24 - Terms - Quality of Service (QoS)

Command to reclassify CoS priority value for hosts to assign to its data packets

"switchport priority extend cos [#]"

Command to instruct device to trust CoS priority of incoming packets

"switchport priority extend trust"

Class of Service (CoS)

A prioritization value used to apply to services, ports, or whatever a quality of service (QoS) device might use. usually used with Ethernet 802.1q frames and contains 3 bits

QoS Model: DiffServ (Differentiated Service)

A technique for ensuring QoS by prioritizing traffic, taking into account all types of network traffic, not just the time-sensitive services such as voice and video. packets should be marked as close to the edge of the network as possible so that core network devices and other devices along the forwarding path will be able to quickly determine the proper QoS treatment to apply to a given traffic flow, based on the PHB that is associated with the DSCP marking Provides "almost guaranteed" QoS while still being cost-effective and scalable Not considered "end-to-end" since it is not guaranteed DiffServ places information in the DiffServ field in an IPv4 packet. In IPv6 packets, DiffServ uses a similar field known as the Traffic Class field. This information indicates to network routers how the data stream should be forwarded.

QoS Policy: network audit

Accomplished by deploying classification tools and conducting interviews with different departments to determine which applications are utilized the most Ex. Network-Based Application Recognition (NBAR), Netflow, or packet sniffers

DiffServ Model: architecture

Architecture is based on a simple model in which traffic entering a network is classified at the boundaries of the network The traffic class is then marked, using a DSCP marking in the IP header packets with the same DSCP markings create BAs as they traverse the network in a particular direction, and these aggregates are forwarded according to the PHB that is associated with the DSCP marking Each DSCP value identifies a BA Each BA is assigned to aPHB Each PHB is implemented using the appropriate QoS mechanism(s)

4 characteristics of marking (coloring)

Based on classification, metering, or both so that other network devices have a mechanism of easily identifying the required treatment Allows network devices to easily classify a packet or frame based on a specific traffic descriptor Typically performed as close to the network edge as possible When an end device marks a packet, a switch or router has the option of accepting or not accepting values from the end device (may reclassify)

3 models for QoS

Best Effort IntServ (Integrated Services) DiffServe (Differentiated Services)

Mitigate End-to-end delay (fixed and variable)

Best way to mitigate this is to increase the link capacity Alternatives (due to cost/time) include queuing or compression techniques

Mitigate lack of bandwidth

Best way to mitigate this is to increase the link capacity Alternatives (due to cost/time) include queuing or compression techniques Bandwidth across a network is limited by the lowest-bandwidth circuit and the # of traffic flows competing for the bandwidth on the path

QoS Guidelines: (3) Perform QoS in hardware rather than software when possible

Cisco IOS routers perform QoS in software Places additional demands on the CPU Cisco switches perform QoS in dedicated hardware ASIC's and therefore do not tax their CPU

4 traffic descriptors

Class of Service (CoS) at Layer 2 incoming interface / IP precedence / Differentiated Services Code Point (DSCP) at Layer 3 source or destination address application

3 characteristics of converged networks

Competition between constant, small-packet voice flows and bursty video/data flows Time-sensitive voice and video flows Critical traffic must get priority

characteristics of real-time interactive video traffic

Comprises different types of packets with different delay and tolerance for loss within the same session Interactive video (video conferencing) has the same delay, jitter, and packet loss requirements as voice traffic One of the most stringent QoS requirements General guideline for overhead is to provide 20% more bandwidth than the data currently requires not tolerant of delay, jitter, or packet loss

characteristics of passive streaming video traffic

Comprises different types of packets with different delay and tolerance for loss within the same session Requirements include a loss of < 5% and a delay of no more than 4-5 seconds not tolerant of delay, jitter, or packet loss

QoS policy

Definition of the QoS levels that are assigned across a network

QoS Policy: business audit

Determine how the application requirements for each business unit maps into the overall business model and goals Important to have executive sponsorship for this process as QoS inherently means that some traffic and users will receive priority over others

QoS Policy: service levels understanding

Determine how the application requirements for each business unit maps into the overall business model and goals Important to have executive sponsorship for this process as QoS inherently means that some traffic and users will receive priority over others

5 characteristics of QoS classification

Determine which treatment that traffic should receive according to behavior and business policies Partitions network traffic into multiple priority levels or Classes of Service Takes place at the access layer and network edge - ideally closest to the source The source agrees to adhere to the contracted terms and the network promises QoS QoS mechanisms use the classification of each packet

