2.traffic and QoS Management
2.traffic and QoS Management
email: [email protected]
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Traffic and QoS management
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Introduction
Telecommunication networks carry different types of traffic, such
as audio (e.g., voice), video (e.g., TV, streaming), and various data
(e.g., Web-based services, etc.).
The ITU has defined traffic and QoS management framework for
the telecommunication networks.
But, all networks today are based on Internet technologies (all-IP)
That requires merging of Internet and telecom traffic and QoS
management approaches.
There are two main types of networks and services:
Internet, which is a global public best-effort network, available to
divergent applications/services on network neutrality basis
Managed/specialized networks and services (again, based on IP and
Internet technologies, but not part of the Internet network).
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Internet and QoS
What to do with different traffic types in Internet public or/and non-public
networks?
On one side, Internet is initially designed to provide best-effort service,
i.e., all IP packets are treated in the same manner.
But, not all IP packets are the same
Web as interactive application is delay sensitive,
Voice is more sensitive to delay and jitter,
Video requirements are highest regarding the required bitrate,
Online games are delay and jitter sensitive
Peer-to-peer file sharing applications (e.g., BitTorrent) are generally insensitive
to delay, jitter, while bandwidth is also important...
and so on…
This "big picture" refers to public Internet, which is also referred to
Internet Access Service (IAS) by many regulators and telecom operators.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Internet protocol vs. QoS
Internet Protocols, both
versions IPv4 and
IPv6, have built-in
fields for traffic and
QoS management:
Type of Service
(ToS) field in IPv4
Traffic Class field in
IPv6 header
However, they are
rarely used in practice
due to autonomous
nature of IP networks.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
UDP and QoS
User Datagram Protocol (UDP) does nothing except adding sources and
destination port numbers, no direct function regarding the QoS.
However, with its simplicity the UDP provides lowest end-to-end delay
(because there are no retransmissions of the lost packets).
With UDP/IP typically the control is implemented into the application over
them (e.g., DNS) or another protocol used over UDP such as RTP (Real-
time Transport Protocol)
Voice over IP uses RTP/UDP/IP for transfer of voice.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
TCP congestion control and QoS
TCP (Transmission Control Protocol) can be considered as being
one of the most important Internet technology protocols.
Why? Because the initial concept of Internet networking for end-to-
end QoS (congestion control) is implemented with TCP at the end
hosts.
TCP congestion control
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
DNS and DHCP vs. QoS
DNS (Domain Name System) connects the two name spaces in Internet:
IP addresses and domain names.
proper functioning of the DNS and lowest possible DNS resolving time is one of
the key performance parameters for QoS in all IP-based networks.
Without DNS all IP-based services will not work (e.g., no Internet access).
DHCP (Dynamic Host Configuration Protocol) does dynamical provision
of network configuration parameters to network interfaces in a given IP
network.
DHCP exists in every network, including telecom operators networks,
enterprise, home networks.
From the QoS point of view, proper functioning of the DHCP server in the IP
network is crucial with aim hosts to obtain an IP addresses on their network
interfaces in a plug-and-play manner and to have Internet connectivity.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Autonomous Systems (AS)
Internet is built on the basis of so-called Autonomous Systems
(ASs), where each autonomous system is identified by a 16-bit or
32-bit AS number.
AS number is allocated by the IANA (Internet Assigned Numbers
Authority), part of ICANN (Internet Corporation for Assigned Names and
Numbers) which currently governs the Internet.
Autonomous System (AS) is a collection of routers under the
control of one authority (e.g., a company, telecom operator, etc.).
There are two main types of exchange of traffic between different
ASs at inter-AS interconnections:
Transit: ISP charges its customers (other ISPs or companies which
have their own ASs) for forwarding/routing their IP packets.
Peering: Similar to the transit, the peering is also a business deal,
however it may be setup without particular financial settlement.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Impact of Autonomous Systems on QoS
The autonomous system is called “autonomous” because it can apply
traffic management schemes independently from other ASs.
Regarding the end-to-end QoS, more ASs on the path of the IP packets
(i.e. the traffic) add more delay and a higher probability of appearance of
bottlenecks.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Internet traffic characterization
Voice traffic
conversational type of traffic (with similar requirements in both directions, from
calling party A to called party B, and vice versa),
has constant bit rate when sending or receiving. Also, voice requires relatively
small IP packets (e.g., 50 – 200 bytes).
Video traffic
It has generally high variable bit rate, which is dependent upon the codec
efficiency on the moving pictures (e.g., MPEG-2, MPEG-4 various codecs, etc.).
