100% found this document useful (1 vote)
203 views173 pages

Level Three Dante Certification Program Advanced Networking Concepts Troubleshooting Domain Manager Audinate Pres

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (1 vote)
203 views173 pages

Level Three Dante Certification Program Advanced Networking Concepts Troubleshooting Domain Manager Audinate Pres

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 173

ADVANCED

DANTE NETWORKING
DANTE CERTIFICATION PROGRAM
LEVEL 3

1
DANTE CERTIFICATION PROGRAM

Training program from Audinate


!
Official certification lets your customers
know that you have the knowledge and
skills to implement Dante networks
!
Ensures a consistent set of methods
and knowledge

Copyright 2017 Audinate Pty Ltd. All rights reserved


DANTE CERTIFICATION PROGRAM

With Dante Certification,


you receive:
• Use of the Level 1, Level 2, and
Level 3 “Dante Certified” logos
• A certificate of completion for each
level passed.

Copyright 2017 Audinate Pty Ltd. All rights reserved


UNDERSTANDING SCALABLE DANTE
NETWORKS
Dante Topics General knowledge
– Motivation networking
◦ Who are Audinate – Motivation
◦ Why use Dante ◦ Why use networked services
◦ Where is Dante Used ◦ What is a network
◦ What is Dante ◦ Where are networks used
◦ How to work with Dante ◦ How to understand networks
– Architectural principles – Architectural principles
– Assessing and optimizing Network – Understanding network
components component features
– Troubleshooting – Covered by networking courses
– Dante in a converged network and general information

Copyright 2017 Audinate Pty Ltd. All rights reserved


ABOUT AUDINATE

Headquartered in Network Develop Dante as


Sydney, Australia engineers first 100%
interoperable
solution
for all audio
manufacturers

5
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHAT WE MAKE

Dante technology
(all of it)
Hardware modules
Development tools
Software products:
• Dante Controller
• Dante Virtual Soundcard
• Dante Via
• Dante Domain Manger

6
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHY USE
DANTE?

7
WHY USE DANTE?

A network is a group of things that connect


!
Interoperability is key
!
More than 1400 Dante-enabled products in the
market
!
More than 350 OEM manufacturers
!
More than 1 million Dante-Enabled endpoints
have shipped

Copyright 2017 Audinate Pty Ltd. All rights reserved


WHY USE DANTE?

If it has the Dante logo it will connect to


any other device with a Dante logo
!
Dante is a commercially developed and
supported solution – improvements and
features are added continually
!
Most widely adopted audio networking
solution ever

Copyright 2017 Audinate Pty Ltd. All rights reserved


WHERE IS DANTE
USED?

10
LIVE SOUND
AMERICAN AIRLINES ARENA – MIAMI, FL
AMERICAN AIRLINES ARENA - MIAMI, FL

11
Copyright 2017 Audinate Pty Ltd. All rights reserved
COMMERCIAL INSTALLATION
CHESAPEAKE ARENA – OKLAHOMA CITY, OK

12
Copyright 2017 Audinate Pty Ltd. All rights reserved
BROADCAST

VICTORIA PARLIAMENT HOUSE – MELBOURNE, AUSTRALIA


13
Copyright 2017 Audinate Pty Ltd. All rights reserved
EDUCATION

GOVERNORS STATE
GOVERNORS UNIVERSITY
STATE UNIVERSITY ––UNIVERSITY
UNIVERSITY PARK,
PARK, IL IL 14

Copyright 2017 Audinate Pty Ltd. All rights reserved


HOUSE OF WORSHIP
WILLOW CREEK COMMUNITY CHURCH – SOUTH BARRINGTON, IL

15
Copyright 2017 Audinate Pty Ltd. All rights reserved
FEATURE FILM PRODUCTION
LA LA LAND
16
Copyright 2017 Audinate Pty Ltd. All rights reserved
POST PRODUCTION

GOLDCREST FILMS – LONDON, UK


17
Copyright 2017 Audinate Pty Ltd. All rights reserved
RECORDING
SYNCHRON STAGE
VSL SYNCHRON – VIENNA,
STAGE AUSTRIA
– VIENNA, AUSTRIA 18
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHAT IS
DANTE?

19
WHAT IS DANTE?
• Dante is a networking technology
- Hardware solutions provided to OEM manufacturers
- Software solutions – DVS, Dante VIA, Dante Controller, DDM
- Network API’s
• Dante technology provides:
- Tightly synchronized (better than 1µs) media playout at every
endpoint in an IP network
- Uncompressed Digital Audio at all professional sample rates
- Simple plug and play discovery and routing across both a
local area network and a routed IP network

• Dante can be deployed on COTS (commercial off


the shelf) network infrastructure

Copyright 2017 Audinate Pty Ltd. All rights reserved


ADVANCED DANTE
NETWORKING: SECTION 1

21
IN THIS SECTION…

Networking
• What is a Network?

• Comparing ways of connecting Audio


devices:
- TDM networking techniques (time centric)

- Packet Switched (address centric)


networking techniques

22
WIRING UP AN “OLD” DIGITAL AUDIO SYSTEM
Meterbridge

Copyright 2017 Audinate Pty Ltd. All rights reserved 23


WITH DANTE

24
Copyright 2017 Audinate Pty Ltd. All rights reserved
CABLING REQUIREMENTS

“OLD” DIGITAL SYSTEM “WITH DANTE”


1x 9-pin serial (HA control) 4x Cat5e cables
1x Proprietary meterbridge cable
3x 75ohm BNC cables (wordclock)
1x 25pin D-SUB M-M (AES)
1x USB cable
1x Thunderbolt cable
4x TOS-LINK cables

25
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHAT DO WE MEAN BY A NETWORK?
• A Dante network is a collection of devices that are connected via IP and can
exchange audio information with each other.

• Dante does much more than this:


- In order to make a modern network useable, devices are automatically discovered
- Devices use human readable names – both for the device and its channel labels
- Diagnostic information is made available from the devices to controller applications
- Devices and channels can be managed by a systems administrator and a secure ecosystem can be
built
• Dante devices exchange information over an IP – Internet Protocol Network
- In all other areas of life the term “network” has become synonymous with IP network
- The biggest single network in the world is the Public Internet – unsurprisingly this too uses IP

26
Copyright 2017 Audinate Pty Ltd. All rights reserved
SO…DOES “NETWORKING” MEAN IP?
• IP networks are the most ubiquitous kinds of networks in the world today
• Why has IP become the dominant networking technology?
- Supported by the most manufacturers (both in terms of software and hardware)
- Has been applied to all industries (from Banking to Space Rockets to huge science
projects)
- IP is hugely flexible:
◦ The network is “neutral”
◦ IP networks reach true global scale
◦ Equipment and cabling takes advantage of “economies of scale” in an unprecedented way
◦ The model IP follows is vastly more scalable than alternatives
◦ Understood by millions of engineers and technicians world wide
◦ Used by billions of end users every hour of every day

27
Copyright 2017 Audinate Pty Ltd. All rights reserved
ALTERNATIVES TO IP (HISTORY)
• A dedicated cable for every task was the starting point
• The cable was part and parcel of the communications system
- “faster” more expensive cable did “high bandwidth” tasks
- “cheaper” more common cable did “lower bandwidth” tasks
• Logical evolution from analog
- Wiring became very complex
- Limitations of distance
- High deployment time
- Desire to reduce cabling

28
Copyright 2017 Audinate Pty Ltd. All rights reserved
ALTERNATIVES TO IP (HISTORY)

• Before IP networks became popular in telecommunications, TDM – Time Division


Multiplex networks were common
- TDM is the method that older digital audio formats used (AES3, AES10 (MADI) AES50)
- A TDM network runs at a set rate – it is considered a “synchronous” network
- “Timeslots” are filled with data (or not) and as long as the transmitting end puts the “right” data
in the “correct” timeslot (as far as the receiver is concerned) then the outgoing resultant data
matches that which went in
- The capacity of the network depends upon the clock frequency, and by extrapolation the
number of “timeslots” that are implied by this

29
Copyright 2017 Audinate Pty Ltd. All rights reserved
ALTERNATIVES TO IP (HISTORY)

• TDM networks are not “neutral”


– In telecommunications, as long as traffic was only voice, this was simple

– Once Telcos started to provide other services, the TDM model showed its
inefficiencies

30
Copyright 2017 Audinate Pty Ltd. All rights reserved
COMPARE THIS WITH TDM

TDM Systems are only efficient at 100% useful capacity

Number of “Boring” bits “Boring” bits Bandwidth % Bandwidth %


Channels Dante unicast 64 channel on 100mbps on 100mbps
24bit MADI link MADI link Dante
1 104 1512 100% 6%
2 80 1488 100% 6%
4 32 1440 100% 6%
64 128 0 100% 88%
65 104 1512* 200% 94%

31
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHY ARE “PACKET SWITCHED” NETWORKS USED
ON COMPUTERS?

