• cdma2000
1xEV-DO: Tecnologia, Serviços e Mercado
----- Original Message
-----
Sent: Tuesday,
January 17, 2006 4:32
PM
Subject: Re:
FW: [wireless.br] cdma2000 1xEV-DO:
Tecnologia, Serviços e Mercado
Para o Evandro e Arakaki
Arakaki escreveu (sobre o artigo "Voice
over IP" abaixo transcrito)
(...) Parece-me que a Rev.A
ainda será sobre tecnologia CDMA e não garantirá a
redução da latência para 30 ms, necessário para uma
conversação agradável por VoIP. Assim, entendo que
apenas a Rev.B do 1XEVDO
conseguirá isso, quando então, além de oferecer
74Mbps pro usuário final, a Qualcomm migrará para a
tecnologia OFDM... (...)
Com absoluta certeza, isto tudo é CDMA2000 (leia-se
Qualcomm). A rev.A do EV-DO promete um bocado de
coisas com respeito a QoS, que, supostamente,
tornaria mais confortável o uso de aplicações
multimídia
(VoIP incluso). Mas isto ainda precisa ser provado
na prática.
Concordo que, a esta altura, os planos para o EV-DO
subiram no telhado, e uma migração direta para OFDM
é muito mais provável, mas mesmo com os níveis de
latência altos do EV-DO rev.0, e sem garantia de QoS
na interface aérea, aplicações VoIP estilo Skype
podem muito bem funcionar (e até surprender pela
qualidade da voz).
Afirmações do tipo "VoIP só é viável se o atraso
máximo for de X ms, com jitter máximo de Y% e perda
máxima de Z%" são enganosas, porque o grau de
"sobrevivência" da chamada diante de um ambiente de
rede
hostil depende diretamente das características da
aplicação (sempre é bom lembrar: isto é
packet-switching networking, e VoIP é mais uma
aplicação - camada 7 do modelo OSI - a ser
transportada pela rede).
Se considerarmos o jeito que a voz é "empacotada" em
telefones celulares (CDMA ou GSM), minha sensação é
que a rede IP subjacente vai ter que ser
extremamente superdimensionada (mesmo com todas as features
de QoS ativadas), porque esta aplicação, montada em
cima de vocoders que comprimem muito, e com quase
nada de processamento adicional (afinal, vc está
tentando "enfiar" esta aplicação em um handset que,
tipicamente, tem um "cérebro" um tanto limitado),
acaba ficando extremamente sensível a variações
inesperadas nos parâmetros de qualidade da rede (delay,
jitter e perda).
O sub-produto desta forma de abordar as aplicações
multimídia são exigências tamanhas sobre a rede IP
que ela, no final das contas, acabaria sendo apenas
uma rede de emulação de circuitos. Só tem um
detalhe: o protocolo IP não foi desenhado para isso,
e nenhum esquema de QoS vai conseguir reproduzir um
ambiente circuit-switching. E se esquemas no nível 2
vem em mente (virtual wires e quetais),
pense de novo, porque eles só tem alcance até os
limites físicos de implementação de cada sub-rede
nível 2. Para seguir adiante o sinal
vai ter de "subir um degrau" e passar por roteadores
IP, e vc volta ao ponto inicial: circuit-switching
emulation de pé quebrado.
Por outro lado, já vi referências a chamadas usando
o soft-phone Skype com qualidade considerada
aceitável mesmo quando o delay (unidirecional)
excedia 600 ms e a perda de pacotes (perda real ou
perda gerada por jitter excessivo) passava de 50%!!!
Então, minha conclusão é:
- existirão usuários
de VoIP sobre a estrutura atual de EV-DO rev.0? Sim.
- Eles ficarão
satisfeitos com a qualidade das chamadas? Depende do
tipo de aplicação VoIP que eles usarem.
