QOS Research Paper
The effect of So in large enterprise networks is more apparent than anywhere else, they need So for their networks to run properly, and with vast amounts Of users there are bound to bottlenecks; SO was designed to allow for effective and efficient mitigation of such problems. So did not come about as an afterthought; it was foreseen during the creation of the IP protocol that such controls were going to be needed eventually and is the reason the TOSS (Type-of-Service) byte was implemented. IP SO) A non-So network generally has four main issues that are resolved tit So; the characteristics of data flow that are used are bandwidth, delay, jitter, and loss. Multiple different methods of applying So are used in large networks, these include RSVP based, IP precedence/Diffuser and MILS and packet shaping. (CNN) The internet today runs on best-effort system, A best-effort service makes no guarantees regarding when or whether a packet is delivered to the receiver, though packets are usually dropped only during network congestion.
The founding fathers of the internet predicted that such a tool would eventually be necessary in most if not the whole internet. They included in the original IP headers a TOSS (Type-of-service) byte “The Type of Service provides an indication of the abstract parameters of the quality of service desired. These parameters are to be used to guide the selection of the actual service parameters when transmitting a datagram through the particular network” (IP SO).
The Internet until the late ass’s and early ass’s was largely based in academic institutions and had a very limited amount of traffic and applications running on it, many implementations of the IP protocol ignored the TOSS byte, as it just was not necessary. The Internet connectionless design allows for much flexibility and robustness but also is prone to congestion problems this was especially problematic at routers that connected networks of varying bandwidth.
The Angle algorithm, which solved this issue, is supported by all IP host implementations. The Angle algorithm brought about the beginning of Internet So-based functionality in IP. A man named Van Jacobson developed the next set of standard Internet So tools, congestion avoidance mechanisms for end systems that are now required in TCP implementations. These methods: slow start and congestion avoidance helped greatly in preventing congestion collapse in the present-day Internet.
They primarily make the TCP flows responsive to the congestion signals (dropped packets) within the network. Two more methods: fast retransmit and fast recovery were added in 1990 to provide optimal performance during periods of packet loss. After this focus shifted to the interaction between routers and end systems, WFM, a packet scheduling algorithm, and WERE, a queue management algorithm, were widely accepted to fill this gap in the Internet backbone.
SO is extremely important because without it the internet we know and love would not function, congestion problems would plague all access points and real time traffic such as video and voice calls would never be able to reach their intended target in time. So allows for critical data flows to reach their destinations generally more quickly than they otherwise would have. So service currently had reached a point at which networks are able to be effectively managed with existing tool and much of the current research is being done on making everything more efficient impaired to just making things work.
An interesting piece of research being done recently is a project called DECOR in which instead of a centralized controller that gathers network information and the relative levels of usage of different resources and calculates optimized task allocation arrangements to maximize some global benefit uses a distributed network of nodes that performs the same functions more locally allowing for there to be no central point of failure. (DECOR) Going into the future So will have further advantages as IPPP is in the process of being implemented as well as the genealogy that the SO is used in conjunction with.
Network speeds as well as the speeds at which routers and switches can process and throughput data will continue increasing though the speed of light will not which is the true speed limit in the end. The basic problems that So is able to solve are: What applications need from the network, How to regulate the traffic that enters the network, how to reserve resources at routers to guarantee performance and whether the network can safely accept more traffic. (CNN) These problems can be solved through the use of tools provided to us in the So suite.
The data flow of a specific service is comprised of the characteristics of: bandwidth, delay, jitter, and loss. Certain applications and service require the flow characteristics to fall within certain values. The bandwidth of a specific flow defines the specific bit rate or the Max throughput for certain medium. If an application requires a certain amount of bandwidth to function properly, the settings upon the router/switch must be appropriately set to allow for adequate So. Services such as Nettling and Hull need high bandwidth to be able to be consumed properly.
Delay in a network must be carefully considered; as it is is an important design and performance characteristic. Measuring delay in a network is the amount of time it takes for a bit to travel from one endpoint to another. The delay or ping as gamers call it is generally measured in milliseconds, although if the nodes are very far apart it can reach seconds. When an engineer is working on a specific networks architecture they must measure multiple types of delay they are: processing delay, queuing delay, transmission delay, and propagation delay.
Processing delay is the amount of mime a router takes to process the packet header. Queuing delay is the time that the packet spends waiting in a specific queue. Transmission delay is the time it takes for the packet to be sent into the medium and propagation delay is the time for a signal to reach its final destination. Services such as videoconferencing and Vivo need very exact delay to ensure that communication is clear and not disarming to participants. Jitter is the packet delay variation. It is the difference in end-to-end one-way delay between selected packets in a flow with any lost packets being ignored.
A real life example Of this would be sending a letter every day where sometimes its takes 2 days for the letter to be delivered and other times 1 or 3. If the expected delivery is 2 days arriving in 1 or 3 days causes an VIVID of +1 and -1 respectively. The positive number is an effect called clumping and the negative known as dispersion. The jitter of a real time applications data flow must be consistent for the data to be correct and understandable, this would be and application such as Keep or first person shooter video games.
