19 - SEII-c Backhauling Implementation
19 - SEII-c Backhauling Implementation
This document contains information proprietary to Gilat Satellite Networks Ltd. and may not be reproduced in
whole or in part without the express written consent of Gilat Satellite Networks Ltd. The disclosure by Gilat
Satellite Networks Ltd. of information contained herein does not constitute any license or authorization to use or
disclose the information, ideas or concepts presented. The contents of this document are subject to change
without prior notice.
AGENDA
Introduction
Operation modes
Backhauling concepts
o The Connection
o Connections Multiplexing
o Streams
o BW allocation algorithm
o Header compression
o Packet aggregation
o Delay control
o QoS
INTRODUCTION
The SkyEdge II-c design takes into account these requirements and provides a dedicated solution
C2P CONNECTIONS
Overview
Connection i
VSAT
Connection j
BackBone
traffic
8 | Proprietary and Confidential
8
Virtual circuit – the Backhauling traffic is isolated from other traffic types, it receives dedicated allocations, it
manages its own QoS.
C2P CONNECTIONS
Main Principles
C2P (Connection Control Protocol) is based DVB-RCS standard, used to establish connections
C2P CONNECTIONS
Parameters
A C2P allocation request receives the highest priority from the HSP
HSP allocates a low jitter allocation pattern for C2P, if the request can’t be met it is rejected
C2P CONNECTIONS
Connection Establishment
The connection is established by the VSAT which is the master of the connection
C2P More
VSAT 1st Packet
connection packets
Logon arrives
established arrive
C2P More
VSAT 1st Packet
connection packets
Logon arrives
established arrive
C2P CONNECTIONS
Connection Teardown
The timer may be configured for infinity, in which the connection will never be torn down, unless the VSAT
loses Backbone connectivity
Timer Expired
Connection down
Packet Packet Packet Packet
t1 t1 t1 Inactivity timer
Example:
Consider a network with redundant servers. eNodeB attempts connection to server #1 which is down. The
packets sent create connection #1. eNoteB then tries connection #2, and succeeds. We now have 2 active
connections, one carries no traffic but may consume BW (If min BW is not zero) and C2P resources.
C2P CONNECTIONS
Connection Properties
C2P MULTIPLEXING
C2P MULTIPLEXING
Introduction
As cellular network architecture is evolving it may require opening of high number of C2P connections
o Each connection requires reservation and allocation of resources that not always fully utilized without the
option of sharing the unused BW between single connections
As cellular network architecture is evolving it may require opening of high number of C2P connections. Each
connection requires reservation and allocation of resources that not always fully utilized without the option of
sharing the unused BW between single connections.
C2P MULTIPLEXING
Example
IP=1.1.1.12.2.2.2
UDP, VLAN=3023
Min=32
Extra=16
IP=1.1.1.13.3.3.3
UDP, VLAN=3023
Min=32
Extra=16 Meta
Min=32
IP=1.1.1.12.2.2.2 Extra=16
SCTP, VLAN=3023
Min=32
Extra=16
IP=1.1.1.12.2.2.2
SCTP, VLAN=3024
Min=32
Extra=16
In this example (on left), we have 4 C2P connections, each consumes 32[kbps] even before the first Byte is sent,
so the requested initial BW is 128[kbps].
In case all connections exceed the minimum BW, they ask for extra 16[kbps] each. This amounts to 64[kbps].
That are usually not used and are reserved to reduce the latency of the next increase in BW rate.
Some connections (e.g. those carrying SCTP) require BW far more than the actual throughput in order to
minimize latency.
Using Meta connection (on right), we keep the same Min, Extra as we had for each of the 4 connections and
apply it to a single connection. Since the assumption is that the extra BW is used to compensate for the next
call/throughput increase, and since not all of them will have an increase at the same time, the same Extra may
be used. The same logics goes for the minimum BW.
For SCTP signalling, where packets arrive in relatively large time intervals, multiplexing makes response time
better using the same values of overhead because the aggregated rate allows the VSAT to more accurately
measure the throughput and ask for appropriate resources.