• Packet switched networks are


neutral – we can run all manner of
applications

• Each of these applications has an


address:
• Some need to talk to each other
inside the computer processor,
so they have addresses to
communicate internally.
• Others need to talk externally
over the network socket.

32
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES

User told me to go to
www.youtube.com

I don’t know Youtube’s


IP address!

If only there was


something like a
Telephone Directory!

33
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES

Ah! DNS! I know a DNS


server’s IP address…
8.8.8.8! Lets call them

8.8.8.8
Google
Public
DNS
Internet

34
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES

Hello –I’m Iooking


for an IP address for
www.youtube.com.
Do you know it?
8.8.8.8
Google
Public
DNS
Internet

35
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES

Hello –I’m Iooking Yes – try


for an IP address for 172.217.23.14
www.youtube.com.
Do you know it?
8.8.8.8
Google
Public
DNS
Internet

36
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES

Better write that


one down so I
don’t have to ask
again

Public
Internet

37
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES
Hi 172.217.23.14:443 Hi
its 192.168.1.23 can 192.168.1.23:536
you call me back at 18 what can I do
TCP port 53618 for you?
please?

Public
Internet

38
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES
OK
Cat videos
meow please

Public
Internet

39
Copyright 2017 Audinate Pty Ltd. All rights reserved
APPLICATION ADDRESSES

Application Local Port Remote IP Remote Port


• The same process repeats PTP UDP 224.0.1.129 UDP 319
for every application Audio Flow UDP 14340 192.168.1.56 UDP 14390

• Each application has its own Audio Flow UDP 14350 192.168.1.60 UDP 14367

unique Internal (port) address Gain control UDP 50135 192.168.1.56 UDP 50231

Application Local Port Remote IP Remote Port


• In an Audio device this could
be: Youtube TCP 53618 172.217.23.14 TCP 443

Facebook TCP 53653 31.13.92.36 TCP 443

Outlook TCP 67123 105.40.225.204 TCP 389

Spotify TCP 57453 194.132.198.198 TCP 443

40
Copyright 2017 Audinate Pty Ltd. All rights reserved
HOW DO WE PUT IT TOGETHER…
BUT KEEP IT SEPARATE?

Dedicated “Pairs”
VS

TDM
IP Network Address Book

41
Copyright 2017 Audinate Pty Ltd. All rights reserved
ADVANCED DANTE
NETWORKING: SECTION 2

42
IN THIS SECTION…

Layered model of a network


• Encapsulation
- How we organize data to be sent
- How we create address centric circuits all the
way to the application
- Understanding Network Layers – the postal
model

43
LAYERED MODELS

OSI Model
7 – Application
• Older – conceptual model
• Good at defining the lower
6 – Presentation layers (1. Physical, 2. Datalink, Application
3. Network)
5 – Session • Less relevant in the upper layers

4 – Transport Transport
TCP/IP Model
3 – Network • Newer
Internet Layer
• Good at defining the upper
layers (Application, Transport,
2 – Datalink Internet)
• Lacks detail in the Network Network Access
1 – Physical Access layer

Copyright 2017 Audinate Pty Ltd. All rights reserved 44


ENCAPSULATION – THE “NETWORK STACK”

Application sends data Some Data

Transport layer puts Data in Packet – knows


which “port” connects to application and
Dst Src
“port” at far end, and which transport prot Some Data
Port Port
protocol (defined by application)

Network Layer adds Source and Destination Dst Src Dst Src
IP addresses IP IP
prot
Port Port Some Data

ARP table is used to put correct Destination Dst Src Dst Src Dst Src
MAC onto frame at Datalink layer MAC MAC IP IP
prot
Port Port Some Data

Placed on wire as bits at Physical Layer 010110011101000110111011100111…

45
Copyright 2017 Audinate Pty Ltd. All rights reserved
ENCAPSULATION – AN ANALOGY
IP stacks are abstract
1. I create some data (write a letter)
2. I put the letter in the envelope
(packetization)
3. I address the envelope
4. At the post office the envelope goes in
a mailbag (Frame (MAC)
encapsulation)
5. Mailbag is put in truck and taken to
next sort facility (physical layer)

46
Copyright 2017 Audinate Pty Ltd. All rights reserved
ADVANCED DANTE
NETWORKING: SECTION 3

47
IN THIS SECTION…
Different modes of communication
• Broadcast
• Unicast
Where they become applicable
• Collision Domains
• Broadcast Domains
Routers and Switches
• Segmenting the collision domain - switching
• Segmenting the Broadcast domain
• ARP
• VLANs
• IP Subnets
• Routers

48
TWO (OR 3) KINDS OF MESSAGES

• “Addressed” messages give huge flexibility advantages


• In a TDM network there is no requirement to use any form of address
- Destinations are interpreted by their timeslot position nothing more

• In a Packet Switched network (just like the mail system) the


addressee is unique
- It is also possible to send “unsolicited” mail – from a sender to many or all
addresses
- This is both very useful, and potentially irritating/bad at the same time

49
Copyright 2017 Audinate Pty Ltd. All rights reserved
TWO (OR 3) KINDS OF MESSAGES

• IP networks are very good at dealing with limiting


“unsolicited mail” through management techniques
• The three types of messages in an IP network are:
• Unicast – one to one communication
• Multicast – one to many communication
• Broadcast – one to all communication

• By contrast a TDM network is only really unicast (but


through “wiring tricks” can be analogous to Broadcast)
– either way there is no strict distinction

50
Copyright 2017 Audinate Pty Ltd. All rights reserved
HOW DIFFERENT MESSAGE TYPES WORK
• Old IP networks used “hubs” or a single cable to connect
computers together
- “layer 1” based network
- all messages are effectively “broadcast” messages within the network
- we can very precisely address a message
- Using a hub or a single “token ring” connection, we effectively remove the
“sorting office” from our postal system
- All computers receive every message, and have to discard anything they
are not interested in themselves

• Packet switched networks use a specific IP address for IP


broadcast messages,
- This maps to a specific broadcast destination MAC address

• In order for this to be useful we need something in the network


to “sort” the mail
51
Copyright 2017 Audinate Pty Ltd. All rights reserved
COLLISION DOMAINS

Message for
192.168.1.7
boring boring

192.168.1.4 192.168.1.5 192.168.1.6 192.168.1.7

52
Copyright 2017 Audinate Pty Ltd. All rights reserved
COLLISION DOMAINS

Message for Message for


192.168.1.7 192.168.1.5

192.168.1.4 192.168.1.5 192.168.1.6 192.168.1.7

53
Copyright 2017 Audinate Pty Ltd. All rights reserved
COLLISION DOMAINS - SEGMENTING

A Switch is considered to “segment the collision domain”

192.168.1.4 192.168.1.5 192.168.1.6 192.168.1.7

switch

54
Copyright 2017 Audinate Pty Ltd. All rights reserved
UNICAST TRANSMISSION
Computer Computer Computer Computer Unicast sessions are private
A B C D conversations between source and
destination.

Switch

Switch

Computer
X

Copyright 2017 Audinate Pty Ltd. All rights reserved


“LAYER 2” NETWORKS
• Switches are only concerned with MAC addresses
• There are 3 types of MAC address
• Host (the MAC address at Layer 2 of the device’s IP stack)
• Multicast – a specific group of destination MAC addresses that tells a switch to send out of
all interfaces (unless managed)
• Broadcast – a destination MAC address that tells a switch to forward out of all interfaces

• Some network technologies only used Layer 2


• MAC address to MAC address
• This network technique is better than just using a technique that cannot identify a device
• However this approach does not scale as well as a “full stack” implementation
• Operating systems require specific drivers to be written to handle the “missing part” of the
stack

56
Copyright 2017 Audinate Pty Ltd. All rights reserved
IP ADDRESSES IN A LAN? WHY?
• A “standard” network stack in an Operating System on a computer connects
ports to the application software running on it – through the IP address

• It is assumed that a consistent method is used to communicate with a


neighboring device and a device in the other side of the world – the “detail” is the
preserve of the network infrastructure.