Video streaming is unidirectional traffic (in downstream i.e., downlink)
the maximum IP packet size for video traffic is limited by the MTU (Message
Transfer Unit) of Ethernet (and similarly in WiFi) which is 1500 Bytes.
Data traffic
typically the non-real-time traffic in Internet
the most prominent examples are WWW (i.e., Web), email, and all non-real-time
OTT (Over The Top) applications/services.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Telecom operator traffic characterization
Telecom voice traffic:
Voice over IP (VoIP) as PSTN/ISDN replacement, with guaranteed QoS end-to-end
based on standardized signaling within the NGN (Next Generation Networks)
standards umbrella.
Operator’s VoIP is not part of the Internet, although it is transferred by using the
same access, core, and transit IP networks as Internet and other IP-based services.
Telecom video traffic: TV/IPTV and VoD (Video on Demand)
When provided by telecom operators it has guaranteed QoS.
It is not part of Internet (similar to carrier-grade VoIP).
Telecom data traffic: all Internet traffic either fixed or mobile
it refers to all OTT (Over The Top) services including OTT voice, OTT video and
other Internet services (e.g., Web, email, social networking, clouds, etc.) provided
on the basis of network neutrality and best-effort approach
no QoS guarantees per service or flow, although Internet Access Service (IAS)
as a whole has certain QoS guarantees (e.g., on minimum bitrate).
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Audio traffic characterization
The audio traffic usually refers to voice.
The conversational voice (telephony) is most sensitive to delay
and jitter.
the recommended delay in PSTN is below 150 ms, while above 400 ms
is not acceptable
Audio is more tolerant to errors than the video due to the human
ear
Voice is a two-way continuous audio streaming
it is typically carried by using RTP/UDP/IP protocol stack
Regarding the QoS, there are differences between
Voice by telecom operators - has guaranteed QoS
Voice over Internet (e.g., Skype, Viber, WhatsApp, etc.) – no
guaranteed QoS
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Video traffic characterization
Most of the video traffic is based on video information which is
statistically compressed.
Most used video codecs nowadays are based on the MPEG-4 standard
type of pictures in MPEG are ‘I’, ‘P’, and ‘B’ frames .
Video traffic has high peak to mean ratio.
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Video playback point
Playback point is introduced with certain delay in respect to the
average delay of the packets
with aim to compensate for the delay variation of most of the packets
Typically live video traffic is based on RTP/UDP/IP protocol stack.
With the availability of broadband access to Internet, many video
sharing sites provide video content by using HTTP/TCP/IP stack
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Non-real-time traffic characterization
Non-real-time traffic has no strict requirements on delay
Most of this non-real-time traffic is carried by using TCP/IP stack
it has lossless transmission end-to-end, provided by TCP on the transport
layer at the end hosts
The non-real-time is considered also as one of the main reasons for
Internet congestion at bottlenecks
TCP
behavior at
congestion
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
QoS on different protocols layers
There are inter-relationship of the QoS parameters on different
layers.
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Traffic management
In an all-IP telecommunications/ICT world, Internet technologies
should be considered for traffic management
and for QoS standardization, monitoring and enforcement.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
How do we classify IP packets?
Classification based on ports
ports are used by transport layer protocols such as TCP and UDP to identity
applications on the top.
Classification based on application type
different applications may have different treatment. For example, carrier grade
VoIP takes precedence over HTTP.
Classification based on user type
For example, home and business users may get normal service, but hospitals,
police or fire departments get highest priority services.
Classification based on subscription
different users may pay different prices for the same services with different QoS.
For example, 50 USD for high speed Internet access and 10 USD for fair-usage
policy on Internet access service.
etc.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
IPv4 and IPv6 flow identification
Packet classification is needed to sort packets into flows (per class).
In IPv4 classification is performed with the following fields in IPv4 headers
(so-called 5-tuple):
Source IP address (from IP header);
Destination IP address (from IP header);
Protocol number (from IP header);
Source port (from transport protocol header, e.g. TCP, UDP);
Destination port (from transport protocol header, e.g. TCP, UDP).
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Techniques to improve QoS
To improve the QoS in Internet (and in packet-switched networks in
general), there are four common methods:
Scheduling
Resource Reservation
Traffic shaping
Admission Control
These techniques are implemented in Internet network nodes:
Routers (in access, core and transport networks),
Switches (in access networks).
However, with to use any of the above QoS techniques one needs a
classifier:
When a router receives a packet, the classifier performs a classification and puts
the packet in a specific queue based on the classification result.
