Sei sulla pagina 1di 9

ATM over Ethernet (PWE) vs Iub/IP positioning and marketing material

TRS PLM Core team meeting Juan Ignacio Solana de Quesada, 22.10.08

For internal use 1 Nokia Siemens Networks

ATM over Ethernet (PWE) vs Iub/IP (I)


Transport Overhead efficiency
ATM and ATM over Ethernet are more efficient (less transport overhead)
than IP based Iub when the traffic mix is predominantly voice. IP based Iub is more efficient than ATM and ATM over Ethernet when the traffic mix is predominantly data. The key factor to predict the transport overhead is average packet size. The even point of efficiency between ATM over Ethernet and IP based Iub is
the average packet size of 200 bytes. If the amount of data calls is less than 30% of the total number of calls, ATM is more efficient. Otherwise, IP based Iub is more efficient.

For internal use 2 Nokia Siemens Networks

ToP results / DLi / 23.07.08

ATM over Ethernet (PWE) vs Iub/IP (II)


First Mile Bandwidth considerations
The physical capacity of the first mile on the BTS connection plays an
important role realted with handling of QoS. In IP based Iub, handling of big low priority IP packets causes a significant
Packet Delay Variation to small high priority IP packets when the physical capacity of the first mile is small. (6Mbs or smaller)

ATM over Ethernet is more suited to handle QoS than IP based Iub when
the physical capacity of the first mile connection to the BTS is 6Mbs or smaller.

A simplified model to determine the amount of Transport overhead as a


function of the traffic mixed. Insert in column B the percentual distribution of the number of AMR, NRT PS and HSPA calls.

For internal use 3 Nokia Siemens Networks

ToP results / DLi / 23.07.08

Back Material

For internal use 4 Nokia Siemens Networks

ATM over Ethernet vs. native Iub/IP


Case 1: short frames (AMR) Iub/ATM
Frame 1 Frame 2 Frame 3

Frame 1

Frame 2

Frame 3

Iub/ATM/Ethernet (PWE)
Iub/IP
IP UDP

IP

MPLS CW

Frame 1

Frame 2

Frame 3

Frame 1

IP

UDP

Frame 2

IP

UDP

Frame 3

Case 2: longer frames

Frame 1

Iub/ATM
Iub/ATM/Ethernet (PWE) Iub/IP
IP
MPLS CW

Frame 1, part 1

Frame 1, part 2

Frame 1, part 3

Frame 1, part 1

Frame 1, part 2

Frame 1, part 3

IP

UDP

Frame 1

For internal use 5 Nokia Siemens Networks

ToP results / DLi / 23.07.08

Overhead Comparison - Details


Iub/ATM emulated over a packet-switched network
Encapsulation: Iub/ATM/IP/Ethernet

Ethernet header IP header MPLS Pseudowire Control Word ATM header (shrinked) AAL2 Start field AAL2 CPS header 18 byte 20 byte PW/IP/ETH = 46 byte fixed overhead 4 byte 4 byte 4 byte (N-to-one mode, per 48 byte AAL2 CPS PDU) 1 byte (per ATM cell) 3 byte (per 45 byte large SSCS-SDU)

ATM/AAL2 = 8 byte

46 bytes fixed overhead per packet + ~8 bytes per 45 bytes FP payload


5 * AMR5.9: 5 * 22 FP byte in 3 ATM cells > 5 * AMR12.2: 5 * 39 FP byte in 5 ATM cells > 5 * PS384: 5 * 510 FP byte in 57 ATM cells > 2 *1506 + 1 * 102 = 5 * HSDPA: 5 * 1472 FP byte in 164 ATM cells > 5 * 1506 + 1 * 1298 = IP MPLS CW Frame 1 Frame 2 202 byte (in 1 IP packet) 306 byte (in 1 IP packet) 3114 byte (in 3 IP packets) 8828 byte (in 6 IP packets) Frame 3

Iub/IP
Encapsulation: Iub/UDP/IP/Ethernet

Ethernet header IP header UDP header 5 * AMR5.9: 5 * 22 FP byte 5 * AMR12.2: 5 * 39 FP byte 5 * PS384: 5 * 510 FP byte 5 * HSDPA: 5 * 1472 FP byte IP UDP 18 byte 20 byte 8 byte > 340 byte (in 5 IP packets) > 425 byte (in 5 IP packets) > 2780 byte (in 5 IP packets) > 7600 byte (in 5 IP packets) Frame 1
ToP results / DLi / 23.07.08

Please note: In order to show the impact of multiplexing on AAL2 layer, we assume always 5 parallel calls. Max number of ATM cells per IP packet is 28

46 bytes overhead

For internal use 6 Nokia Siemens Networks

Overhead comparison (I)


The amount of overhead introduced for Transport is quite different for ATM over Ethernet
(PWE) and Iub/IP The ATM AAL-2 offers an almost constant overhead ratio of 18% independently of the packet length.
This is so because the FP packets are chopped to AAL-2 CID packets of constant length (44 bytes). The Iub/IP mode has a constant overhead per packet. The overhead ratio decreases inversely to the packet length

The crossing point is for packets lengths of about 200 bytes on Frame Protocol Layer.
Overhead from FP to Ethernet as a function of FP frame length
45,00% 40,00% 35,00% 30,00% 25,00% 20,00% 15,00% 10,00% 5,00% 0,00% 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 Native Iub/IP

ATM over ethernet (PWE)


For internal use 7 Nokia Siemens Networks

ToP results / DLi / 23.07.08

Overhead comparison (II)


The almost constant overhead ratio of ATM AAL-2 mode allows a very important
QoS characteristic of ATM. Big FP packets can be interleaved according to QoS classification. A high priority small FP packet (i.e. belonging to a Voice call) does not need to wait for
the completion of the transmission of an ongoing big FP packet (i.e. belonging to a HSPA data call).

Packet interleaving is not possible on Iub/IP mode.


Such an effect is typically called the serializing effect.
The serializing effect can lead to significant Packet Delay Variation. The serializing PDV can be a problem when the BTS last mile technology allows only a
limited bandwidth.

For internal use 8 Nokia Siemens Networks

ToP results / DLi / 23.07.08

Serializing PDV depending on the first mile BTS capacity


Native Iub/IP: Max PDV caused by Packet Length (ms)
as function of Limited Physical Bandwidth (Mbs) 2,50 2,02 ms 2,00

1,50

6 Mbs 9 Mbs

1,35 ms 1,01 ms 0,81 ms 15 Mbs

1,00 12 Mbs 0,50

0,00 0 200 400 600 800 1000 1200 1400 FP Frame length (bytes)

In combination with DSL technologies on the last mile BTS bakhaul, the max PDV
caused by serializing effect on the Iub/IP mode is in the range of 1 to 2 ms. In the ATM over Ethernet mode, the PDV is kept always as small as 50us. This is due to the CID packet interleaving provided at AAL-2.

For internal use 9 Nokia Siemens Networks

ToP results / DLi / 23.07.08

Potrebbero piacerti anche