• Important note: it is as important to understand what sticks the “layers” of a


network together as it is to understand the layers themselves

• For a device to understand how to communicate with other devices in the


network we use a process called ARP – Address Resolution Protocol to
resolve IP addresses to MAC addresses

57
Copyright 2017 Audinate Pty Ltd. All rights reserved
SENDING AND RECEIVING UNICAST
WHO IS
192.168.1.3

Device Device Device


A B C

Switch

Interface MAC Address

GE0/1
GE0/2
GE0/3
58
Copyright 2017 Audinate Pty Ltd. All rights reserved
SENDING AND RECEIVING UNICAST
WHO IS
192.168.1.3 ARP Table:
192.168.1.1 –
aa:bb:cc:dd:ee:01

Device Device Device


A B C

Switch

Interface MAC Address

GE0/1 aa:bb:cc:dd:ee:01
GE0/2
GE0/3
59
Copyright 2017 Audinate Pty Ltd. All rights reserved
SENDING AND RECEIVING UNICAST
Hello
ARP Table: 192.168.1.1
192.168.1.3 –
I am
aa:bb:cc:dd:ee:03
192.168.1.3

Device Device Device


A B C

Switch

Interface MAC Address

GE0/1 aa:bb:cc:dd:ee:01
GE0/2
GE0/3 aa:bb:cc:dd:ee:03
60
Copyright 2017 Audinate Pty Ltd. All rights reserved
BROADCAST DOMAINS

Boring
So let’s take a look at why IT
Broadcast message:
who has 192.168.1.47? managers are scared of
broadcast:

Here’s a small network. We


will issue an ARP and one
Boring computer responds. We only
switch Me
have to send a few ARPs and
everyone will have
discovered each other.

But what happens when the


network gets bigger?

61
Copyright 2017 Audinate Pty Ltd. All rights reserved
BROADCAST DOMAINS
Broadcast
message: who
Boring
Boring Boring Boring has Boring
192.168.1.47?
Boring Boring Boring Boring Boring
Boring Boring Boring Boring Boring Boring
Boring Boring Boring
Boring Boring Boring
Boring

Broadcast Broadcast
message: who message: who Boring
Boring Boring Boring Boring Boring Boring has
has
Boring Boring Boring
Boring Boring Boring Boring Boring Boring 192.168.1.47?
Boring Boring
192.168.1.47? Boring
Boring Boring Boring

Boring
Boring
Me! Boring Boring Boring

Boring Boring
Boring Boring
Me! Boring Boring Boring Boring Boring Boring

Boring
Boring
Me! Boring Boring Boring

Me!
Broadcast
Boring message: who
Boring Boring Boring
Boring Boring
Boring Boring
has Boring Boring
Boring Boring Boring Boring Boring Boring Boring Boring
192.168.1.47?
Boring Boring Boring Boring
Boring

Boring Boring Boring Boring


Boring Boring
Boring Boring Boring Boring Boring Boring Boring Boring
Boring Boring Boring Boring
Boring Boring Boring Boring
Boring Boring

62
Copyright 2017 Audinate Pty Ltd. All rights reserved
Surely there is a better way to deal with this?

63
Copyright 2017 Audinate Pty Ltd. All rights reserved
SEGMENTING BROADCAST DOMAINS – GOOD PRACTICE
Broadcast
message: who has
192.168.1.47? Boring Boring

Boring Boring Boring

Boring
Boring Me!

Boring Boring Boring

Boring
Boring Boring

64
Copyright 2017 Audinate Pty Ltd. All rights reserved
SEGMENTING BROADCAST DOMAINS – BUT
STAYING CONNECTED

Broadcast Broadcast
Domain A Domain B

65
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLANS - HISTORY
Originally when Network managers segmented networks without VLANs a different
switch would be used for each “group” of computers

router
device
device device

switch switch switch

device
device device device

66
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLANS - HISTORY

This led to “empty” ports on switches, and lots of extra wiring


Empty ports = wasted money!

router
device
device device

switch switch switch

device
device device device

67
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLANS - HISTORY
Why not put all these physical LANs on one switch?
Saves “wasted” ports – still same cabling

router
device
device device

Ports 1,2,3 Ports 4,5,6 Ports 7,8,9,10


= VLAN2 = VLAN3 = VLAN4

device
device device device

68
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLANS - HISTORY
Port based VLANs certainly saved money on expensive switch ports
Now 802.1q “tagging” addresses duplicate cabling – tag allows frames to be on same wire

router
device
device device

Ports 1,2,3 Ports 4,5,6 Trunk Ports 7,8,9,10


= VLAN2 = VLAN3 Port = VLAN4

device
device device device

69
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLAN TAGS – AND PVID (PORT VLAN ID)
• The switch will allow the
Ethernet Frame frame to leave through
Device any port that is a member
1. Preamble
of the same VLAN
2. SFD-Start of Frame
Delimiter
• The switch will prevent the
3. Destination MAC Address frame leaving through any
4. Source MAC Address
5. Ethertype port that is not a member
6. Payaload
7. CRC Block of the same VLAN

Switch
Switch adds
802.1q tag

70
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLAN TAGS – AND PVID (PORT VLAN ID)

• To access ports in the same VLAN on another


switch create a port that is “tagged” with
Switch multiple VLANS
- Cisco calls this a “trunk” port
• A Trunk – or multi tagged connection allows
frames with different VLAN tags to use the
same physical interface
• Tagged frames are only allowed to leave the
switch through ports in the same VLAN
Switch • When a tagged frame leaves a normal local
access layer VLAN port, the tag is removed.
• In a trunk port the tag is not removed.

71
Copyright 2017 Audinate Pty Ltd. All rights reserved
IPV4 IP ADDRESSES – HOST AND NETWORK
• Routers segment Broadcast domains – because they will not forward broadcast
traffic
– Routers do not care about MAC addresses (they obscure them – a good security feature)

• VLANs can be used to separate broadcast domains using 802.1q tags


• Routers can join IP subnets together
• There are 2 “kinds” of IP address
– “Host” addresses – the IP address that a device has
– “Network” address – the IP address that describes a group of hosts

• The size of an IP subnet is defined by the Subnet Mask. The subnet mask:
– Is a 4 octet number that is supplied via DHCP or an essential part of Static IP configuration
– Allows devices to know which IP addresses are “local” (same subnet) or “remote”
– Local IP addresses cause the device to issue an ARP
– Remote IP addresses are forwarded to the MAC address of the “default gateway” (if one is defined)
– If there is no defined “default gateway” the message will just be dropped

72
Copyright 2017 Audinate Pty Ltd. All rights reserved
IP SUBNET MASKS
• There are 32 bits worth of IPv4 addresses – this could be 1 network with 4.3 billion
devices in it
• Imagine a Broadcast domain that size!
• This is why we have IP subnets (to divide up that broadcast domain)
• The subnet mask is the “boundary” between the “network” address section of an IP
address and the “host address” section of an IP address
• A Valid Subnet mask is always a continuous line of binary 1s until the mask is reached,
then it becomes 0s
Subnet Mask Subnet Mask (binary) Subnet Number Number of
(decimal) Mask of hosts possible
CIDR per networks
network
255.255.255.0 11111111.11111111.11111111.00000000 /24 254 16.7 million
255.0.0.0 11111111.00000000.00000000.00000000 /8 16.7 256
million
255.255.255.252 11111111.11111111.11111111.11111000 /29 6 537 million

73
Copyright 2017 Audinate Pty Ltd. All rights reserved
VLANS AND SUBNETS
• A single VLAN is a single Broadcast domain
• An IP subnet is also a broadcast domain
• VLANs can therefore be described to a router as an IP subnet
• WARNING! – static IP addressing two IP subnets in the same
VLAN is not wise!
– When an IP broadcast packet is created in a VLAN – it is STILL
mapped to the broadcast MAC address
– At Layer 2 a device cannot distinguish between a broadcast frame
from another IP subnet
– Therefore an IP broadcast in a different IP subnet in the same VLAN
will still go to devices in the “other” IP subnet – This is a potential
security risk

• Tip – Make each VLAN its own IP subnet


74
Copyright 2017 Audinate Pty Ltd. All rights reserved
ROUTER “VIRTUAL INTERFACE” “VIRTUAL LAN”
Interface 1 –
Source MAC Address: (Device A) Untagged in VLAN 2 Frame tagged on ingress to
Destination MAC address: (default gateway) Interface 1 - VLAN 2

Device
A Frame packet Frame 802.1q
packet
S
tag =2

Source IP Address: (Device A) W


Destination IP address: (device B)
I
Router packet
Interface 2 – Tagged in T
802.1q Tag removed on
egress from Interface 3
VLAN 2 and VLAN 3
C Router creates new frame Router destroys original
encapsulating packet – frame, extracts packet
H NB new 802.1q tag
Device
Frame packet
B
Frame 802.1q
tag =3 packet
Source IP Address: (Device A)
Destination IP address: (device B)
Source IP Address: (Device A)
Source MAC Address: (default gateway) Destination IP address: (Device B)
Destination MAC address: (Device B) Interface 3 –
Untagged in Source MAC Address: (default gateway)
VLAN 3 Destination MAC address: (Device B)
75
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHY HAVE DIFFERENT SIZED SUBNETS?
/30 network gives
me 2 hosts
This link is
actually a
small (2 node)
network!
/25 gives me
126 hosts

/29 gives me 6
hosts

76
Copyright 2017 Audinate Pty Ltd. All rights reserved
FOR MOST EFFICIENT IP SCHEME SEGMENT
BIGGEST TO SMALLEST

Subnet ID First Host Last Host Broadcast IP Subnet Mask Number


of Hosts

192.168.1.0 192.168.1.1 192.168.1.126 192.168.1.127 255.255.255.128 126


192.168.1.128 192.168.1.129 192.168.1.134 192.168.1.135 255.255.255.248 6
192.168.1.136 192.168.1.137 192.168.1.138 192.168.1.139 255.255.255.252 2

Should I use Static or Dynamic IP addresses?