3 characteristics of policing / shaping

Drop misbehaving traffic to maintain network integrity or shape traffic to control bursts Policers drop / re-mark traffic Shapers delay traffic Used to enforce a rate limit that is based on metering, with excess traffic being dropped/marked/delayed

2 characteristics of congestion management

Each interface must have a queuing mechanism to prioritize the transmission of packets based on the packet marking Normally implemented on all output interfaces

QoS Guidelines: (1) Classify and mark applications as close to their sources as possible

Enables end-to-end (ETE) DiffServ PHB's Not recommended to trust endpoints to set CoS or DSCP markings correctly because users can easily abuse provisioned QoS policies if they are permitted to make their own traffic For this reason, "as close as possible" is included in the design principle Ex - if DSCP EF received priority services throughout the enterprise, users could easily configure the NIC on a PC to mark all traffic to DSCP EF, thus hijacking network priority queues to service their non real-time traffic

PHB Standard: Assured Forwarding (AF)

Guarantees BW and allows access to extra BW when available Defines a method by which BA's can be given different forwarding assurances Each level has allocated BW based on the QoS policy Must detect and respond to long-term congestion within each class by dropping packets, while handling short-term congestion (packet bursts) by queuing packets Implies the presence of smoothing or filtering function that monitors the instantaneous congestion level and computes a smoothed congestion level Instead of using Strict PQ, more balanced queue servicing algorithms are implemented (fair or weighted fair queueing) If congestion occurs within a class, the packets with the higher drop probability are discarded first Sophisticated drop selection algorithms like RED are used to avoid tail drop issues Bits 7 through 5 of DSCP = 001, 010, 011, or 100 DSCP values of aaadd0, where aaa is the class and dd is the drop probability Has 4 standard classes (af1, af2, af3, af4) that are represented by the aaa values of 001, 010, 011, 100 Each class should be treated independently and should have allocated bandwidth that is based on the QoS policy

Link Efficiency Mechanisms: compression methods

Header compression and payload compression mechanisms reduce the sizes of packets, reducing delay and increasing available bandwidth on a link based on eliminating redundancy Protocol header is an item of repeated data that does not change much over the lifetime of that flow Using header compression mechanisms, most header info can be sent only at the beginning of the session, stored in a dictionary, and then referenced in later packets by a short dictionary index Layer 2 payload compression and header compression are performed on a link-by-link basis These compressions techniques cannot be performed across multiple routers because routers need full Layer 3 header info to be able to route packets to the next hop

IEEE Standard: Class of Service (CoS)

IEEE 802.1q

3 steps to determine QoS policy

Identify traffic across the environment and its requirements (network, business, service levels) Divide traffic into classes with similar QoS requirements Define QoS policies for each class that will meet QoS requirements

WRED: Probability Denominator

If the average queue length is greater than the minimum threshold, but less than the maximum, WRED will either queue the packet or perform a random drop If the average queue length is less than the minimum threshold, the packet is passed to the output queue If the queue is already full, the packet is tail-dropped Otherwise, the packet will eventually be transmitted out of the interface

Purpose of QoS

Implemented to help provide consistent, predictable performance (in addition to bandwidth) Achieved by providing tools to manage network congestion, shaping network traffic, using expensive wide-area links more efficiently, and setting traffic policies across the network Gives some sessions priority over others Network traffic from business-critical and delay-sensitive applications must be serviced with priority and protected from other types of traffic

IP Precedence (IPP)

In the original definition of the IP header's Type of Service (ToS) byte, the first 3 bits of the ToS byte, used for marking IP packets for the purpose of applying QoS actions.

5 characteristics of policers

Instantaneous decisions Ideally placed as ingress tools (drop before CPU cycles) Can be placed at egress to control the amount of traffic per class When configured traffic limit is exceeded, policer can either drop traffic or re-mark to another class of service Class-based policing allows excess traffic to be resent with a lower priority Significant # of TCP resends can occur Does not introduce jitter or delay

4 quality issues

Lack of bandwidth End-to-end delay (fixed and variable) Jitter (variation of delay) Packet loss

Layer 3 marking

Layer 3 packet marking with IP precedence and DSCP is the most widely deployed marking option because Layer 3 packet markings have end-to-end significance. Layer 3 markings can also be easily translated to and from Layer 2 markings

Congestion Management: Queuing / Buffering

Logic of ordering packets in output buffers Only activated when congestion occurs Queuing algorithms mange the front of the queue, and congestion mechanisms manage the back of the queue When queues fill up, packets can be reordered so that the higher-priority packets can be sent out of the exit interface first Devices can either wait for queues to fill up and then start dropping packets, or drop packets before the queues fill up Primary way to manage congestion in a network

