of future designs. The propriety technologies like VMware vsphere 5.0, KVM and Virtual Box can also be considered for making your own Data Center. But mind well to have fully certified architecture set up for your purpose. The purposes may differ and so the architecture of every Data Center. No two data center can have exactly same architecture even though the purposes are different. References [1] Anycast-Based Optimizations for Inter-DataCenter Interconnections [Invited] Molka Gharbaoui, Barbara Martini, and Piero Castoldi B168 J. OPT. COMMUN. NETW./VOL. 4, NO. 11/NOVEMBER 2012 [2] DAC: Generic and Automatic Address Configuration for Data Center Networks Kai Chen, Chuanxiong Guo, Haitao Wu, Member, IEEE, Jing Yuan, Zhenqian Feng, Yan Chen, Songwu Lu, Senior Member, IEEE, and Wenfei Wu, IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 20, NO. 1, FEBRUARY [3] The Cost of a Cloud:Research Problems in Data Center Networks - Albert Greenberg, James Hamilton, David A. Maltz, Parveen Patel Microsoft Research, Redmond, WA, USA. [4] The Distributed Data Center: Front-End Solutions by Stefano Testa and Wesley Chou IEEE Computer Society /04/$ [5] ElasticTree: Saving Energy in Data Center Networks Brandon Heller, Srini Seetharaman, Priyahadevan, Yiannis Yiakoumis, Puneet Sharma, Sujata Banerjee, Nick McKeown, Stanford University, Palo Alto, CA USA [6] Communication within Clouds: Open Standards and Proprietary Protocols for Data Center Networking by Carolyn J. Sher DeCusatis and Aparico Carranza, New York City College of Technology Casimer M. DeCusatis, IBM Corporation /2012 IEEE IEEE Communications Magazine September [7] Converged Optical Network and Data Center Virtual Infrastructure Planning Konstantinos N. Georgakilas, Anna Tzanakaki, Markos Anastasopoulos, and Jens Myrup Pedersen VOL. 4, NO. 9/SEPTEMBER 2012/J. OPT. COMMUNICATION NETWORK /12/ /2012 Optical Society of America [8] Virtual Infrastructure Management in Private and Hybrid Clouds by Borja Sotomayor University of Chicago, Rubén S. Montero and Ignacio M. Llorente Universidad Complutense de Madrid, Ian Foster Argonne National Laboratory, University of Chicago - IEEE INTERNET COMPUTING. [9] A Mathematical Programming Approach for Server Consolidation Problems in Virtualized Data Centers by Benjamin Speitkamp and Martin Bichler - IEEE TRANSACTIONS ON SERVICES COMPUTING, VOL. 3, NO. 4, OCTOBER-DECEMBER [10] On Architecture Design, Congestion Notification, TCP Incast and Power Consumption in Data Centers by Yan Zhang, Student Member, IEEE, and Nirwan Ansari, Fellow, IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013 Blue Ocean Research Journals 11