ABSTRACT Cloud computing is being projected as the next generation of the digital age. Many still have a question to what is cloud computing. The cloud is a general term that was used to describe the telephone system. The actual complexity of a telephone network is in its network not in its end- to-end systems which just act as dummy nodes. The internet which has come to be so widely used is based on computers which have a much level of complexity than a telephone set. Therefore the designers of the internet found it easier to make a simple network which just forwards the data from one place to another. But now we return once again back to the old telephone style data transfer where most of the complexity lies in the communication between the end user and the server. Cloud computing is not something which is entirely new or entirely different from the internet. But it requires much faster data transfer and lower latency than the internet. Therefore some of the protocols followed may not be fully compatible with this. It has the characteristics of client-server model, peer-to-peer model, grid computing, utility computing etc. That means newer protocols and newer standards have to be invented and implemented in order to make cloud computing really effective. The industry right now is much hyped about cloud computing. But the network which essentially is the backbone of cloud computing is not fully ready yet to support this new technology. Still much development is going on and newer and more efficient protocols are being developed. In this report we explain why some of the protocols used in the internet have potential drawbacks in cloud computing and also what are the newer protocols to come such as FASP as well as a brief introduction to cloud computing and its characteristics. INTRODUCTION A cloud is generally used to describe the internet. It is a collection of web-based technologies which would enable user to access or use the applications from any computer. Cloud computing is not an entirely new technology or trend. Roaming profiles for employees within a corporate network which saves their favorites and their documents or applications which are based on web browsers are used now. But what makes cloud computing so radical is that till now what all has been going on has just used the internet for communication and not much for applications. The usage of software services is defined by cloud computing as a communication between the client and the cloud. A cloud can be defined as a pool of resources from which a user picks the resources needed. So basically the user will be paying only for the necessary features and/or for the amount of usage. The five essential characteristics defined by the market leaders are On- Demand self-service, broad network access, rapid elasticity, resource pooling and measured service. Cloud computing is essentially the internet. Since the amount of data and computing power sent over the network increased so tremendously, the network itself has to have the capability to cope up and deliver required performance. Cloud computing has been broken down into three segments based on the services. They are software as a service, platform as a service and infrastructure as a service. One must remember that cloud computing is not an entirely new technology but derives many of its characteristics from technologies like grid computing, utility computing, peer-to-peer, client-server model etc. Cloud computing is still an evolving technology, one whose boundaries and specifications are not fully defined yet. But there is a lot of potential in this and this can be seen as the future of many software companies. Hence there is a lot of demand for reliable cloud providers and some of the major industry players are Amazon, Google, SalesForce.com, Microsoft, IBM, etc. Cloud computing is termed as the next generation technology one which we can see rising up to become the benchmark for software in the next couple of years. Difference between cloud computing and hosted services Hosted services are commonly used in corporate environments which enable a user to remotely login to a system (called “roaming profiles”). In hosted services the storage, networking and servers are shared among users in the local network or through a remote server. The scaling in this network is done manually by the network administrator. Cloud computing differs by the fact that if offers an on demand service which is paid by the amount and time of usage and the scaling is done based on the data flow which is determined during the usage. In hosted services the data flow is fixed and cannot be altered by the user and also only a set of features is available and the user cannot have remote access to services that are not installed in his profile. Network Automation: Even though technology has advanced such that virtual servers can be deployed in a matter of seconds, the process of constructing required infrastructure such as assigning IP addresses, management of DNS records and virtual assets (virtual machines etc.,) for networks has to be still configured manually. Network automation plays a critical role in cloud computing. Because devices connecting and leaving the network are in huge numbers for a highly populated cloud, it is not possible that the old