• Normally Dynamic is easier, and less prone to human error
• Static is sometimes preferable for servers (but not essential)
• Static IP addresses are only essential when using Organization SSL certificates
(the only “excuse” acceptable to obtain a public static IPV4 address in Europe)

77
Copyright 2017 Audinate Pty Ltd. All rights reserved
ADVANCED DANTE
NETWORKING: SECTION 4

78
IN THIS SECTION…

Quality Of Service
• What QoS is, and what it is not
• When should you use it?
The “Third Kind” of communication
Multicast
• Why is Multicast useful
• How do we manage Multicast?
• IGMP – Deep Dive

79
QUALITY OF SERVICE
(QOS)

80
WHAT IS QUALITY OF SERVICE (QOS)?

• Quality of Service (QoS) is the business of traffic prioritization


and management
• The aim is to provide a specific level of service to an application
using the network
• How do I specify a level of service?
- Throughput
- Delay
- Delay variation
- Packet loss

81
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHAT IS QUALITY OF SERVICE (QOS)?
• QoS comes in two basic forms:
- Class based, using techniques like queueing and prioritization (e.g.
Diffserv)
o Relative

o Specify what is important, what is less important

o Simpler to implement

- Reservation based, using techniques like rate control, traffic shaping,


and admission control (e.g. QoS NSLP, or Intserv)
o Absolute

o Specify how much, how often, and on what terms - then make a
decision if it is possible
o Complex to implement

82
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHAT IS QUALITY OF SERVICE (QOS)?

Audio over IP only really cares about some aspects:

• Delay variation – helpful to minimize latency and achieve good


PTP performance under a wider range of conditions

• Aggregate throughput – we need to get a lot of audio packets


through the network

• Packet loss – we need all the audio to arrive

83
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHY DO IT?

• QoS techniques can help extend performance further into a


region where bandwidth limitations might otherwise cause
problems
- Help to squeeze more out of the infrastructure
- Running a link at over 60-70% utilization
- Mixing 100Mbps and 1Gbps in the same network
- Running a shared services network (e.g. audio and video, enterprise IT etc)

• QoS is not magic, it cannot manufacture better performance

84
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHY DO IT?

• Simple fact: The best QoS available is bandwidth


- Cheap, plentiful, easy to use
- Dante will present a predictable peak load to the system – this should enable
you to ensure that the network is appropriately dimensioned

• Key things to remember:


- Dante devices transmit relatively small, fixed size packets that are “nicely”
distributed by the transmitter
- This helps to minimise queueing in the switch, and greatly reduces the need for
QoS to achieve good performance

85
Copyright 2017 Audinate Pty Ltd. All rights reserved
CLASS BASED QOS

• Class based QoS is the most widely deployed technique in COTS Ethernet
switches
- We can thank VoIP for this!
• Uses a tag written into the IP or Ethernet header of each packet
- Just a number, no special meaning
• Switch needs to inspect each packet and map to a specified queue at the
output port
• The queues are then emptied following a set of rules that govern the process
- These rules define the level of service experienced by the traffic in each class

86
Copyright 2017 Audinate Pty Ltd. All rights reserved
CLASS BASED QOS
• Has many advantages
- Simple
- Low per-packet overhead
- Implement in switch silicon
- Distributed
- Flexible – network can apply tags, change them, interpret them using a
system specific policy etc.
- Very good at protecting high priority traffic from lower priority

• Disadvantage
- Can overload within a traffic class – e.g. everything high priority…
- When everything is important… nothing is
- Not a limitation in vast majority of cases

87
Copyright 2017 Audinate Pty Ltd. All rights reserved
DIFFSERV
• DSCP values are mapped to a queue at each output port in the switch
- In a Dante centric network the highest priority queue is used for PTP traffic (very small
payload)
- The next level of priority is used for Audio data – slightly larger
- The level of priority below that is used for time critical control data
- And finally the last queue is available for everything else

• Queues in switches send frames onto the wire after applying this “sorting”
criteria
- Don’t forget this is per-switch, so all switches should have the same configuration if you want
consistent behaviour

• Extremely well suited to ensure that high priority traffic is not affected by lower
priorities
- PTP amounts to only a few packets per second per device
- Audio is several thousand packets per second per flow
88
Copyright 2017 Audinate Pty Ltd. All rights reserved
EXAMPLE
• Airline check in counter or boarding lanes
- Customers are assigned to a class based on importance to the airline
(Platinum, Gold, the rest…)
- Equivalent of a DSCP Tag
- Important customers have a priority check-in counter
- Equivalent of a dedicated queue
- When it is not busy, all customers served quickly
- When it gets busy, important customers are served quickly

• Why does this work?


- There are relatively fewer priority customers than other categories
- Minimal wait for priority customers at check-in, at boarding…

• When does this fail?


- At Chicago O’Hare United terminal, where everyone is Platinum
- If everyone is important, no one is…

89
Copyright 2017 Audinate Pty Ltd. All rights reserved
SIMPLIFIED QUEUE MODEL
Egress
Queue
Queue 1
Egress Port
Queue 2
Ingress Port DSCP
Mapping Queue 3

Queue 4

Switch Egress
Fabric Queue
Queue 1
Ingress Port Egress Port
Queue 2
DSCP
Mapping Queue 3

Queue 4

90
Copyright 2017 Audinate Pty Ltd. All rights reserved
QUEUEING MODES

Switches can be configured to manage queues using different algorithms:


- Strict Priority – Always serve the most important class
- Weighted Round Robin – Share service time amongst the queues following
per-queue weights
- Shaped Round Robin – Share service time amongst queues in a statistically
shaped manner (Cisco use this term)

91
Copyright 2017 Audinate Pty Ltd. All rights reserved
QUEUEING MODES

• Dante needs a strict priority queue for PTP traffic


- This will ensure that if a sync packet tagged DSCP 56 arrives it will always be
passed through the as quickly as possible
- Using Strict Priority minimizes delay variance experienced by PTP traffic
through the system, particularly if under heavy load
- WRR or SRR will increase delay variance and reduce PTP performance

• Many data-center switches default to SRR or WRR modes – you need to


check for this if using QoS on higher spec switch products

92
Copyright 2017 Audinate Pty Ltd. All rights reserved
DANTE DSCP VALUES

DSCP
Priority Usage Hex Decimal Binary
Label
High Time critical PTP events CS7 0x38 56 111000
Medium Audio, PTP EF 0x2E 46 101110
Low (reserved) CS1 0x08 8 001000
None Other traffic BestEffort 0x00 0 000000

These values chosen to align with default mapping in many switches

93
Copyright 2017 Audinate Pty Ltd. All rights reserved
CISCO SG300

• Switches like the Cisco SG300 can


have QoS queueing set up very simply
• All egress queues can be set to strict
priority
- Highest priority queue is emptied first,
followed by next in order etc.
• It is simple to set up – set switch to
Trust DSCP, and then assign DSCP
tags according to previous table… job
done

94
Copyright 2017 Audinate Pty Ltd. All rights reserved
SOME OTHER SPECIFICS – TRUST DSCP?

• Tells the switch to trust that the endpoints will tag traffic correctly,
not abuse high priority
• Useful with services like Dante that do not require user
configuration – the endpoint can only generate correct DSCP tags
• What if I turn this off? - the switch will “strip” DSCP tags from all
packets and ignore them
• Most campus, enterprise, and carrier networks will NOT trust
endpoint DSCP tags, and instead apply a policy as defined by the
system admin

95
Copyright 2017 Audinate Pty Ltd. All rights reserved
SOME OTHER SPECIFICS – DSCP REMAPPING

• Many switches allow you to strip or remap DSCP tags on ingress

• Define ACLs on ingress and handle matching packets

• Very common at borders between networks

• Useful for protecting PTP in non-trusted environments, or making PTP fit


within the scheme in operation on that network

• Example: Match PTP traffic on UDP port 319, 320, 321, and apply DSCP 56

96
Copyright 2017 Audinate Pty Ltd. All rights reserved
SOME OTHER SPECIFICS – CLASS OF SERVICE (COS)

• What about COS, or Class of Service?


