Académique Documents
Professionnel Documents
Culture Documents
CM 4.0
150 TARTS IP endpoints registered per region, where each of the two PNs had its own
isolated region.
The originating TARTS IP endpoints were used to generate traffic over two ISDN PRI
trunks that then terminated to a second set of 150 TARTS IP endpoints on the second PN.
This provided our outbound and inbound simulated PSTN trunk traffic respectively
– no inter-port network VoIP traffic, all bearer traffic traversed via the ISDN PRI trunks
– Poisson Distribution for random call patterns
Call durations were 10 seconds
BHCC call rates were 1K, 2.5K, 5K, 7.5K, and 10K
Bandwidth was limited and delay introduced by a PacketStorm WAN Emulator
Up-to 300 msec delay, 0.1% Packet Loss for IPSI traffic and an IPSI bandwidth limitation
of 128Kbps were introduced.
An Radcom and an Ethereal Packet Sniffer was used to collect and analyze TCP/IP
Packet data.
Kbps
40
20
0
1K 2.5K 5K 7.5K 10K
BHCC
BHCC Usage Per Station Average IPSI Bandwidth Average IPSI TCP/IP
Per PN (Kbps) full duplex packets per second
5.9 Kbps for L2 overhead + 52.2 Kbps for TCP/IP CCMS Packets equals a
minimum average bandwidth requirement of 58.1 Kbps.
For example; for 5K busy hour calls using encrypted PPP links to control remote
port networks, as described in the previous example, you would guarantee
128Kbps (69.3Kbps + 64Kbps) for IPSI signaling bandwidth across the WAN link.
BHCC Ethernet PPP MLPPP Frame Relay
1K 64Kbps 64Kbps 64Kbps 64Kbps
1K w/ encryption 64Kbps 64Kbps 64Kbps 64Kbps
2.5K 128Kbps 128Kbps 128Kbps 128Kbps
2.5K w/ encryption 128Kbps 128Kbps 128Kbps 128Kbps
5K 128Kbps 128Kbps 128Kbps 128Kbps
5K w/ encryption 128Kbps 128Kbps 128Kbps 128Kbps
>=7.5K 192Kbps 192Kbps 192Kbps 192Kbps
>=7.5K w/ encryption 192Kbps 192Kbps 192Kbps 192Kbps
• In the above figure, the green line represents traffic from the standby IPSI
which in turn overlaps the blue line with traffic from the standby server. As
you can see there is very little traffic.
• 2.4 Kbps bandwidth is consumed by the standby IPSI.
© 2005 Avaya Inc. All rights reserved. 13
Where does IPSI Signaling Fit?
8 Class QoS Model Voice
Higher Priority Traffic can starve Video
equal and lower priority traffic
– Voice Bearer Traffic and IPSI
Signaling Traffic currently share H323 Call Signaling
the same queue and compete
for bandwidth within this queue Network Management
– Voice Bearer cannot Tolerate
Delay or Jitter Critical Data
– IPSI can tolerate up to 300
msec Delay
Bulk Data
– Voice Quality tolerates 3%
Packet Loss
Best Effort
– A Port Network will become
unstable with greater than 3%**
Packet Loss across the link Scavenger
P P P D T R M R IP Precedence = 101 = 5
Precedence
7 - 111 Network Ctl DSCP uses an additional 3 bits of the ToS Byte to
6 - 110 Internetwork Ctl
5 - 101 Critical differentiate traffic into 64 classes
4 - 100 Flash Override
3 - 011 Flash Delay Throughput Reliability Monetary Reserved
2 - 010 Immediate
1 - 001 Priority Cost
0 - 000 Routine 0 Normal 0 Normal 0 Normal 0 Normal *ALWAYS* set
1 Minimize 1 Maximize 1 Maximize 1 Minimize to zero
Voice Voice EF 46
class-map VoIP-Bearer
match ip dscp 46
class-map H.323-VoIP-Control Classify the traffic of
match ip dscp af31 interest for QoS Policy
class-map IPSI-Control
match ip dscp af42
! Define QoS Policy
Treatment
policy-map QoS-Policy As Percent or Kbps
class IPSI-Control
bandwidth 128
class H.323-VoIP-Bearer
priority percent 60
class VoIP-Control
bandwidth percent 5
class class-default
random-detect
!
You can change the QoS profile assignment for each of the 64 code points using the following command:
Co-Pilot
– Steve Regini
System Test
– Andy Cornejo
– David James
– Ana Kesselring
– Rob Pospisil
ATAC
– Tim Kaye
– Ken Lin
1% to minimally
differentiate Napster,
KaZaa, Gaming from BE