All packets from a same class get the same treatment from the packet scheduler.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Packet scheduling: First In First Out (FIFO)
Packet scheduler schedules the packets to meet their QoS
requirements:
Packet scheduler manages the forwarding of different streams using a
set of queues and timers.
It is implemented at the point where the packets are queued (waiting
area in Figure below) and then sent on the link (departures in the Figure):
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Packet scheduling:
Weighted Fair Queuing (WFQ)
WFQ (Weighted Fair Queuing) also assigns packets to different classes
and admit them to different queues.
In WFQ the higher priority means a higher weight.
For example, if the weights of the queues are 3, 2, and 1, then in one cycle 3
packets are processed from the first queue, 2 from the second queue, and 1 packet
from the third queue.
Each queue however is served in FIFO manner.
WFQ can be combined with PQ, thus resulting in fair queuing with
priority (it avoids starvation of lower priority traffic).
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Resource reservation end-to-end
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
25
Traffic shaping
Purpose: need for a mechanism to control packet flow to manage
high vs. medium vs. low priority flows.
Traffic shaping shapes the bitrate according to given fixed data rate,
which is accomplished with control mechanisms:
Leaky-bucket algorithm:
Shapes bursty traffic into fixed-rate traffic by averaging the data rate.
It may drop the packets if the bucket is full.
Token-bucket algorithm:
Used for policing of traffic: if token is available, packet may pass; otherwise,
packet is queued or dropped.
The token bucket allows bursty traffic at a regulated maximum rate.
Without Traffic Shaping
Line
Rate
With Traffic Shaping
Shaped
Rate
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Admission Control
No matter how good the scheduler works, one may still have no QoS in
practice
When traffic demand exceeds available resources.
What is Admission Control?
It is traffic management technique which is used to reject a flow if insufficient
resources (e.g., in the access, core or transit network).
It is different than traffic policing (which is used to reject/drop a packet).
In telephone network (PSTN/PLMN) is easy to do admission control – why?
Due to circuit-switching (allocation of a pair of channels per call, i.e., fixed
resources per call).
Teletraffic engineering provides efficient means for planning the telephone
networks (e.g., well-known Erlang-B formulae).
In Internet flows from different applications/services (e.g., voice, video, Web,
email, gaming, cloud, etc.) have different QoS requirements:
Therefore, admission control is much harder for implementation.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
What influences the traffic management?
The type of traffic, such as voice (it is low rate) or video (high
bitrates and bursty by nature) or data (variable data rates and bursty
by nature);
The number of users attached to a given network;
The capacity of the network which is installed (measured in bits
per second, or shortly bit/s or bps, which are equivalent between
each other);
The type of network equipment and its capabilities (e.g. switches,
routers, gateways, base stations, etc.);
Business plans (by the ISP, i.e. telecom operator) and regulatory
legislation (e.g. national regulation);
Capacity of employees at the ISPs and regulators.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Traffic management: Intervention
All traffic management involves a decision basis and an
intervention.
For example, exceeding a monthly usage allowance is a decision basis,
and the response of cutting the data rate according to policy is an
intervention.
Three common decision inputs are:
The user identity (or profile), specifying a QoS package for that user;
Whether or not a usage cap has been exceeded (these caps are often
set by the user’s tariff);
The particular traffic type (e.g. VoIP, IPTV, OTT, etc.).
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Traffic management intervention types
Characteristic
Impact of negative
Intervention ISO intervention on data type
type Possible Applied
Model Comments
actions in
Level TCP/IP UDP
FTP RTP
TCP/IP traffic
can be
Packet Prioritize or Core Layers 3 Retransmission effectively
Data loss
prioritization De-prioritize network and 4 of packets managed by de-
prioritizing this
traffic type
Video is best
Reduced
Reduced managed by
quality
throughput prioritizing or
Bandwidth Guarantee Access Layers 2 (Codec
(Service giving
allocation or Cap network and 3 may drop
maintained, but guaranteed
to a lower
at lower speed) bandwidth in the
rate)
access network
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Generic traffic management architecture
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Network agnostic traffic management
In telecom operators’ networks, Deep Packet Inspection (DPI)
boxes are deployed in the core network nodes to inspect the
packets to determine traffic types.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Traffic management vs. network capacity
The extent and complexity of traffic management is associated with
how much congestion is being experienced, or
how close network traffic is to the limit of the network’s capacity
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Traffic management positive and
negative effects on QoS
Traffic management
Can reduce congestion, Other users' traffic may take
applied to other users’
allowing fair use for all users priority over user's own traffic
traffic
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
ITU QoS framework
QoS framework by ITU-T
is organized into three
planes (Y.1291):
Control Plane: admission
control, QoS routing, and
resource reservation.
