Understanding Quality of Service (QoS)
Along with the advancement of information and telecommunications technology, the need for a network will increase, especially the use of IP on networks, especially the Internet. To measure the quality of TCP/IP internet or intranet network connections, QoS or Quality Of service is needed, where there are several methods to measure connection quality such as bandwidth consumption by users, connection availability, latency, losses, etc.
As we all know that QoS is very necessary for real-time applications in the Internet. A QoS can be described as a unit of parameters that describe the quality (for example, bandwidth, buffer usage, priority, CPU usage, and others) of a data. The basis of the IP protocol is to provide the best effort QoS. There are two main bases of QoS for the Internet and IP that are based on the network, namely: Integrated Services and Differentiated Services.
It is hoped that with QoS, the quality of a TCP/IP internet network connection can be measured or the internet can be accessed faster and better.
1. Definition of QoS
Quality of Service or QoS is used to measure the quality level of TCP/IP internet or intranet network connections. There are several methods to measure connection quality such as bandwidth consumption by users, connection availability, latency, losses etc.
Now let's discuss the terms in Quality of Service:
1.1 Bandwidth
Bandwidth is the capacity or capacity of an Ethernet cable to pass through a certain amount of data packet traffic. Bandwidth can also mean the amount of data packet consumption per unit of time expressed in bits per second [bps]. Internet bandwidth is provided by an internet provider with a certain amount depending on the customer's rent. With QoS we can set it so that users do not use up the bandwidth provided by the provider.
1.2 Latency
If we send 3 Mbytes of data when the network is quiet, it takes 5 minutes but when it is busy, it takes 15 minutes, this is called latency. Latency when the network is busy is around 50-70 msec.
1.3 Losses
Losses are the number of packets lost when sending data packets to their destination, the best quality of LAN/WAN networks has the smallest number of losses.
1.4 Availability
Availability means the availability of a web service, smtp, pop3 and application when the LAN / WAN network is busy or not.
1. Integrated Service
The Integrated Services (IS) model was defined by the IEFT working group to be the foundation of the Internet. This Internet architecture model includes best effort services and new real-time services that provide the functionality to reserve bandwidth on the Internet and networks. Integrated Services were developed to optimize network and network resource utilization for new applications, such as real-time multimedia, which require QoS guarantees. Because of routing delays and congestion losses, real-time applications do not work well on the Internet. Video conferencing, video broadcasting, and audio conferencing software require guaranteed bandwidth to provide audio and video of acceptable quality. Integrated Services make it possible to divide Internet traffic into standardized effort traffic for data applications and QoS guaranteed usage.
Figure 47. Data Order Control
Figure 7.12 shows the operation of the Integrated services model residing within hosts and routers.
Figure 48. Integrated Services Model
Integrated Services uses the Resource Reservation Protocol (RSVP) to signal reservations. Integrated Services communicate via RSVP to create and maintain data paths within host endpoints and within routers along the flow of a flow.
As shown in Figure 1 above, an application that wants to send a reserved flow data packet communicates with the RSVP reservation. The RSVP protocol tries to provide a reserved flow according to the requested QoS, which will be accepted if the application meets the policy restrictions and the router is able to handle the requested QoS. RSVP tells the packet classifier and packet scheduler at each node to process the packet according to the available flow. If an application sends a data packet to the classifier in the first node, which has mapped this flow into a specific class of service according to the requested QoS agreement, the flow is identified by the sender's IP address and the data will be transmitted to the packet scheduler. The packet scheduler will forward the packet, depending on the class of service to the next router and finally the data packet is received by the receiving host.
Because RSVP is a simple protocol, QoS reservations are made only in one direction, from the sending node to the designated receiving node. If the application in the example wants to cancel the reservation for data flow, the application will send a message to the reservation according to the QoS reservation in all routers along the path. The specification of the integrated service is defined in RFC 1633.
2. Differentiated Service
The differentiated services mechanism does not use per-flow signaling, and as a result, does not consume per-flow state during routing. Different service levels can be allocated to different services belonging to the same user group, which means that all traffic is divided into classes or groups with different QoS parameters. This reduces maintenance costs compared to Integrated Services.
The Differentiated Services (DS) concept is currently under development by the IEFT working group. The DS specification is described in several Internet drafts and is not yet available in an RFC. This section provides an overview of the ideas and basis for providing differentiated services on the Internet. A component of the DS is the Service Level Agreement (SLA). An SLA is a service contract between a customer and a service provider that specifies the details of traffic classification and the corresponding delivery of the requested service according to the customer's wishes.