C2P MULTIPLEXING
Introduction
The connection that multiplexes a number of single connections called Meta connection
o Connections from different IP address pairs, VLANs and protocols may be multiplexed into a single Meta
connection
o All the streams of the “child” connections become streams of their Meat connection
o The Meta connection manages the BW (RTN allocations) and common parameters to all of its child
connections
o The rate estimation algorithm applies to the entire Meta connection
o The connection establishment process is identical to the existing method
– By Login or By Traffic
STREAMS
BACKHAULING STREAMS
Overview
Streams are established between the VSAT and the DPS, by the VSAT who is the master of the stream
BACKHAULING STREAMS
Diagram
Default Stream
Stream i
Stream k
Traffic complying
Control Stream with C2P
definitions (To C2P)
BACKHAULING STREAMS
Features
BACKHAULING STREAMS
Establishing
During the establishment process, streams are sent over the default stream, without header compression, in
low priority
Packets matching the connection filter, but not belonging to any configured stream may be treated, based on
connection-level configuration:
o As non-Backhauling packets, sent to BackBone
– No priority, no header compression, doesn’t consume connection resources
o In the default stream
– Low priority, no UDP header compression
BACKHAULING STREAMS
Each packet matching the stream filter will create a stream, if it is not yet established
o In some cases, this could lead to a very large number of streams, exceeding the available number of streams
o It is possible to configure a “trunk” of streams sharing the same ToS
o A Trunk consumes a single stream, but doesn’t allow UDP header compression
BANDWIGTH ALLOCATION
BW ALLOCATION ALGORITHM
Parameters
The combination of these parameters ensures the availability of resources ahead of time
Each parameter is configurable allowing the user to decide on the tradeoff between user experience and BW
efficiency
BW ALLOCATION ALGORITHM
Minimum BW
Min BW
o Requested by the connection upon its establishment
o Connection never requests less than minimum BW
o When a packet arrives to an idle VSAT, the fact that the BW is already available reduces the response time
o When set to zero, if the measured BW is also zero, zero BW is requested
BW ALLOCATION ALGORITHM
Extra BW
Extra BW
o The extra BW bridges the gap between the detection of rate increase and the arrival of allocations (~ RTT)
o Assuming a given traffic pattern, in the next millisecond it may increase (e.g. one more call established). If no
extra BW is available, the packets will be delayed by at least RTT until the new allocations are available
o Without extra BW, jitter, delay and packet loss may occur
o When set to zero, no extra BW is requested leading to increased efficiency at the cost of voice quality
BW ALLOCATION ALGORITHM
Maximum BW
Maximum BW – prevents the VSAT from asking too much BW (e.g. asking for Extra when it is known the cell is
already at its maximum capacity)
o Example – a cell is limited to 5 Mbps
o The current measured BW is 5 Mbps, and the extra BW is 1 Mbps. Without the Max BW parameter, the
connection would ask for 6 [Mbps]
BW ALLOCATION ALGORITHM
BW Decrease Guard Time
BW ALLOCATION ALGORITHM
Operation
Measurement starts when C2P connection is established. Until channel is established, packets are sent over
Backbone. When C2P is established, and measurement is large, but BW not yet granted, packets are queued,
subject to delay control.
BW ALLOCATION ALGORITHM
Operation Diagram
In this diagram we can follow the VSAT BW requests (red line) above the measured BW (blue line). The
requested BW is equal to Measured BW + Extra BW. The BW requests are always in limited by the Min and Max
values. When the measured BW exceeds the Maximum, VSAT will request for Max BW. From the moment that
the measured traffic will start decreasing, VSAT will continue requesting for the same BW until the decrease
guard timer is expired. After that VSAT will modify the request to a new measured + extra BW.
BW ALLOCATION ALGORITHM
BW Cases & VSAT Behavior
Measured BW < previously measured BW for a time < guard time Requested BW remains the same
Measured BW < previously measured BW for a time > guard time Requested BW reduced to measured BW +
extra BW (limited by min and Mx)
HEADER COMPRESSION
HEADER COMPRESSION
Overview
Per-stream configuration enables no compression, IP level only compression and UDP+IP header compression
The internal header on the OB contains the connection index in addition to the information contained in the IB
header
The IB connection ID is carried in the ATM-level VPI/VCI parameters without any additional cost
Packets are associated with a stream in order to enable header decompression on the receiving side of the
satellite link
HEADER COMPRESSION
IB Compression
UDP header
IP header
20[B]
8[B] Payload
Internal
header Payload
4[B]
Internal
header Payload
4[B]
HEADER COMPRESSION
OB Compression
Internal
header Payload
5[B]
IP packet (OB)
IP header
20[B]
Payload
Internal
header Payload
5[B]
HEADER COMPRESSION
Compression Example
Note that in RTP mode we have a smaller header of 3 bytes, and the savings climbs to 50%
PACKET AGGREGATION
PACKET AGGREGATION
RTN Carrier
In the RNT carrier DVB-RCS ATM mode is used, each burst (TRF) contains 2 or 4 ATM cells
Packets are being collected so as to fill the burst, as much as possible, whether it is a 2-ATM or a 4-ATM burst
Best for small packets, as it shares burst overhead between multiple packets
Increased efficiency
1 2 n
Burst
39 | Proprietary and Confidential
39
PACKET AGGREGATION
FWD Carrier
The aggregation is configured per FWD Priority – High, Medium and Low
o For each CPE, packets of three FWD priorities may be aggregated and share one IP Neto header
FWD BH packets aggregation is configured per BH FWD priority. For each priority the aggregation may be
enabled or disabled. In addition for each priority it’s possible to configure:
DELAY CONTROL
DELAY CONTROL
Background
It is possible that a traffic surge or resource unavailability will cause packets to stay for long duration in the
queues
DELAY CONTROL
Main Principles
Once transmission opportunity occurs, the current time is compared with the packet arrival time
If extra BW is large enough, packets should not be dropped due to delay Packet stays in line
too long. Drop!