practices of IP address management and DNS services can be applied to them. Cloud computing requires the network to handle dynamic IP address allocation as well as the DNS records to be updated constantly. Virtualization causes more complexity for network management and the solution for all these is network automated. Data Centers are the heart of networking in cloud computing. The performance of the cloud depends on the efficiency of the data center. Automation plays a major role in better maintenance of virtual machines. Previously data center managers have to co- ordinate between server, network and storage resources for a given environment. But this cannot be applied to cloud whose environment changes so frequently and dynamically. It would be a waste of manpower to attempt manual configuration. This is not limited just for change in the virtual environment but also for the management and operations of the data center. Groups of staff taking care of separate areas such as network, server and storage resources cannot work independently. In order to quickly reorganize the environment as required by business these groups must work with more coordination and communication. Network automation can improve operations through the following ways: The data flows vary greatly in cloud computing which maybe low at one moment and suddenly due to the deployment of resources it may rise to a peak exhibiting unpredictable bursts of traffic. In order to increase the performance of routing and switching to reduce the end delay of the transaction a congestion free network is needed. In case of outage the network must be able to shift its resources almost instantly without any loss. Due to the large amount of connections flowing in and out, the architecture can be made less complicated by simply increasing port density. This increases the speed of access for the connections as well as less hardware and simple architecture. On demand allocation of resources is also possible through automated network configuration. It can be capable of performing such functions as VLAN management, applying QoS policies, dynamic bandwidth allocation and reducing power consumption by effectively utilizing the resources. FASP (Fast and Secure Protocol) TCP has proved to be the backbone of the internet. It offers reliable data transfer which prevents data loss. Though this is useful for applications which do not tolerate data loss, increasing amount of data transfer is for multimedia streaming which is loss tolerant but time dependent. Therefore it uses UDP. But UDP also has some shortcoming regarding this. Both the protocols are compatible with the cloud network but they are hardly efficient. Therefore there is a need for a new protocol that support the cloud computing and handles the tremendous speed it requires as well as provide a sufficiently reliable transfer protocol. TCP is essentially made for the internet therefore it does its best in sending relatively small amounts of data over short distances. But Cloud computing involves transfer of large amounts of data through large distances. Therefore TCP can be a cause for bottlenecks in clouds. The number of errors increases with the amount of data and the distance travelled. Since TCP is a reliable transfer protocol if there is a loss of 1% a 100Mbps connection will get reduce to less than 10Mbps for large distances. Many new and innovative data transfer protocols have been proposed compatible to cloud computing. One of them is Fast and Secure Protocol (FASP) which is going to be used for Amazon elastic cloud. “FASP is an alternate bulk data moving protocol” says Michelle Munson, Aspera’s CEO and co-founder. FASP does not wait for acknowledgment from the client but instead assumes that all packets have reached their destination. FASP only resends those packets which have been conformed as dropped. “Instead of sending lots of small packets, it sends fewer large packets” says Simon Hudson from cloud2. This uses the network bandwidth more efficiently while supplying more data at a faster rate. Traffic management is also a criteria that has to been handled for cloud computing. Because of the large amount of data being transferred between the client and the server, packets can be easily blocked or lost due to poor network traffic control. FASP solves this problem by monitoring all the data that is passing into the network and changing the size of the packets and the order and speed at which they are sent depending on the available bandwidth and other parameters. This method ensures that the flow of data is regulated and the network does not get saturated so as to cause loss of data due to congestion. Platform as a Service (PaaS) Platform as a service the legal transfer of a computing platform and solution stack as a service. PaaS fling nearby entries the distribution of forces in preparation of application without the cost and complexity of purchasing and managing the inherent hardware and software and providing users with access to data and technology resources. Giving all of the facilities the full needed to sustenance the oscillation of edifice and delivering specific use of web and services fully useable from the internet. PaaS proffering may contains facilities for diligence design, diligence development,examine,deployment and hosting internet service.Moreover,it also offers storage, application instrumentation, database integration. There are four types of PaaS:- 1) Open Chopine as a Surround: - allows programmer to select any OS,any programming language, any server, any database. 