3. Differentiated Service Architecture
Differentiated Services architecture Unlike Integrated Services, QoS guarantees are made with Differentiated Services that are static and stay long-term in the router. This means that applications using DS do not have to provide QoS reservations for specific data packets. All traffic passing through a DS-capable network can receive specific QoS. Data packets must be marked with a DS field that is interpreted by routers on the network.
1. Per-hop behavior (PHB)
The DS field uses six bits to specify the Differentiated Services Code Point (DSCP). This code point will be used by each node on the net to select the PHB. Two currently unused (CU) bits of the field are reserved. The value of the CU bits is ignored by differentiated services-compliant nodes, when PHP is used for received packets. Example of DS routing.
Figure 13. DS Routing
2. Organization of the DSCP
- There are several IANA considerations regarding DSCP. The codepoint space for DSCP distinguishes between 64 codepoint values. The proposal would divide the space into tree pools.
- Pool1 can be used for standard actions. The other pool may be used for local experimental use, with one of the two pools being reserved for local experimental use in the near future.
Table 1. DSCP pools
3. Differentiated Services domains
- The provision of QoS guarantees was not created for end-to-end connection specifications, but for the formulation of good Differentiated Services domains.
- Can represent different administrative regions or autonomous systems, different trust regions, different network technologies, such as cell or frame-based techniques, hosts, and routers.
- A DS region consists of boundary components that are used to connect different DS regions to each other and interior components that are only used within the region.
A DS area normally consists of one or more networks under the same administration. For example, a corporate intranet or an Internet Service Provider (ISP). The administration of the DS area is responsible for ensuring that sufficient resources are reserved and provisioned to support the SLAS offered by the area. The network administrator must use measurement techniques to monitor if the network resources within the DS area are sufficient to satisfy all QoS requests.
Figure 49. Use of internal components and boundaries for two DS regions
3.1 DS boundary nodes
All data packets passing from one DS area to another must pass through boundary nodes, which can be a router, a host, or a firewall. A DS boundary node handling traffic leaving a DS area is called a boundary node and a boundary node handling traffic entering a DS area is called an ingress boundary node. Normally, DS boundary nodes act as both ingress nodes and nodes, depending on the direction of the traffic.
4. Components found in traffic conditioners
4.1 Classifier
The classifier selects packets based on their packet headers and forwards packets that satisfy the classifier rules for further processing. The DS model defines two types of classifier packets:
- Multi-Field (MF) Classifier can classify on DS fields as well as on other IP, for example, IP address and the port number, like RSVP.
- Behavior Aggregate (BA) Classifier, only classifies the bits in the DS field.
4.2 Meter
Traffic meters measure if the packet delivery was selected by the classifier according to the traffic profile that describes the QoS for the SLA between the customer and the service provider. A meter passes status information to another conditioning function to trigger a specific action for each packet, whether it does or does not comply with the requested QoS requirements.
4.3 Markers
DS sets the DS field of the next IP packet for certain bits. PHB is set in the first 6 bits of the DS field so that marked packets are delivered within the DS area according to the SLA between the subscriber and the service provider.
4.4 Shaper/dropper
Packet shapers and droppers cause conformity to some configured traffic properties, for example, bucket filter tokens, such as "Service classes". They use different methods to bring the flow into compliance with the traffic profile. The shaper delays some or all of the packets. Droppers generally have a finite-size buffer, and packets cannot be dropped if there is enough buffer space to hold the delayed packets. The dropper discards some or all of the packets. This process is known to maintain flow order. A dropper can be implemented as a special case of a shaper by setting the shaper buffer size to zero packets.
4.5 DS interior components
The interior components in the DS area select the delivery method for the packet based on its DS field.
Figure 50. DS Interior
5. Source domains
Traffic sources and intermediate nodes within a source area may exhibit traffic classification and conditioning functions. Traffic sent from a source area may be marked by the traffic source directly or by intermediate nodes before leaving the source area.
The first PHB that marks the data packet is not done by the sending application itself. The application does not order the availability of Differentiated Services in a network. Therefore, applications using the DS network are not rewritten to support DS. This is an important difference to Integrated Services, where most applications support the RSVP protocol directly when some code changes are required.
Figure 51. Initial Marking on Data Packages
6. Conclusion
QoS is essential for real-time applications on the Internet. A QoS can be described as a unit of parameters that describe the quality (for example, bandwidth, buffer usage, priority, CPU usage, and others) of a data. It is expected that with QoS, a network can measure the quality of the TCP/IP internet network connection or the internet with the access process being faster and better. Quality of Service or QoS is used to measure the level of quality of the TCP/IP internet or intranet network connection. There are several methods for measuring connection quality such as bandwidth consumption by users, connection availability, latency, losses, etc. QoS control methods include Packet Scheduler, Token Buchet Filter, First In First Out, Randomly Early Detection. The properties of QoS include Integrated Service and Differentiated Service, as well as Differentiated Services Architecture.