3 congestion management techniques

Low Latency Queuing (LLQ) Class-Based Weighted Fair Queuing (CBWFQ) Weighted Random Early Detection (WRED)

Typical definition of a low-bandwidth link

Low bandwidth links have speeds less than or equal to 768kbps

3 parameters to define traffic class

Min/max bandwidth Priority (low 0 - 7 high) Values are carried in Layer 2 802.1Q frame headers as a Class of Service (CoS) value, or Layer 3 IP packets as IP precedence or as part of Differentiated Services Code Point (DSCP) value Congestion management technique (LLQ, CBWFQ, WRED)

Mitigate Jitter (variation of delay)

Mitigate this with a dejitter buffer Buffers packets and plays them out in a steady stream Adds total delay, but allows for smooth delivery of real-time traffic If the jitter limit is exceeded, the packet is dropped

Mitigate packet loss

Mitigate this with upgrading link bandwidth, using queuing techniques, or preventing congestion by shaping/dropping packets before congestion occurs

QoS Guidelines: (4) Enable queuing policies at every node where the potential for congestion exists

Most campus links are underutilized Means that campus networks can be designed to accommodate oversubscription between access, distribution, and core layers Oversubscribing allows for uplinks to be utilized more efficiently Common campus oversubscription values are 20:1 for the access-to-distribution layers and 4:1 for the distribution-to-core layers Only way to provide service guarantees Potential for congestion exists in campus uplinks because of oversubscription ratios and speed mismatches in campus downlinks Ex - Gigabit or Fast Ethernet links

QoS Guidelines: (2) Police unwanted traffic flows as close to their sources as possible

No benefit in forwarding unwanted traffic only to police and drop it at a subsequent node Especially when the unwanted traffic is the result of attacks Can overwhelm processors with traffic

DiffServ Model: Per Hop Behavior (PHB)

PHB is defined as the probability of timely forwarding observable forwarding behavior that is applied at a DiffServ-compliant node to a DiffServ BA Different PHBs are used in a network based on the DSCP of the IP packets Refers to the packet scheduling, queuing, policing, or shaping behavior of a node on any given packet belonging to a BA DiffServ model does not specify how PhBs must be implemented Cisco IOS Modular QoS CLI (MQC) policy maps can be used to configure PHBs

Queuing: Tail Drop

Packet drops that occur when a queue fills, another message arrives that needs to be placed into the queue, and the networking device tries to add the new message to the tail of the queue but finds no room in the queue, resulting in a dropped packet. Avoids global synchronization of TCP streams Selective dropping of packets while queues are filling up is called congestion avoidance

QoS Levels

Platinum (VoIP) Gold (video) Silver (default best-effort for non-essential) Bronze (guest services)

Congestion Management: Scheduling

Process of deciding which packet should be sent out next Occurs regardless of congestion Methodical output of packets at a desired frequency Can be applied to different traffic classes to weight the traffic by priority

Link Efficiency Mechanisms

QoS mechanisms exist for optimizing throughput and reducing delay, but do not create bandwidth Decrease latency and assist in meeting the service-level requirements of delay-sensitive traffic

5 guidelines for enterprise campus QoS

Queuing, classification, marking, and policing are important QoS functions that are optimally performed within the campus network at the access layer ingress edge (1) Classify and mark applications as close to their sources as possible (2) Police unwanted traffic flows as close to their sources as possible (3) Perform QoS in hardware rather than software when possible (4) Enable queuing policies at every node where the potential for congestion exists (5) Protect the control and data planes

characteristics of voice traffic

Real-time Comprises constant and predictable bandwidth and packet arrival times Most stringent QoS requirements not tolerant of delay, jitter, or packet loss

Congestion Management: round-robin scheduling

Round-robin: packets in queues are served in a set sequence No starvation with this scheduler, but delays can badly affect the real-time traffic

3 characteristics of congestion avoidance

Specific packets are dropped early based on marking Typically implemented on output interfaces whenever a high-speed link or set of links feeds into a lower-speed link Queuing / Buffering Scheduling Link Efficiency Mechanisms

Link Efficiency Mechanisms: 3 types of Layer 2 payload compression methods

Stacker Predictor Microsoft Point-to-Point Compression (MPPC)

6 requirements for successful QoS deployment

Strategically defining QoS objectives Analyzing application service-level requirements Designing and testing QoS policies Implementing QoS policies Monitoring service levels Repeat as business conditions evolve