2) Diligence delivery-only surround:-allows only few programs to run. 3) Stand alone development surround:-Open source. 4) Append development facilities:-Someway are equivalent to macro language. Key Characteristics:- 1) Service to take complete care of the merged development surround 2) Internet supported user port innovation tools 3) Multi-renter architecture 4) Merged along internet services 5) Pillar for development squad coactions 6) Utility form orchestration. Computing Platform A computing platform is classification of hardware architecture & software framework that permit software to pass. Normally platforms include a computer’s architecture, OS, software languages and associate client interface. A platform is a pivotal element in building programs. It can be defined as an assignment to establish software .It really intends execution of the software is not control by the type of OS given. OS examples: - Linux, Mac OS, Microsoft Windows, Solaris etc. Hardware examples:-Super computer, RISC, Graphic Processing Unit (specially for video game) etc. Solution Stack Solution stack is a place of software classification or objects needed to fetch amply structural solution.eg:-POS (product of service) For example:-to create a web application it requires a web server, software language, database, os. SaaS (Software as a Service) It is also mean as “software on claim”,is software that is position over the web and is position to zip down a firewall on a LAN or PC.With SaaS,a provider authorize an application to customers either as a service on demand,done with a subscription,in a “pay as you go” model,or at no charge. This near to application delivery is part of the substitute computing model where all of the technology is in “cloud” access over the web as a service. Infrastructure as a Service (IaaS) Cloud infrastructure service is also known as “Infrastructure as a Service”,typically platform virtualization surround as a service.Instead than buying servers,sprogramming language,etc..Users alternatively purchase those resources as a completely sourced service.Suppliers typically bill such services on a utility computing and amount of resources with them.Iaas developd from virtual private server. IEEE 802.1Q Protocol IEEE 802.1Q or VLAN Tagging was devised by the 802.1 work group for sharing of the physical Ethernet link among a number of independent multiple logical networks. This protocol forms the underlying principle of the CISCO NEXUS 1000V switch which is most commonly used for cloud computing. Together with the VMWare VSphere Operating System, this protocol along with QoS, PVLAN, ERSPAN and CLI forms the heart of the cloud computing environment. The following HYPERVIZOR diagram actually shows the implementation of the IEEE 802.1Q protocol in the CISCO NEXUS 1000V switch. This protocol explains the meaning of the Virtual LAN with respect to bridging between the Medium Access Control layer and the IEEE 802.1D-The spanning Tree protocol. This protocol allows multiple nodes on different Virtual LAN’s to communicate between each other through a network layer switch. Medium Access Control Data Communication Protocol: This layer, also known as MAC, is a sub-layer of the Data Link Layer. It provides the addressing and the channel Access Control mechanism so that multiple nodes or different terminals are able to communicate in a Local Area Network. The hardware that is used for this is known as the Medium Access Controller. Addressing Mechanism: This mechanism provides a unique MAC address to each device in the network. This ensures that the packets are delivered to the proper device because even though the devices may have different IP addresses, it is possible that they may be in the same sub-network. Thus this type of addressing mechanism assigns a unique address to the device even within the subnets. To be more specific, there would not be any other device in the world with the same address as yours(analogous to your phone number). The address usually consists of 6 groups of hexadecimal digits or numbers which are separated either by a colon (:) or by hyphen (-) in the order of transmission. Ex. 01:05:ab:5c:7d:8e Channel Access Control Mechanism: This mechanism does the job of making it possible different nodes or stations which are connected to the same physical network to share it. Some of the examples of the networks are ring, bus, hub etc. This is also known as the Multiple Access Control Protocol. It either detects or avoid collisions among the packets if the packet mode contention channel access method is used or reserves resources if a circuit switched channelized Control Access Method is used. One of the examples of the Channel Access Control Mechanism is the CSMA/CD(Carrier Sense Multiple Access with Collision Detection). The 802.1D Spanning Tree Protocol This protocol was devised to ensure that there is no looping in a bridged network especially in a bridged LAN. Bridging is usually used in packet switched networks and makes no assumptions on where a particular address is