The maximum delay is a configurable property of the connection
QOS
QOS
Allocations
C2P priority is the highest in the network, compared with all levels of BackBone priority
Traffic from any other connection or from backhauling on that VSAT can’t interfere with the established
connection traffic
QOS
Main Principles
QOS
BW Management Options
Various Backhauling BW Management mechanisms are applicable for RTN and FWD carriers on several levels
o C2P level
– Each connection may only request up to its Max BW parameter
– Applicable for RTN allocations
o CPE Level
– The total requests in the CPE are limited by the CPE MIR
– Implemented in SLA profile
– Applicable for RTN allocation and FWD traffic
o MG Level
– The total allocations for all the CPEs in a MG are limited by the MG MIR
– Implemented in SLA profile
– Applicable for RTN allocation and FWD traffic
QOS
BW Management Cont.
Various Backhauling BW Management mechanisms are applicable for RTN and FWD carriers on several levels
o NS level
– The total FWD Backhauling traffic of the whole NS is limited to this value
– Applicable only for FWD traffic
o Priorities
– Applicable for RTN and FWD, different implementation
QOS
C2P BW Allocations
QOS
BW Management - CPE
The total requests of the CPE are limited by the CPE MIR
The MIR is enforced for the RTN BW allocation and for FWD traffic
CPE may have a separated MIR for a Backhauling traffic according to configuration in the SLA profile
o Backhauling MIR may be included or excluded from the VSAT MIR
– Included is only applicable for the RTN direction
There are 2 models for Backhauling MIR to choose from per direction:
“Included” – There is one MIR for the VSAT. The Backhauling traffic is limited by the lower between the VSAT
MIR and Backhauling MIR. Setting the Backhauling MIR lower than the VSAT MIR ensures BackBone traffic not
completely oppressed by Backhauling. Applicable only for RTN traffic
“Excluded” – the Backhauling connections may not exceed Backhauling MR, all of the other traffic in the VSAT
may not Exceed VSAT MIR. Backhauling requests plus Backbone requests may be in total larger than VSAT
MIR
CBR and BH allocation have the same priority, so first comes, first served.
In case CPE is configured with CBR and the BH MIR mode is “Included”, the CBR allocation
will be reduced proportionally to the BH traffic when BH connection will become active.
In “Excluded” mode CBR is not affected by the BH traffic.
QOS
BW Management - MG
The total BW for all the CPEs in a managed group are limited by the MG MIR
The MIR is enforced for the RTN BW allocation and for FWD traffic
There are 2 models for Backhauling MIR to choose from per direction:
“Included” – There is one MIR for the MG. The Backhauling traffic is limited by the lower between the MG MIR
and MG Backhauling MIR. Applicable only for RTN traffic
“Excluded” – the Backhauling connections may not exceed Backhauling MR, all of the other traffic in the VSAT
may not Exceed VSAT MIR. Backhauling requests plus Backbone requests may be in total larger than VSAT
MIR
QOS
BW Management - NS
The total Backhauling BW may be separated from the Backbone BW for the whole NS
o Applicable only for FWD traffic
QOS
RTN Priority Mechanism
When a transmit opportunity arrives, the VSAT will start emptying the queues starting with the highest priority
As long as the higher priority queue is not empty, the lower priority queues are not served
Delay control kicks in ones delay exceeds a threshold, and drops the packets
QOS
FWD Priority Mechanism and MIR
The streams are classified into 3 absolute priority levels, based on configuration sent to DPS upon stream
establishment
o The High priority is sent without delay or jitter, and is not dropped
– Typically used only for signaling
o The Medium priority is sent without delay
– Typically used for voice traffic so should be very jitter sensitive.
– If the MIR value is reached, packets are dropped
o The low priority traffic is allowed a small queueing delay in order to absorb traffic spikes
– Typically used for Backhauling data traffic
– If the medium priority consumed all of the MIR, by the end of a timer, the traffic in the queue is discarded
QOS
FWD Priority Mechanism and MIR
The High priority is sent without delay or jitter, and is not dropped. It is assumed high priority is used mostly for
signalling, hence constitutes a small portion of the network BW. The high priority traffic is deducted from MIR,
leaving a smaller MIR for the remainder of the traffic
The Medium priority is sent without delay. It represents typically voice traffic so should be very jitter sensitive. If
the MIR value is reached, packets are dropped
The low priority traffic is allowed a small queueing delay in order to absorb traffic spikes. If the medium priority
consumed all of the MIR, by the end of a timer, the traffic in the queue is discarded
QOS
FWD Priority Mechanism and MIR
BackBone traffic is a lower priority than the lowest Backhauling priority, so Backhauling and Backbone may
share the same outbound, but Backhauling has priority
MIR enforcement for BackBone and for Backhauling are two completely separate processes (like the
“Excluded” mode in IB)
QOS
FWD Priority Mechanism and MIR
THANK YOU
58