Call Admission Control (CAC)

Strategies used to prevent real-time traffic from exceeding the capacity of the network Should be used instead of policing and shaping on real-time traffic (counter-productive)

Congestion Management: 3 scheduling mechanism examples

Strict priority: queues with lower priority are only served when the higher-priority queues are empty Round-robin: packets in queues are served in a set sequence Weighted fair: queues are weighted, so some are served more frequently than others

Link Efficiency Mechanisms: 4 types of Cisco IOS header compression methods

TCP header compression Real-Time Transport Protocol (RTP) header compression class-based TCP header compression class-based RTP header compression

WRED: algorithm

The router constantly updates the WRED algorithm with the calculated average queue length, which is based on the recent history of queue lengths When a packet arrives at the output queue, the QoS marking value is used to select the correct WRED profile for the packet The packet is then passed to WRED for processing Based on the selected traffic profile and the average queue length, WRED calculates the probability for dropping the current packet ("Probability Denominator")

characteristics of data traffic

Typically not real-time May be bursty in that they create unpredictable traffic patterns Causes widely varying packet arrival times May use as much bandwidth as possible Least stringent QoS requirements Many applications use TCP since data traffic can not normally be dropped

WRED: usage with TCP

WRED addresses packet loss caused by tail drop Tail drop causes global TCP sync problem where traffic reduction during congestion leads to future traffic increase, which again congests WRED configures different tail drop thresholds for each IP precedence or DSCP value so lower-priority traffic is more likely to be dropped, avoiding global TCP snc

Congestion Management: weighted fair scheduling

Weighted fair: queues are weighted, so some are served more frequently than others Solves starvation and gives priority to real-time traffic Method does not provide bandwidth guarantees (drawback) Resulting bandwidth per flow varies based on the # of flows present and the weight of each of the other flows

IPP and DSCP

While IP precedence is the old way to mark ToS, DSCP is the new way The transition from IP precedence to DSCP was made because IP precedence only offers 3 bits, or eight different values, to describe different classes of traffic DSCP is backward-compatible with IP precedence

QoS Guidelines: (5) Protect the control and data planes

Without QoS, attack-generated traffic drown out applications and cause denial of service through unavailability Enabling QoS policies within the campus maintain network availability by protecting and servicing critical applications

Queuing: Weighted Fair Queuing

algorithm that divides the interface bandwidths by the # of flows, thus ensuring proper distribution of the bandwidth for all applications Provides good service for real-time traffic, but there are no guarantees for a particular flow

Link Efficiency Mechanisms: Link Fragmentation and Interleaving (LFI)

allow traffic types (voice & interactive) to be sent either ahead of interleaved with larger, more aggressive flows Layer 2 technique in which large frames are broken into smaller, equally sized fragments and then transmitted over the link in an interleaved fashion with more latency-sensitive traffic flows( like VoIP) Smaller frames are prioritized and a mixture of fragments is sent over the link Reduces the queuing delay of small frames because the frames are sent almost immediately Therefore, link fragmentation reduces delay and jitter by expediting the transfer of smaller frames through the hardware transmit queue

CoS: priority value 0

assigned to Cisco IP phone traffic received from a host on its access port required to prioritize transmission of voice packets over data packets since packets from both share a physical link to the switch

CoS: priority value 5

assigned to IP phone voice data traffic

CoS: priority value 3

assigned to voice signaling traffic

5 QoS mechanisms

classification marking congestion management congestion avoidance policing / shaping

DiffServ Model: Behavior Aggregate (BA)

collection of packets with the same DSCP value crossing a link in a particular direction Packets from multiple applications and sources can belong to the same BA

Queuing: Class Based Weighted Fair Queuing (CBWFQ)

combination of bandwidth guarantee with dynamic fairness of other flows Does not provide latency guarantee and is only suitable for data traffic management Define traffic classes based on match criteria (protocols, ACLs, input interfaces) A queue is reserved for each class Minimum bandwidth during congestion is assigned per class After a queue has reached its configured queue limit, enqueuing of additional packets to the class causes tail drop or random packet drop to take effect Packet weight is derived from configured class bandwidth Therefore, bandwidth assigned to the packets of a class determines the order in which packets are sent No class may be granted strict priority Poses a problem for voice traffic

Queuing: Random Early Detection (RED)

congestion avoidance technique monitors the buffer depth and performs early discards (drops) on random packets when the minimum defines queue threshold is exceeded addresses tail drops does not address queuing, BW starvation, or BW guarantees