Loss identifies the amount of packets being lost by the network during remission. Generally loss will occur at congested network points and when packets become corrupt in transmission. These congestion points drop packets when the incoming packets outpace the queue size limit of the output queue. Occurrences are also common due to insufficient input buffers during packet arrival. Packet loss is generally specified as a fraction of packets lost while transmitting a certain number of packets over some time interval.
Specific applications cannot function correct when packets are lost these include any sort of real time data flow, these are generally UDP connections, while TCP has built in functions to help deal with packet loss allowing for looser requirements. Generally packet loss will be rare on a network with correct architecture, packet drops are expected on the internet at large though as it is a best-effort system. To ensure that an application is receiving its specific needs from the network theses characteristics are all used.
To regulate the traffic coming into the network a technique known as traffic shaping is used. This allows for the data coming into a specific network to be averaged out and to limit burliness. The goal of the network is to ensure an environment in which a wide variety of traffic is free to be sent, including bursts yet allows for everyone to receive a proper amount. The network will dynamically adjust data flows to ensure that the distribution is fair for the service that is being used. Traffic shaping reduces congestion on the network and helps to ensure all clients an adequate connection.
When a client or set packets begins to be in excess of their proper share they may be given lower priority or dropped completely. When this is performed it is called traffic policing, Winnow State shapes the traffic around campus to ensure the outwork is being used for legal purposes as well as so that everyone has a fair amount of bandwidth. Routers/switches use an algorithm known as the leaky bucket to ensure that short term bursts of traffic are allowed but lengthy ones eventually limited. The lengthy bursts will smoothed out by the algorithm which decreases congestion on the network.
This technique is used to ensure that routers/switches will have the necessary resources to accommodate all traffic. The network also must decide if it can take on additional connections, it reaches it decision after considering its current opacity and the commitments it has made to other data flows. If the additional flow is accepted the network will reserve capacity in advance at routers further down the line. The network must ensure all routers in the network are prepared for the new flow other wise congestion can Occur and break the So guarantee.
Algorithms are used to find the single best path between a source and destination. If one cannot be found that is generally when a new connection would be rejected. Large networks need quality of service because of the stability it provides for he network and more specifically for certain applications. Mission critical applications such as financial packages, sales engines, and portal sites, and any application that drives an organization’s cash flow or operations should be given priority across the network infrastructure.
Large organizations and businesses must also understand that most of the time their network will not be under 100% load but that when it is they must prepared. So allows one to ensure a network can handle real-time data activities. The organization must firmly take into account three concepts: networking provisioning, ensuing, and classifying. (WAR SO) Network provisioning is the amount of bandwidth that is available to the traffic that may occur on a certain network.
When the telephone network was created the designers over provisioned immensely to ensure that calls would always have enough bandwidth, the manner in which IP protocol was designed allows for over provisioning to occur but not on the same scale as with the telephone system. Generally traffic that stays within an organization will be over provisioned which ensures that there is always an overabundance available for application, seers, services, etc. Large networks begin to have problems when data needs to flow into the Internet as a whole.
Costs become higher and higher to achieve an over provisional amount of bandwidth. So allows for this bottleneck to be overcome, specific real time data can be targeted and given priority over less important data flows. Certain activities such as telephone and video conferencing require that all participants receive a specific amount of network throughput and latency to allow for efficient communication. These services generally run over JODI, which means they are unacknowledged and the packets are just expected to arrive without confirmation.
There are certain amounts of latency, jitter, and bandwidth that are necessary for the human mind to understand audio and video effectively. The TIT (International Telecommunications Union) has created a standard for end-to-end delay (G. 114) that will allow for satisfactory communication between two parties for audio and video communication. (WAR SO) This standard G. 1 14 states the end-to-end delay should be no more than 150 milliseconds (ms). Though further investigation has shown that up to 200 ms is okay. Jitter should not exceed 50 ms, and total delay with end processing occurring should not exceed 300 ms.
SO is used explicitly in managing data flows such as these to ensure an optimal connection. Queuing is another aspect that So addresses, when data is sent through a network it goes it may go through only one switch/router or many. When a data flow reaches said switch/router it is many times put into a queue before being sent out again (if the interface is at full utilization this occurs), this queue can cause major problems for real time data, if the router/switch’s buffer overflows room too much bandwidth packets can be lost.
If say a large data transfer is occurring at the same time as an audio call on a network without SO, the audio call may seem disjointed and choppy which will cause difficulty in communication. This issue is solved through the use of priority queues, which work by having several different queues of which the router/switch will sort the packets into. The highest priority queue will empty first followed by the next and so on. This can cause a network to reach 100% utilization, which can cause problems for lower priority applications. IP SO) Many times to avoid jacket loss a system architect will increase the buffer sizes for the routers. This is counter intuitive as the TCP protocol was designed to experience packet loss. The real time LID packets are not made better by the increased buffer size either, to solve this issue the network must be setup in a way that real time will always have bandwidth available to ensure delivery and effective connections. This can cause major problems; an example of one is a TCP request that ends up duplicated. The situation develops as follows. A sender transmits data using TCP.