- COS relies on a Layer 2 method for tagging frames
- Part of the 802.1q header (remember VLANs?)
- Only tagged VLAN traffic will have an 802.1q header
- Most endpoints can’t generate 802.1q tags
- Not useful end-to-end and will most likely be stripped once the traffic hits a gateway
- Note: Even though they might change, DSCP values can be retained end-to-end,
no matter what link type

• What about IPP or IP Precedence?


- Supports less resolution than DSCP, so used less frequently in larger systems

97
Copyright 2017 Audinate Pty Ltd. All rights reserved
ENTERPRISE SWITCH FEATURES
• On larger (more powerful, and more expensive) enterprise
switches QoS can be more “flexible” (read complicated)
• Switches in this category allow the network operator to cater for
many different services
- Leverage the flexibility and power inherent in IP

• It is important to understand exactly how the QoS settings on the


particular platform are implemented
- Manufacturer training courses are best for this
- There can be a lot of detail here

98
Copyright 2017 Audinate Pty Ltd. All rights reserved
IS IT LIKELY TO MAKE A DIFFERENCE?

• Using QoS will “catch” momentary “spikes” in demand


- These are generally caused by untagged DSCP traffic

• If you are moving 512 channels of Dante over the network link in question, you still
have “headroom” to accommodate other services (more than 100mbps)
- How many people have 100mbps Internet connection?
- How many people see 100mbps of Internet traffic on the same network segment as AV?

• You will generally need to have a pinch point somewhere in the network in order to
see a positive benefit from QoS configuration
• You are also likely to see negative impacts from over configuration in environments
where QoS configuration is not necessary

99
Copyright 2017 Audinate Pty Ltd. All rights reserved
MULTICAST

100
MULTICAST IP ADDRESSING
• A multicast group is defined by a “multicast destination IP address”
- Any host within the address scope can “listen” to the group address
- Really really useful when there are many more receivers than transmitters
• Special reserved IP and MAC addresses indicate multicast destinations
- MAC addresses start with 01:00:5E
- IPv4 range: 224.0.0.0 – 239.255.255.255
- IPv6 Range: FF00::/8
• Multicast IPv4 addresses do not map 1:1 to Multicast MAC addresses
- Up to 32 Multicast IP addresses for each Multicast MAC address
• This creates challenges for a L2 switch uniquely identifying multicast streams
- Hence the use of “IGMP Snooping”

101
Copyright 2017 Audinate Pty Ltd. All rights reserved
INTERNET GROUP MANAGEMENT PROTOCOL (IGMP)
• Internet Group Management Protocol (IGMP) is a protocol used to manage
multicast traffic in IPv4 routed networks
- IPv6 replaces IGMP with “Multicast Listener Discovery (MLD)”

• Used to help manage multicast flooding


- Without it, multicast traffic needs to go everywhere - not ideal
- Multicast management is generally a good thing for high rate, media traffic

• Protocol is responsible for working out which networks or hosts need a specific
multicast group
- The router sends query messages to all hosts on each network
- Hosts respond with the groups they want to join
- What about inside the VLAN?

• All Dante devices implement IGMP (as a host)

102
Copyright 2017 Audinate Pty Ltd. All rights reserved
IGMP QUERIER
• Most switches today are able to function as an IGMP Querier
- Querier periodically sends query message to all hosts on 224.0.0.1
- What group addresses do you want?
- Hosts respond with “membership report” to 224.0.0.2 (or 224.0.0.22)
- I want A.B.C.D
- You should only have one active Querier per VLAN
- Multiple queriers *can* be configured amongst some manufacturers equipment

• Intermediary L2 switches “snoop” on IGMP traffic


- Membership report identifies which interfaces need each multicast group
- Switch then maps groups to ports and builds a multicast forwarding table
- Multicast traffic only appears on the ports it is needed

103
Copyright 2017 Audinate Pty Ltd. All rights reserved
IGMP AT WORK This is an IGMP
General Query

IGMP Querier
Switch with IGMP Snooping enabled
1 2 3 4 5 6

Device Device Device Device Device


Device A
B C D E F

Copyright 2017 Audinate Pty Ltd. All rights reserved


IGMP AT WORK This is an IGMP
General Query

IGMP Querier
Switch with IGMP Snooping enabled
1 2 3 4 5 6

Device Device Device Device Device


Device A
B C D E F

General report
General report General report
239.255.255.32 General report
239.255.255.32 239.255.255.32
239.255.255.33
239.255.255.64 239.255.255.33
239.255.255.64

Copyright 2017 Audinate Pty Ltd. All rights reserved


IGMP AT WORK This is an IGMP
General Query

IGMP Querier
Switch with IGMP Snooping enabled
1 2 3 4 5 6

Device Device Device Device Device


Device A
B C D E F

General report
General report General report 239.255.255.33
General report
239.255.255.32 General report
239.255.255.32 239.255.255.32 239.255.255.33
ME! General report
239.255.255.33
239.255.255.64 239.255.255.33 239.255.255.64
239.255.255.64

Copyright 2017 Audinate Pty Ltd. All rights reserved


IGMP AT WORK This is an IGMP
General Query
Snooping table
Port1: 239.255.255.32,
239.255.255.33,
239.255.255.64.
IGMP Querier Port3:
Switch with IGMP Snooping enabled 239.255.255.32,
239.255.255.64
1 2 3 4 5 6 Etc…

Device Device Device Device Device


Device A
B C D E F

General report
239.255.255.33 General report General report
239.255.255.33 239.255.255.33
General report
239.255.255.32 General report
ME! 239.255.255.32 239.255.255.32
ME! 239.255.255.33
ME! General report
239.255.255.33
239.255.255.64 239.255.255.33 239.255.255.64
239.255.255.64

Copyright 2017 Audinate Pty Ltd. All rights reserved


IGMP SNOOPING OPTIMISATION

• Running IGMP to manage IPv4 multicast involves:


- Specific roles in the network Querier
- Timers: Query Intervals, Timeouts etc.
- Specific Multicast Addresses: 224.0.0.1, 224.0.0.2, 224.0.0.22 etc.
- Multiple versions: IGMPv2, IGMPv3 etc.
- Protocol implementations from different switch vendors – vendor specific
optimisations: e.g. fast leave, querier proxies etc.

• It can be complex…
- Especially if you start randomly changing timers or switch configurations
- Random changes will have random effect

108
Copyright 2017 Audinate Pty Ltd. All rights reserved
SIMPLIFIED RECIPE

• Set the Querier interval quite short – 15 or 30 seconds is OK


- Minimizes potential for gaps in audio, increases responsiveness

• Leave timeout values at default settings


- Longer is better in most cases

• Ensure that there is only active querier per VLAN.


- Ideally this is a core network switch, or the router port
- Symptom: Intermittent audio
- Diagnosis: Wireshark shows IGMP Query messages from multiple sources

Copyright 2017 Audinate Pty Ltd. All rights reserved


SIMPLIFIED RECIPE

• Avoid using fast leave


- Doesn’t offer anything

• Avoid IGMP proxies


- Unless you are CERTAIN you know how it behaves

• Ensure that “block unregistered multicast” is not set on Netgear switches


- Blocks traffic it should not (mDNS, PTP)

• Dante devices support IGMPv3

Copyright 2017 Audinate Pty Ltd. All rights reserved


IGMP SNOOPING PERFORMANCE CONSIDERATIONS
• Remember – switches operate using MAC address table
• There are possibly multiple IP addresses per MAC address
- Work needs to be done by the switch to resolve this

• Some switches also use a CPU to forward multicast traffic


- This can lead to variability in forwarding delay
- Potential for overload as the amount of multicast increases

• This means some switches can be slow dealing with multicast


- Potentially can cause unacceptable latency for multicast audio or PTP

• Common Symptoms Include:


- Devices losing sync periodically
- Choppy audio when receiving a multicast stream

Copyright 2017 Audinate Pty Ltd. All rights reserved


MULTICAST PROPAGATION
• It is a common requirement that the network provides L2 VLANs over a L3 core
network
- Multicast is sometimes a challenge in those environments
- Many system admins don’t like multicast traffic in these environments

• PTP likes the forwarding delay through the network to be consistent and
symmetric
• In some scenarios this may not be the case
- Asymmetric non-multicast technologies such as MPLS, GPON etc.
- ACLs incorrectly configured etc.