Data Plane: buffer
management, congestion
avoidance, packet
marking, queuing and
scheduling, traffic
classification, traffic
policing and traffic
shaping.
Management Plane: SLA,
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-
traffic restoration, metering Broadband”, Wiley, UK, 2019.
and recording, and policy.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Internet QoS framework
Best effort
This is traditional Internet model, without any QoS guarantees.
The IP networks just route packets until they reach the destination.
Integrated services (IntServ)
This is first standardized mechanism by IETF, which is based on resource
reservation in routers on the path by using signaling.
So routers have to store traffic and QoS information per connection (i.e., flow).
Differentiated services (DiffServ)
This is the most commonly used method for traffic differentiation in Internet, in
which all packets are classified into limited number of classes,
so routers have to store only information per class (not per connection, i.e., per
flow).
Multi-Protocol Label Switching (MPLS)
This is “default” approach for QoS provisioning in transport IP networks, which
may be combined with DiffServ as well as other protocols such as BGP (Border
Gateway Protocol), VPN (Virtual Private Networks) protocols.
Policy-based QoS
This is typically used for QoS provisioning between network providers (e.g., BGP
policies).
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Integrated Services (IntServ)
Integrated Services (IntServ) is an architecture for providing QoS
guarantees in IP networks for individual application sessions.
Every router between the two end hosts which need to establish a session
must reserve required resources (e.g., bandwidth) by using the QoS call
setup signaling
Signaling based on use of Resource Reservation Protocol (RSVP).
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Differentiated Services (DiffServ)
Which are DiffServ functions?
Classification
Traffic conditioning.
DiffServ is based on Type of Service (ToS) field in IPv4 header and Traffic
Class field in IPv6.
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Basic model for a single provider
The QoS model which is used inside the given network provider’s
domain (e.g., given telecom operator, either fixed or mobile) is
targeted to limited number of services offered to the end users
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Comparison of Best-Effort, DiffServ, and IntServ
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
QoS with MPLS in transport networks
Legacy IETF approach which is standardized for QoS traffic management
in transport networks is Multi-Protocol Label Switching (MPLS)
Short, fixed-length labels are associated to streams of data
in the ingress edge MPLS router LER (Label Edge Router)
Packet belonging to those streams are forwarded along the LSPs (Label
Switched Paths) based on their labels
no IP header examination in LSRs (Label Switching Router )
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
DPI (Deep Packet Inspection)
Deep Packet Inspection (DPI) is a type of packet filtering
targeted to all headers and all payloads from network layer (layer 3) up to
application layer (layer 7), i.e., all layers that are consisting the end-to-end
communication between the end hosts.
in some cases layer 2 can be also considered in this group of layers for the
inspection.
Then, what is the potential QoS role of DPI?
Well, it can be used to provide classification of individual sessions and their
application/traffic type, which may be needed for traffic management.
Source: Toni Janevski, “QoS for Fixed and Mobile Ultra-Broadband”, Wiley, UK, 2019.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Deep Packet Inspection (DPI)
Deep Header Inspection (DHI)
Deep Application
Shallow Packet Medium-depth Packet
Network application Identification (DAI)
Inspection (SPI) Inspection (MPI)
QoS support
Traffic shaping (L3 byte shaping) - X - -
Network analysis
User behavior - X X X
Usage patterns - X X X
Charging/billing support
Time-based information - X - -
Traffic volume-based information - X - X
Link-oriented DPI
Layer 2-related policy conditions X X X X
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Conclusions
Internet and managed IP networks carry heterogeneous traffic
voice, video, data
and also emerging critical data (towards future).
Traffic management is targeted to:
Higher efficiency of IP networks for different services
Carrier grade (VoIP, IPTV, VPN) and OTT (all Internet services)
Supporting business goals of providers
Increasing the quality of experience of users
ITU and IETF QoS frameworks are based on traffic management
techniques (buffering, priority, admission control, DPI…)
In all cases the goal is end-to-end QoS…
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski
Sources:
Toni Janevski, “QoS Technoloigies for Fixed and Mobile Ultra-Broadband”,
John Wiley & Sons (Wiley – IEEE series), UK, April 2019.
Toni Janevski, “NGN Architectures, Protocols and Services”, John Wiley &
Sons, UK, April 2014.
“QoS Technologies and Regulation for Fixed and Mobile”, Prof. Dr. Toni Janevski