3 traffic types

data voice video

Type of Service (ToS)

generally used to indicate the Layer 3 IPv4 packet field and comprises 8 bits, 3 of which are designated as the IP precedence field. IPv6 changes the terminology for the same field in the packet header to "Traffic Class"

Link Efficiency Mechanisms: payload compression

increases the amount of data that can be sent through a transmission resource Primarily performed on Layer 2 frames and therefore compresses the entire Layer 3 packets

QoS Model: IntServ (Integrated Services)

introduces to supplement the best-effort delivery by setting aside some bandwidth for applications that require bandwidth and delay guarantees Provides guaranteed QoS to IP packets and delivery Can limit scalability of the network Expects applications to signal their QoS requirements to the network

Quality of Experience (QoE)

measures end-user perception of the network performance (subjective metric)

4 traffic classes (highest --> lowest priority)

mission-critical transactional/interactive best-effort scavenger

3 types of Link Efficiency Mechanisms

payload compression header compression Link Fragmentation and Interleaving (LFI)

QoS Model: Best Effort

predominately-used model designed for no-guarantee delivery of packets QoS is not applied to traffic and packets are serviced in the order received

TCP Flow Control

receiver controls sender, so sender won't overflow receiver's buffer by transmitting too much, too fast operates by increasing the transmission rates of traffic flows until packet loss occurs When packet loss occurs, TCP drastically slows down the transmission rate and then again begins to increase the transmission rate

Queuing: Weighted Random Early Detection (WRED)

same as RED, except that traffic weights skew the randomness Idea behind WRED is to maintain the queue length somewhere between the minimum and maximum thresholds and to implement different drop policies for different classes of traffic Selectively discards lower-priority traffic when the interface becomes congested Useful for networks with traffic using mainly TCP due to re-transmission and congestion avoidance addresses tail drops does not address queuing, BW starvation, or BW guarantees

Queuing: Custom Queuing

set of 16 queues with a round-robin scheduler To prevent traffic starvation, it provides traffic guarantees Does not provide strict priority for real-time traffic (drawback)

Queuing: Priority Queuing (PQ)

set of four queues that are served in strict-priority order Can starve traffic in the lower-priority queues since they are served only when the higher-priority queues are empty

Queuing: First In First Out (FIFO)

single queue with packets that are sent in the exact order that they arrived

Queuing: Low Latency Queuing (LLQ)

suitable for mixes of data and real-time traffic Provides latency and bandwidth guarantees essentially CBWFQ with strict priority Allows voice traffic to be dequeued and sent first

Traffic Identifier (TID)

term used to describe a 4-bit field in the QoS control field of wireless frames (802.11 MAC frame) Used for wireless connections and CoS is used for wired Ethernet connections

Class Selector (CS)

term used to indicate a 3 bit subset of DSCP values Class selector designates the same 3 bits of the field as IP precedence Provides interoperability between DSCP-based and IP precedence-based devices in a network Bits 4 through 2 of DSCP = 000 (xxx000 DSCP) identifies a Class-selector DSCP value for a CS PHB can be calculated by multiplying the class # by 8 Ex - the DSCP value for CS3 would be = (3*8) = 24 used for backward compatibility with non-DiffServ-compliant devices (RFC 1812 devices)

Classifier

tool that inspects packets within a flow to identify the type of traffic that the packet is carrying

4 characteristics of shapers

traffic-smoothing tools that work in cooperation with buffering mechanisms deployed between enterprise network and service provider to ensure enterprise traffic is under contracted rate fewer TCP resends than with policers introduces delay and jitter

PHB Standard: Default (tail drop)

used for best-effort service Bits 7 through 5 of DSCP = 000 Essentially specified a packet marked with a DSCP = 000000 (recommended)

PHB Standard: Expedited Forwarding (EF)

used for low-delay service Ensures minimum departure rate Guarantees amount of BW with prioritized forwarding Polices excess BW so that other classes not using this PHB are not starved Strict Priority Queuing (PQ) is typically used for EF traffic Applications like VoIP, video, and online trading programs require this kind of service Bits 7 through 5 of DSCP = 101 Packets requiring EF PHB should be marked with a DSCP binary value of 101110, or decimal 46 Non-DiffServ-compliant devices will regard the EF DSCP value as IP precedence 5

DiffServ Model: Differentiated Services Code Point (DSCP)

value in the IP header that is used to select a QoS treatment for a packet Classification and QoS revolve around the DSCP in the DiffServ model set of 6 bit values that can describe the meaning of the Layer 3 IPv4 ToS field DSCP is encoded in the header of both IPv4 and IPv6 packets