• Common symptoms include:


- Multiple PTP masters
- Devices failing to maintain sync

• Turning ON unicast delay requests for all devices can help diagnose, and
possibly resolve the problem

Copyright 2017 Audinate Pty Ltd. All rights reserved


TESTING MULTICAST SWITCH PERFORMANCE

Audio Testing

Increase
Create some Check the See how far Check
Subscribe Turn on IGMP number of
Dante latency view the latency latency
them to a snooping in multicast
Multicast in Dante view has performance
receiver the switch flows through
flows controller moved again
switch

PTP Testing

Check the Increase


Create some Subscribe clock Turn on IGMP See if clock number of Check sync
Dante them to a histogram in snooping in histogram multicast performance
Multicast receiver Dante the switch has spread flows through again
flows
controller switch

113
Copyright 2017 Audinate Pty Ltd. All rights reserved
ADVANCED DANTE
NETWORKING: SECTION 5

114
IN THIS SECTION…

Understanding network resources for


Dante
• Bandwidth use by Dante
• Effect of topology on bandwidth use

115
DEFINING RESOURCE REQUIREMENTS
• Any project in anything requires this phase of planning

• Required resources for a Dante network:


ü Enough Transmit flows to serve all receivers
ü Enough Bandwidth to carry flows
ü Unblocked logical connections (“wire” is not cut, and is “plugged in”)
ü Enough Receive flows available on devices connecting to transmitters

116
Copyright 2017 Audinate Pty Ltd. All rights reserved
WHAT DANTE PUTS ONTO THE WIRE
• In order to make a complete audio network solution we need:
- Audio transport
- Clocking
- Device discovery
- Device and routing control
• Using the IP over Ethernet model allows us to do this on 1 cable
• Some devices also take their power from this same cable!
• The OSI model gives a framework to do this very tidily

117
Copyright 2017 Audinate Pty Ltd. All rights reserved
ARCHITECTURAL PRINCIPLES
• What are the main considerations when architecting a network?
- Identifying required services
- Bandwidth utilization
- Specifying an optimal infrastructure for predicted bandwidth demands
- Identifying potential “bottlenecks” – designing them out or managing them
- Ensuring a scalable architecture (the network will grow over time)
- Understanding possible future service requirements

• A long and daunting list – luckily there are established “rules”

118
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps
0 mbps 6 mbps
0 mbps

Switch 6 mbps

119
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps
0 mbps 12 mbps 6 mbps 0 mbps
0 mbps

Switch 12 mbps

120
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps 0 mbps
6 mbps 12 mbps 6 mbps 0 mbps
0 mbps 6 mbps

Switch 18 mbps

121
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps 0 mbps 0 mbps


12 mbps 12 mbps 6 mbps 0 mbps
0 mbps 6 mbps 6 mbps

Switch 24 mbps

122
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps 0 mbps 0 mbps


12 mbps 12 mbps 6 mbps 0 mbps
0 mbps 6 mbps 6 mbps

Switch 24 mbps
0 mbps
0 mbps

Switch 0 mbps

Internet

123
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps 0 mbps 0 mbps


12 mbps 12 mbps 6 mbps 0 mbps
0 mbps 6 mbps 6 mbps

Switch 24 mbps
0 mbps
0 mbps
8 mbps
Switch 8 mbps
8 mbps 0 mbps
Internet
0 mbps

124
Copyright 2017 Audinate Pty Ltd. All rights reserved
BANDWIDTH UTILIZATION – DANTE FLOW
Device Device Device Device Device
A B C D E

6 mbps 0 mbps 0 mbps


12 mbps 18 mbps 6 mbps 0 mbps
0 mbps 6 mbps 6 mbps

Switch 30 mbps
0 mbps
6 mbps
8 mbps
Switch 14 mbps
8 mbps 0 mbps
Internet
6 mbps

125
Copyright 2017 Audinate Pty Ltd. All rights reserved
UNDERSTANDING RESOURCES - BANDWIDTH
• A flow is a sequence of packets sent to an IP address / port
• A Unicast flow contains 4 channels
- 2ch on Ultimo2 devices
- At 48KHz sample rate a unicast flow uses 6mbps bandwidth
- If only one channel is “subscribed” the other 3 channels are “sent” as silence
- As other channels are added between the 2 devices audio content from the successive channels
“dumped” into the flow
- New flows not created if there is “space” in existing flows

• A Multicast flow contains up to 8 channels


• At 48KHz sample rate a multicast flow uses up to 12mbps bandwidth
- Only selected channels are included in the flow
- Cannot add new channels to an existing multicast flow
- Avoid creating multiple “single channel” Multicast flows that follow similar paths

126
Copyright 2017 Audinate Pty Ltd. All rights reserved
UNDERSTANDING RESOURCES - FLOWS

• A flow limit is an internal device resource limit

• Different device types have different limits

• It does not matter to the device whether the


flow is unicast or multicast (as the destination
IP address simply changes)

Copyright 2017 Audinate Pty Ltd. All rights reserved


UNDERSTANDING RESOURCES - FLOWS

• Each Dante device has a number of “Transmit Flows” and


“Receive Flows” available to it
- An Ultimo device has 2 transmit and 2 receive flows
- A Dante Virtual Soundcard has 16 transmit and 16 receive flows
- A Brooklyn II device has 32 transmit and 32 receive flows
- A Dante PCIe card has 32 transmit and 32 receive flows
- A Dante HC device has 128 transmit and 128 receive flows

• Flows are used to increase the overall efficiency of the


system

Copyright 2017 Audinate Pty Ltd. All rights reserved


UNDERSTANDING RESOURCES - TX FLOWS

• A Tx flow is created in 2 different ways


In Unicast mode –
- The Tx flow is created when the receiving device requests one (or more) channels
- If room exists channel is added to existing TX flow
- A single unicast flow can contain up to four channels

In Multicast mode –
- The Tx flow is created immediately (it doesn’t need a receiver to request it)
- A single Multicast flow can contain up to eight channels
- Cannot add channels to a multicast flow

• All Dante flows are UNICAST by default


• Only intentional programming can create a multicast flow – in Dante
controller – this is done by a human

Copyright 2017 Audinate Pty Ltd. All rights reserved


UNDERSTANDING RESOURCES – TX FLOWS

• A Dante flow is Unicast by default

• Optimisation can be achieved using Multicast when


distributing one to many

• When channels are made Multicast – those


channels will be transmitted in a new multicast flow

• If a device is using all flows – no more subscriptions


can be made by receivers

• Tx Flow usage is visible on the Transmit tab of


Dante Controller

Copyright 2017 Audinate Pty Ltd. All rights reserved


UNDERSTANDING RESOURCES – RX FLOWS

• A receiver will utilize a new RX flow resource each


time a subscription is created that either:
- Requires a new unicast TX flow on the transmitter, or
- Is the first subscription to a channel already in a multicast flow

• Rx Flow usage is visible on the Receive tab of


Dante Controller

Copyright 2017 Audinate Pty Ltd. All rights reserved


SIMPLE NETWORK RESOURCE UTILIZATION

Recap
An audio network consists of 3 things:
• Devices that create flows (microphones, stage
boxes, DSPs, playback sources etc.)
• Infrastructure that carries the data
• Devices that receive flows (DSPs, Amplifiers,
Powered Loudspeakers, Recorders etc.)

Copyright 2017 Audinate Pty Ltd. All rights reserved


ADVANCED DANTE
NETWORKING: SECTION 6

133
IN THIS SECTION…

Subscriptions in Dante
• Subscription process
• Multicast subscriptions
• Topology and QoS

134
SUBSCRIPTION PROCESS (UNICAST)
1. Plug laptop (running Dante Controller) into switch
2. Plug Microphone into switch
1. Microphone “announces” itself (multicast)
“Hello my
name is
2. Dante controller “hears” the announcement
Bob” Paul, Please and populates routing page
Can I do
subscribe your 3. Plug Speaker into switch
channel1 to
that Hmm
channel1@Bob 1.Speaker “announces” itself (multicast)
latency? … Bob
2.Dante Controller “hears” the announcement
Hmm …
Paul
and populates routing page
Here is
channel1 at
4. We decide to “subscribe” Paul to Bob (from Dante
2ms latency Controller)
“Hello my 5. Dante controller “tells” Paul to ask Bob for
name is Paul” channel1
6. Paul “asks” Bob for channel1 at it’s latency setting
Bob, can I have 7. Bob checks that latency setting is = or > own
channel1 at 1ms
0mbp latency please? setting
s 8. If this is true flow is created with Paul’s latency
6mbps
6mbp 9. If not flow is created with Bob’s latency
s Paul here, my
channel1
10. 6mbps bandwidth is used between Bob and Paul
0mbps
successfully (in one direction)
subscribed to 11. 0mbps bandwidth is used on cable between
channel1@Bob
switch and computer
12. Paul reports “good” subscription (multicast)
13. Controller updates record (green check)

Copyright 2017 Audinate Pty Ltd. All rights reserved


MULTICAST SUBSCRIPTIONS
• Dante is Unicast by default (for good reason)
• In Unicast – bandwidth is only used between a receiver and a
transmitter:
- When an active subscription is made
- And on the shortest path between receiver and transmitter