7. QUESTIONS
- Why is QoS so necessary in internet networks?
- Explain with a picture how disciplined queuing packets can skip the data packet queue?
- Explain the differences in the QoS properties of differential services and integrated services?
- Explain how the classifier packet divides the IP into classes?
- Why does integrated service use RSVP in data ordering?
How to Control QoS
1. Packet scheduler
Packet schedulers manage the delivery of different packet flows within hosts and routers based on class of service, use of queue management and various scheduling algorithms. Packet schedulers must ensure that packet delivery conforms to the QoS parameters for each flow. A scheduler must also be able to maintain order or shape traffic to match a particular level of service. Packet schedulers must be implemented at the point where packets are sent.
The scheduler package is also called queuing discipline. Queuing discipline is a queue in each ethernet card that is used to store data packet queues, incoming or outgoing data packets via qdisc. Data packets entering the qdisc will be separated by the filter section to determine the port / ip address that will be regulated for traffic flow. The class or traffic classification section will be discussed in the next section, while the purple qdisc is used to output data packets to the ethernet card.
Each network device has a queuing discipline that is related to QoS. The main function of the queuing discipline is to control how packets are enqueued on a particular device.
2. Token Buchet Filter (TBF)
Token bucket filter (TBF) limits bandwidth with the shape & drop method, the working principle uses a flow of tokens entering the bucket at a constant rate, if the token in the bucket runs out then the data packet will be queued and the excess is discarded, each data packet issued is identical to a token. Tokens in the bucket will run out faster if the flow of data packets exceeds the speed of tokens entering the bucket, so we assume that traffic exceeds the configuration limit.
Figure 41. Token Bucket Filter
TBF Parameters
- The bandwidth limit rate is set by the administrator, if the data packet flow exceeds this value then the data will be discarded (dropped) or delayed, the bandwidth will be cut.
- Limit / latency limit is the number of bytes that can be queued before the token is available, while latency is the length of time (in milliseconds [msec]) that a packet can be queued.
- Burst / buffer / maxburst Bucket capacity in bytes, data packets exceeding this value will be discarded or delayed.
- Peakrate The maximum rate limit handles momentary bandwidth spikes provided that data packets cannot exceed the bucket and MTU capacity.
3. First In First Out (FIFO)
The FIFO queue technique refers to FCFS (First Come First Server), the first data packet that arrives is processed first. The data packet that comes out first is entered into the FIFO queue, then issued according to the order of arrival. The FIFO queue technique is very suitable for networks with medium bandwidth of 64kbps but is quite consuming processor and memory resources.
Figure 42. FIFO queue
The image above shows the arrival of several data packets at different times, the first packet (1) from flow 8 that arrived earlier is issued to the port first by the FIFO queue. To set the queue we need the "tc" command with qdisc pfifo, the limit parameter to determine the maximum limit of the queue.
In the FIFO method, if the traffic exceeds the set value, the data packet will be entered into the queue, the data packet is not discarded, it is only delayed for a while. The FIFO method is suitable for internet connections with medium bandwidth of 64kbps, to avoid bootle neck on the LAN network. Data packets that exceed the configuration limit will be entered into the queue and when the LAN network is not busy, the data packets in the queue will be issued.
4. Random Early Detection (RED)
Random Early Detection or can be called Random Early Drop is usually used for gateways / backbone routers with very high traffic levels. RED controls network traffic to avoid congestion during high traffic based on monitoring changes in minimum and maximum queue values. If the queue content is below the minimum value then the 'drop' mode does not apply, when the queue starts to fill up to exceed the maximum value then RED will randomly discard (drop) data packets so that congestion on the network can be avoided.
The RED parameters are as follows
- min = Minimum average value of the queue.
- max = Maximum average value of the queue, usually twice the minimum value or with the formula; max = bandwidth [Bps] * latency [sec]
- probability = The maximum number of data packet marking probabilities ranges from 0.0 to 1.0.
- limit = The highest limit of the queue in real terms, the number of data packets that exceed the limit value will definitely be discarded. The limit value must be greater than 'max' and is expressed by the equation. limit = max + burst
- burst = Used to determine the speed of calculating the antrain value affecting the real queue (limit). For practice, we set the value with the formula; burst = (min+min+max) / 3 * avpkt
- avpkt = Average value of data packets / traffic passing through the RED gateway, should be filled with 1000.
- bandwidth = Bandwidth of the ethernet card.
- ecn = Explicit Congestion Notification provides the RED gateway with the facility to notify clients if congestion occurs.