located in the network. It uses the concept of flooding and the source address in the packet header to locate the device. Once this device has been located, its address is stored in the MAC address table so that it need not be broadcasted again. Bridging constrained by the fact that it uses flooding, so it is limited to LAN’s. This protocol is implemented widely in the 802.1Q protocol to avoid bridging loops when nodes in different VLAN’s As the name of the protocol suggests, it creates a spanning tree within a mesh network of bridges and disables those links that are not a part of the tree which would leave us with only one active path between two active nodes. A spanning tree can be explained by the following diagram. Here, all the vertices of the grid are connected but note that no cycle is formed, which forms the basis of the spanning tree protocol. Our main aim is not to form bridging loops in a network and avoid further broadcasts as well as multicasts. It can also be said to be the maximum set of the edges of the grid which connects all the vertices or the minimum set of the edges which forms no cycle. The network diagram shown above will be used to explain the spanning tree. Here the the boxes represent the bridges in the network whereas the numbers in these boxes represent the bridge id and the clouds represent the network segments. The spanning tree is calculated by the following steps. Find the root bridge-The root bridge is the one with the minimum bridge id. #3 in this case. Each of these bridges have a unique bridge id and a configurable priority number. For comparison of 2 bridge id’s, their priority is compared first. If they are equal, then their MAC address is compared. For example, if two bridges have a priority of 10 and one has a MAC address of 0300.0000.2222 and the other has a MAC address of 0300.0000.4444, then the previous one will be chosen as the root bridge. Select the least cost path to the root bridge-The tree has the property that any message from a device in the tree will traverse the least cost path from that device to the root. The cost of traversing a path is the sum of the cost of that path. This is governed by the following rules:- After the root bridge has been chosen, each of the bridges in the network determines the cost of each possible of itself to the root bridge. From these, it chooses that path which has the least cost. The port that connects to that path would then become the root port (RP). The bridges on the network segment determine the least cost path from the network segment to the root. The port which then connects this bridge to the network segment becomes the Designated Port (DP). Any port which is neither a root port nor a designated port becomes a Blocked port (BP). Example Application of the VLAN Tagging Protocol: Consider for example that a company has many departments and it wants to conserve the resources by using only one physical network and assigning different logical networks for each of the departments that the company has. For this, the network administrator assigns a unique VLAN to each department. An Edge device, which are particularly routers assign a frame address to each data frame that is received from a given department. After this frame is transmitted to the respective VLAN, this tag is removed from the data frame before transmitting it to other computer in the same department. This method is analogous to the one that is used in the Cloud Computing because the main aim of cloud computing is to use a centralized physical resource and nodes in different networks would share these resources thus forming a logical connection. The frame address format that is used is as follows:- Tag Protocol Identifier(TPID)-This field is used to distinguish between tagged frames and untagged frames and is located in the same position as the Ether Type field. It is set to a value of 0×8100 to identify it as a tagged IEEE 802.1Q frame. Priority Code Point (PCP)-This is a 3 bit field and it indicates the 802.1p priority. This field is used to distinguish between the priority of different frames with 0 being the best effort and 7 being the highest. This can be used to prioritize between different types of traffic i.e. voice, data, video etc. Canonical Format Indicator (CFI)-This is a 1 bit field which indicated whether the MAC address is in canonical format or not. A canonical format indicates that the least significant bit is transmitted first. Most of the modern devices use the canonical format. VLAN identifier (VID)-This is a 12 bit field which specifies the VLAN to which the frame belongs. If this value is 0, then it does not belong to any VLAN and the 802.1Q tag only contains the priority. The hexadecimal value of 0xFFF is usually reserved. The rest of the values can be used allowing up to 4094 tags. Reference http://blogs.cisco.com/sp/announcing_cisco_infrastructure_as_a_service_iaas_solution_for_service_pr ov/ http://www.strongmail.com/resources/blogs/it_email_infrastructure/2010/02/saas-paas-iaasfaads.php http://www.google.com/#q=cloud+computing&hl=en&prmd=ivnbl&source=univ&tbs=nws:1&tbo=u&ei =giHjTIC0BpLCsAPi2- Bm&sa=X&oi=news_group&ct=title&resnum=4&sqi=2&ved=0CGIQqAIwAw&fp=7193df428e16df5e www.hyperwiser.com www.
Pages to are hidden for
"project"Please download to view full document