• When a multicast flow is created


- It doesn’t care about a receiver
- If there is no subscriber – it still uses bandwidth - everywhere (in an
unmanaged network)

• Dante CANNOT dynamically switch Unicast to Multicast


- BUT if a multicast flow is created AFTER unicast subscriptions have already
been made, they will switch to use the new multicast flow

136
Copyright 2017 Audinate Pty Ltd. All rights reserved
EXAMPLE - IN A DANTE NETWORK
64 ch 64 ch 64 ch 64 ch 64 ch 64 ch 64 ch
device device device device device device device

Switch

64 ch 64 ch 64 ch 64 ch 64 ch 64 ch 64 ch
device device device device device device device

Unicast Multicast
• 14 Devices, 64 channels each • IF all devices set to full Multicast
• 32 flows each maximum –unicast b/w per node =192 mbps • Switch would still be fine…
• 192 x 14 = 2688 mbps – more than a gigabit • Devices would not!
• Maximum Rx bandwidth per node is still 192 mbps unicast • 8 x 12 mbps multicast flows = 96 mbps x 14 nodes = 1.344 gbps
(run out of flows) • More than each cable can handle!
• Minimum backplane bandwidth of switch 28 gbps (still more • BUT only still 4.8% of the backplane bandwidth
than 10x actual bandwidth) • Sending 832 channels into a network where they “flood” and the
• “Worst” flow structure possible = 32 x 2 channel flows on a maximum receive possible is 64 channels seems like a huge problem
device …
• Only a “2 way” split achievable unicast

137
Copyright 2017 Audinate Pty Ltd. All rights reserved
LET’S REVISIT OUR ORIGINAL EXAMPLE
• With IGMP snooping off and all
devices multicasting – we are
64 ch 64 ch 64 ch 64 ch 64 ch 64 ch 64 ch
trying to push 1.344 gbps down
device device device device device device device each gigabit cable
• Obviously never going to work!
• Turn on IGMP snooping
Switch IGMP Snooping • Each device can only subscribe to
= ON
32 flows - This is a maximum
bandwidth of 384 mbps per port
64 ch 64 ch 64 ch 64 ch 64 ch 64 ch 64 ch • The backplane bandwidth is
device device device device device device device unchanged at about 4.8% capacity
• This is why multicast
management is always a priority
over QoS

Copyright 2017 Audinate Pty Ltd. All rights reserved


DANTE WITH QOS
64ch
device
64ch
device
64ch
device
64ch
device
64ch
device
64ch
device
64ch
device
• Assume all 14 devices on top
switch are transmitting on a 1:1
mapping to the 14 devices on the
switch bottom switch
• 14 x 64 = 896 channels
• 16 flows per device
64ch 64ch 64ch 64ch 64ch 64ch 64ch @ 6mbps x 14 = 1344mbps
device device device device device device device
• Simply put – on a gigabit link…
• The Bandwidth has run out!
64ch 64ch 64ch 64ch 64ch 64ch 64ch
device device device device device device device
• Solution? 10 gigabit link –
Problem solved!
switch
• All of these devices are
transmitting a constant data rate
64ch 64ch 64ch 64ch 64ch 64ch 64ch
device device device device device device device • QoS will do NOTHING here!

139
Copyright 2017 Audinate Pty Ltd. All rights reserved
TOPOLOGY AND QOS
• The previous example shows a “Classic” model “Collapsed Core” Model

bottleneck being created by a edge edge edge edge edge edge edge edge

sub optimal topology


• Luckily there are established distro distro

methods for designing network


topologies
core core
• The “three layer hierarchical
model” describes the concept
• As Switch and router speeds distro distro

have increased “collapsed core”


has become more normal edge edge edge edge edge edge edge edge

140
Copyright 2017 Audinate Pty Ltd. All rights reserved
ADVANCED DANTE
NETWORKING: SECTION 7

141
IN THIS SECTION…

Advanced clocking
• How does a network synchronize media
clocks?
• What does an external clock do?
• Network synchronization
• Synchronizing time
• Time in PTP

142
HOW DOES A NETWORK SYNCHRONIZE MEDIA CLOCKS?

1 “cycle” of Wordclock
24 “bits” in 1
Wordclock Cycle

In AES3 there are 2 • Simplified example


Channels (2x24bits) • “Real” AES3 contains other data frames too
• This is more like a format called LJ I2S

143
Copyright 2017 Audinate Pty Ltd. All rights reserved
SO WHAT DOES AN “EXTERNAL” CLOCK DO?

Phase Locked Loop

144
Copyright 2017 Audinate Pty Ltd. All rights reserved
HOW A DISTRIBUTED WORD-CLOCK WORKS

Master Clock
Audio Device
Device

• Speed = Distance/time How far does the signal travel in 1uS?


- 2.8x108 x 1x10-6 = 280m
• How often do we need to synchronize over this distance?
• In Live sound and Broadcast – a lot of the time (think Golf, Football,
Stadium Concerts, Motorsport… etc)
• Dante guarantees microsecond sync at any point in the network…. How?
145
Copyright 2017 Audinate Pty Ltd. All rights reserved
NETWORK SYNCHRONIZATION
That’s a
good
idea! yeah

Don’t tell me
how the clock
works… just
tell me what
time it is!

yeah yeah

146
Copyright 2017 Audinate Pty Ltd. All rights reserved
SYNCHRONIZING TIME

• The idea of distributing time over a network started


with British Railways

• As rail networks grew – it became necessary to


have a timetable

• It was not acceptable for a train to just show up,


and leave at arbitrary times – think of how
annoying that was!

• How did this problem get solved? (remember – no


telephones!)

147
Copyright 2017 Audinate Pty Ltd. All rights reserved
SYNCHRONIZING TIME

Set
watch Set Edinburgh
to Big station time to
Ben Put watch
watch on
train
Go to

London Edinburgh

148
Copyright 2017 Audinate Pty Ltd. All rights reserved
TIME IN PTP
• Of course – synchronizing in seconds is
nowhere near good enough for audio!

• Luckily IEEE1588 aka PTP Precision Time


Protocol is a lot better!

• IEEE1588 has a resolution of nanoseconds

• The PTP master device has the reference


time

• It transmits this reference time onto the


network (think train)

149
Copyright 2017 Audinate Pty Ltd. All rights reserved
TIME IN PTP

• The problem with a railway system is that the


physical clock would have to be loaded and
unloaded from a physical train – only one station
could synchronize at a time

• Wouldn’t it be better if we could simultaneously


achieve this?

• Luckily on a network – that’s exactly what we do

• Just like the railway – it takes time for the


message to propagate across the network – PTP
also nicely takes care of that!

150
Copyright 2017 Audinate Pty Ltd. All rights reserved
PTP MECHANISM
Set time to
0900:01.00 1. The Master clock sends a
0325364
Sync Message 1435 I counted 1410 sync message.
its Monday 8th May nanoseconds 2. This allows the slave
2017 between these
0900:01.000325364 times – speed up devices to set their individual
my clock time pieces
Follow up 3. BUT devices run at slightly
Ref 1435 – was
actually sent at
different ”speeds”
0900:01.000326789 4. To address this the master
clock sends a sync follow-up
Set time to message.
0900:01.00 5. The receiving clock can then
0325364 compare the time that it
I counted 1445
nanoseconds
received the sync message
between these against the actual time from
times – slow the follow-up to determine if
down my clock it is running faster or slower
than the master.

151
Copyright 2017 Audinate Pty Ltd. All rights reserved
PTP HAS TO WORK ACROSS A NETWORK

• We saw how frames can even slightly potentially “get in the way”
of each other

• This is also true for PTP frames (they aren’t magic)

• In reality the slave device needs to “get an average” of the


messages it receives from the master to not “jump to conclusions”

• PTP also considers how to be robust in a network

152
Copyright 2017 Audinate Pty Ltd. All rights reserved
PTP HAS TO WORK ACROSS A NETWORK
Delay Requests
Delay request 1066
Received at
0900:02.001325745 Without ”real” information the slave doesn’t have
Reply sent
0900:02.008564367
a good idea about the network delay

1. A Slave sends a delay request message


back to the master asking what time it
received the request.
2. The master replies with the time it received
the delay request.
3. This gives the slave a good idea of what the
Delay request
message 1066 – its delay caused by the network is.
0900:02.00567283 4. It can use this information, along with the
What time did you
get this? sync packets to predict how much it should
“pull” its local timepiece to stay aligned.

