Agenda
Why do we care?
The way we do it
How long and how much?
Traffic flows – P2P vs. Conferencing
Endpoint placement
Simulation flows
What are we looking for and why?
Traffic distribution diagram
2
No jitter
No packet loss
3
Need to have enough data for analysis since things may be happening,
but not that often
Need to see if there are any patterns
Can stress the network! (but that is the idea)
Should be discussed with customer
4
Generate user datagram protocol (UDP) flows
Flows are point-to-point
The traffic is similar to real-time transport protocol (RTP) traffic with payload of a certain codec (typically G.711)
Expressed in Kbps
Cannot always specify raw Kbps
Some tools use specific codecs (such as G.711) for simulation
Conversion process can be used (Kbps -> NxG.711)
Some allow also for TCP background traffic
Can be used to simulate Lync non-RTC traffic
(e.g. SIP + IM/P, App/DesktopSharing)
Often generate variable flow (waves of traffic)
Helps the network to adapt
More real traffic than ”constant heat”
5
Where Lync users might be connected
They do not have to be on the users desk, in the rack is enough
Where Lync servers might be connected
Customer might not know where they want to place it yet
Security policies might be hard to overcome
6
Somewhere close to LAN distribution
or LAN core is good enough
Firewalls and other security or network optimization devices are
common in DC LANs
They might be have impact on Lync traffic in DC
So you want to know that and test that !
7
Simulated directly
Simulated as multiple P2P flows to central site(s)
Determine if simulation tool can simulate video traffic
Alternative can be to use audio simulation as video
8
These values are the main indicators for call quality from
a networking perspective
Packet loss:
Percentage of packets that get dropped along the way for any reason, such as congestion in the WAN,
or failing links
High average packet loss might mean overloaded (lossy) network
Sporadic high max packet loss (spikes) might indicate failing links
If these are high, then:
Lync client might turn forward error correction (FEC) on, consuming more bandwidth
Voice quality might be impacted due to codec switch
From wide band to narrow band, or from G.722 to Siren
9
Eventually, either user or Lync client might decide to drop call
delay
Time it takes for a packet to travel from endpoint A to endpoint B
High average delay will simply indicate geographical distance
High average delay might also indicate network congestion
High max delay (spikes) indicates network congestion
Can result in jitter and late buffer drops (think of it as a domino effect)
If these are high, then:
Lync users might experience a ”walkie-talkie” effect
Eventually, might lead to jitter and voice quality can be impacted
10
Jitter
Jitter is the variation of latency
Jitter is amount of time we need to set as a buffer to build a steady stream of packets for continuous
play-out
High jitter typically indicates improper or lack of QoS
High average jitter can result in extra delay in the conversation
Buffer has to be increased
A feedback loop to back delay – might lead to ”walkie-talkie” effect
High max jitter (spikes) might lead to late buffer drops
Results are similar to packet drops in the network
11
http://en.wikipedia.org/wiki/Mean_opinion_score
MOS
But they usually use Narrowband MOS
May use it only as guideline for you, but do not present is as result of the network assessment
12
?
13
14
Standard network assessment is scoped for up to 20 sites
Need to make sure that we test in representative sites
Where modeling phase reveals that the link bandwidth might be barely enough
Connections that will travel between continents are prone to packet loss and delay
Connections of expected high utilization (users talk often)
There might be existing knowledge of WAN links with a bad quality
Example: ”That’s the site where my CEO is”
15
Bandwidth from and to Site X
Flows from Site X to Site Y
17
1.
Start with P2P allocation first
Don’t decide completely on your own, discuss with the customer
”Do users from Site A talk to users from Site B often?”
“Does their business hours overlap?”
Go through sites one by one, starting from smallest and moving to largest
1.
The PSTN flows will go from Branch X to its central site – easy
2.
Conferencing flow – also from Branch X to its central site – easy
NOTE: in a network assessment, for multiple central sites, you might want to distribute flows 2 & 3 in between
many central sites
18
Peer Traffic Distribution – Conceptually
Peer traffic will be distributed based on
geography
In a global deployment, we split the
globe into three time zones, EMEA,
Americas, Asia. For example, if a probe is
placed in EMEA, then its 3 peer locations
will be 1 x EMEA, 1 x Americas and 1 x
Asia
This ensures we have a true globally
distributed peer setup
19
Conference and PSTN Traffic Flows
Conferencing Traffic
This scenario will be
using the traditional
Hub and Spoke mesh
test
There will be one Hub
that we will refer to as
the datacenter
There will be up to a
maximum of 20 Spokes
that will be referred to
as a site, probe, or
remote location
Each of the site probes
will generate the
following test traffic
back to the datacenter
probe:
• One audio stream
bi-directional
• One video stream
bi-directional
20
PSTN Traffic
This scenario will be
using the traditional Hub
and Spoke mesh test
There will be one Hub
that we will refer to as
the datacenter.
There will be up to a
maximum of 20 Spokes
that will be referred to as
a site, probe, or remote
location.
Each of the site probes
will generate the
following test traffic back
to the datacenter probe:
• One audio stream
bi-directional
21
22
23
24
Site Name
Conf
P2P Audio Audio
Kbps
Kbps
PSTN
Audio
Kbps
Data Center
79.60
1402.50 5562.00
Headquarters
Regional Office
Branch office
557.20
437.80
0.00
688.50
561.00
153.00
25
2781.00
2224.80
556.20
119 Kbps
79.60
Site Name
Data Center
Headquarters
Regional Office
Branch office
26
P2P Audio Conf Audio
PSTN
Audio
Kbps
Kbps
Kbps
0.00
1402.50
5562.00
557.20
437.80
0.00
688.50
561.00
153.00
2781.00
2224.80
556.20
119
79.60
Kbps
Site Name
Conf
P2P Audio Audio
Kbps
Kbps
PSTN
Audio
Data Center
0.00
1402.50 5562.00
Headquarters
Regional Office
Branch office
119.4
0.00
0.00
688.50
561.00
153.00
27
437.80
Kbps
Kbps
2781.00
2224.80
556.20
119
Kbps
79.80
2781
Kbps
Site Name
P2P Audio Conf Audio PSTN Audio
Kbps
Data Center
Kbps
Kbps
0.00
1402.50
5562.00
119.40
688.50
2781.00
Regional Office
0.00
561.00
2224.80
Branch office
0.00
153.00
556.20
Headquarters
28
556.20
2224.80
Kbps
119
473.80
688.50 Kbps
79.80
2781
Kbps
Site Name
Data Center
PSTN
Audio
Kbps
Kbps
Kbps
0.00
1402.50
5562.00
119.40
688.50
2781.00
Regional Office
0.00
561.00
2224.80
Branch office
0.00
153.00
556.20
Headquarters
29
P2P Audio
Conf
Audio
561
Kbps
153 Kbps
556.20
2224.80
Kbps
119
473.80
Done!
Site Name
Data Center
PSTN
Audio
Kbps
Kbps
Kbps
0.00
1402.50
5562.00
119.40
688.50
2781.00
Regional Office
0.00
561.00
2224.80
Branch office
0.00
153.00
556.20
Headquarters
30
P2P Audio
Conf
Audio
2781
Kbps
561
Kbps
688.50 Kbps
79.80
153 Kbps
556.20
2224.80
Kbps
119
473.80
http://www.microsoft.com/download/en/details.aspx?id=19011
34
© 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the
part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.