153
Copyright 2017 Audinate Pty Ltd. All rights reserved
CONTINUING ON…

• The master continues to send sync and follow up messages


• The slaves continues to send delay request messages
• Dante has a “sync interval” of 0.25 seconds – 4 packets per second
• The slaves obtain tight synchronization very quickly

154
Copyright 2017 Audinate Pty Ltd. All rights reserved
CONTINUING ON…
• Now… if we remove the Master clock some time later
- The slaves have a very good idea of what the “relative speed” of the master clock was
(and can maintain this on their own) … think a speed limit sign on the highway (and
assume you follow the rules)
- The slaves also have a very accurate idea of what the time is at the master device –
and will have set their own timepieces accordingly

• This means
- Any Slave can now easily (and quickly) become the master
- The subsequent slaves are synchronous enough with the “new master” that their audio
clock is unaffected
- The synchronization of the system is not dependent on a single permanent fixed master
- Time is Time – I can “chop” it into however many parts (Hz) I like
- Audio and Video clocks are simply the inverse of time – 1Hz = 1S^-1 after all

155
Copyright 2017 Audinate Pty Ltd. All rights reserved
ADVANCED DANTE
NETWORKING: SECTION 8

156
IN THIS SECTION…

Troubleshooting
• Fundamentals of troubleshooting
• Dante port addresses
• Network ports
• IP addresses
• MAC addresses
• Common symptoms and causes
• Talking to an IT department

157
TROUBLESHOOTING
• Networks are just about connecting devices together
- Understanding how connections can be “broken” is the key to troubleshooting
• Recap – Dante puts the following “traffic” into the network
- Timing – PTP IEEE1588
- Audio Data
- Commands for creating routes and controlling Dante functions
- Discovery services (so that commands can be sent to devices)
• Just because you cannot “see it” doesn’t mean it cannot be seen!
- Think about applications attaching to the network stack using “port addresses”
- Many Operating systems use software “firewalls” to prevent certain applications
accessing the network stack
- If you can “see” a master clock (identified at top of Dante controller routing screen)
but no devices, or device names with no information, 99% of the time – the firewall
on the machine running Dante controller is blocking something

158
Copyright 2017 Audinate Pty Ltd. All rights reserved
TROUBLESHOOTING - DANTE “PORT” ADDRESSES

Traffic used by Dante is as follows:


mDNS 224.0.0.251:5353
Control and Monitoring 224.0.0.230 – 232:8007-8706
PTP 224.0.1.129 – 132:319-320
Multicast Audio 239.255.0.0/16:4321
Unicast Audio RX Unicast IP:14336 – 14600
AES67 Multicast Audio 239.XX.0.0/16:5004

– All Dante traffic is UDP/IP


– This means that if any traffic to/from these ports and IP addresses is
blocked, then the ”wire” carrying that particular service can be
considered “cut!”

159
Copyright 2017 Audinate Pty Ltd. All rights reserved
NETWORK PORTS – 2 WAYS TO CUT THE WIRE

• A computer’s operating system may use a “firewall”


- Look at which applications within the Operating System are trying
to access the network
- Apply rules to these connections (permit or deny)
- Look at which destination IP addresses (and ports) are being used
– can permit/deny on this basis
- Can determine whether a network is “private” or “public” and
decide to permit/deny on this basis
- Can look at how much data has been transferred to/from a specific
port and/or IP address and decide to trigger rules on a “threshold”

• Firewall Appliances are simply computers in the network at


gateway points – they work in a similar way to an OS Firewall

160
Copyright 2017 Audinate Pty Ltd. All rights reserved
NETWORK PORTS – 2 WAYS TO CUT THE WIRE

• The Network Switch Fabric can have ACLs applied


(Access Control Lists)
- ACLs are very powerful tools for filtering traffic in the network
- Many advanced IT Networks will be applying ACLs
- Normally IT departments will not wish to reveal or discuss
these (potential security concern)

• It is reasonable to expect that the required service be


provisioned with the required resources – at a Port
level this is defined in previous slides

161
Copyright 2017 Audinate Pty Ltd. All rights reserved
IP ADDRESSES
• 2 kinds of IP addresses to be concerned about:
– In Dante, Multicast IP addresses only work in a single IP subnet
– All communication on a Dante network spanning multiple subnets is unicast (device
IP to device IP)

• Multicast IP addresses
– Some switches block IP multicast by default (this will prevent the following
functioning properly):
◦ Dante Discovery in the local area network
◦ Dante Clock (PTP is sent multicast as standard)
◦ Multicast Dante Audio

162
Copyright 2017 Audinate Pty Ltd. All rights reserved
IP ADDRESSES
• 2 kinds of IP addresses to be concerned about:
– In Dante, Multicast IP addresses only work in a single IP subnet
– All communication on a Dante network spanning multiple subnets is unicast
(device IP to device IP)
• Unicast IP addresses
– For a unicast communication to be successful:
◦ In a LAN – both devices must be within the same IP subnet
◦ In a Routed network – the LANs containing the devices to be connected must be
joined together by an IP Router
◦ The IP subnet configuration of the Router must be correct (Interfaces correctly
addressed)
◦ Routes must exist between the IP subnets in question
◦ More advanced routers can filter traffic using ACLs just like a switch or a firewall

163
Copyright 2017 Audinate Pty Ltd. All rights reserved
MAC ADDRESSES
• MAC addresses rarely cause any issues (unless an ACL or security appliance
is configured with a deny rule)
• At Layer 2 – ensuring all interfaces on a switch are in the expected VLAN is
the most common configuration issue
• And Finally, Layer 1
- Is the wire physically cut?
- Is the interface open? (managed switches can have their interfaces shutdown using
a software command)
- If using fiber – is the signal integrity good (fiber ends cleaned? Interface dust free?)
- Use respectable test equipment to certify cables (cat 5/6/7 etc) and check signal
integrity and loss of a fiber assembly

164
Copyright 2017 Audinate Pty Ltd. All rights reserved
COMMON SYMPTOMS AND CAUSES

Multiple Master Clocks shown Devices “cannot hear” multicast sync messages from
other devices – assume they are master (caused by
blocked multicast traffic in network)

Clock gives unlock/lock warning Possible Multicast block – check the clock histogram
in Dante controller to confirm (Dante clocks can run
for a surprising amount of time before falling out of
sync badly enough to affect audio… hours)

Device Name shows in Dante Check 1. Firewall settings on your computer’s


Operating System
Controller – no + visible to expand 2. That the Interface being used by Dante controller
channels & status view missing on your computer is in the same IP subnet as the
Data Dante devices that you wish to control

165
Copyright 2017 Audinate Pty Ltd. All rights reserved
COMMON SYMPTOMS AND CAUSES

Device Name • Check Firewall settings on your


shows in Dante computer’s Operating System
Controller - no + • Check that the interface being
used by Dante controller on
visible to expand your computer is in the same IP
channels & status subnet as the Dante devices
that you wish to control
view missing Data

166
Copyright 2017 Audinate Pty Ltd. All rights reserved
COMMON SYMPTOMS AND CAUSES

• Possible Multicast block – check


the clock histogram in Dante
controller to confirm
Clock gives • Unicast Delay Requests can be a
unlock/lock quick tool to test if this is the case
warning • Dante clocks can run for a
surprising amount of time before
falling out of sync badly enough to
affect audio… hours

167
Copyright 2017 Audinate Pty Ltd. All rights reserved
COMMON SYMPTOMS AND CAUSES

• Devices “cannot hear” multicast


Multiple Master sync messages from other devices
– assume they are master
Clocks shown • Normally caused by blocked
multicast traffic in network

168
Copyright 2017 Audinate Pty Ltd. All rights reserved
TALKING TO AN IT DEPARTMENT
• This course covers the requirements of a Dante Audio over IP
network in some detail
• Configuration details for a particular IT infrastructure vendor’s
equipment are specific and unique
• HOWEVER – the concepts covered in this course are universal in
IP networking (the terminology may be subtly different on different
infrastructure vendors)
– Understanding how IP connections are made (ports, IP addresses,
Routing, Switching, Multicast, Broadcast etc.) helps explain
requirements
– Understanding bandwidth requirements is useful
– Understanding that QoS rules can be applied to optimize performance
– Understanding that IGMP can be used to mitigate against the negative
perception of multicast

169
Copyright 2017 Audinate Pty Ltd. All rights reserved
TALKING TO AN IT DEPARTMENT

• Networking is about making a lot of parts of an


unique “jigsaw” work together
• This course explains the requirements and
performance of the “Dante piece” of the jigsaw
• Remember – an IT department have to make many
pieces fit together – remain patient, it can and will be
made to fit together nicely.

170
Copyright 2017 Audinate Pty Ltd. All rights reserved
NEXT
STEPS

171
TAKE THE LEVEL 3 TEST

http://www.audinate.com/certify

• Create an Audinate account if you don’t


have one
• Login to your account
• Take Level 3 test
• Certificate is automatically generated

172
Copyright 2017 Audinate Pty Ltd. All rights reserved
THANK
YOU

173

You might also like