- Isto vai melhorar no
EV-DO rev.A? Provavelmente sim, mas ainda precisamos
ver como as features de QoS na interface
aérea irão coexistir e cooperar com as features
de QoS da rede IP.
BTW, embora não saiba os detalhes, suspeito que o
cenário GSM (aí considerando GPRS, HSDPA e HSUPA) é
mais ou menos parecido.
[ ]'s
Smolka
-------------------------------------------------------------------
Voice over IP
Voice over IP (VoIP)
refers to the set of mechanisms that allow the sending
of voice communications over a packet-based IP (Internet
Protocol) network. 1xEV-DO Rev A introduces new
features including end-to-end application-level Quality
of Service (QoS) to the wireless network that allow
service providers to deliver high quality voice services
across their EV-DO infrastructure.
Drivers for VoIP over EV-DO Rev. A
For service operators, there are three primary drivers
for VoIP: network efficiency, more efficient use of
spectrum and the ability to enhance voice service
portfolios.
Network efficiency
Network efficiencies are achieved by using existing
resources more efficiently and by consolidating multiple
networks, systems, and support organizations. As voice
communications move toward digital techniques, the logic
of merging separate networks for voice calls and for
data sessions is inescapable. Converging the network
onto one common, packet-based infrastructure will
dramatically reducing the number of systems, employees
and facilities required.
Spectrum efficiency
Operating parallel voice and data networks is an
inefficient use of spectrum - especially in a CDMA
environment where 1xRTT and EV-DO technologies use 1.25
megahertz of spectrum for each radio channel. Using
separate radio channels for voice and data prohibits
load balancing between the two services, and reduces
overall trunking efficiency. As in the case of wireline
networks, a transition to VoIP enables wireless carriers
to converge all uses, voice and data and other
multimedia, onto the same spectrum, eliminating the need
for redundant channels, increasing trunking efficiency,
and making better overall use of the spectrum.
Advanced Service Creation
VoIP enables the rapid creation of advanced services.
Once carriers have a VoIP network in place they own a
rich service creation environment in which new services
can be created much more easily than they can be in the
MSC-centric architecture of legacy cellular networks.
New services beyond the traditional telephone services
(call-waiting, multi-party calling, call forwarding,
etc.) are possible when voice and data communications
travel over the same network. Examples include
push-to-talk, video telephony, click-to-dial,
whiteboarding, and advanced accessibility features.
1xEV-DO Rev A and VoIP
1xEV-DO Rev A adds
several advancements for VoIP and other multimedia
traffic.
Increased channel capacity on both the forward and
reverse links
Compared to EV-DO Rev. 0,
EV-DO Rev. A increases the peak forward link data rate
to 3.1 Mbps and the peak reverse link data rate to 1.8
Mbps. The dramatically enhanced uplink data rate will
enable Rev. A networks to support significantly more
voice connections than is possible under Rev. 0.
QoS support over the air link with multiple flows
The EV-DO forward link uses TDM to send packets to
various users, which requires a scheduling function to
decide which user should gain access to the air link at
any given time. EV-DO Rev. A has added the ability for
the forward link scheduler to coordinate the use of the
EV-DO air link with the various devices that will be
using it. Therefore, data devices, which need high data
rates but are insensitive to packet delay and jitter,
can be handled in one way, while voice devices that need
lower data rates but that are highly sensitive to packet
delay and jitter will be handled a different way. On
the reverse link EV-DO Rev. A permits the use of higher
power for QoS packets in order to reduce the number of
transmissions and retries necessary to successfully send
these packets.
Support for short and multi-user packets
Unlike many data applications, voice applications
transmit relatively short packets on a regular basis. In
EV-DO Rev. 0, the rigid structure of the physical layer
leads to inefficient transmission of these short packets.
EV-DO Rev. A has addressed this problem in two ways.
Unlike Rev. 0, which allows packets sent over the
traffic channel to be a minimum of 128 bytes long, Rev.
A supports a wide variety of shorter packets, with
physical layer packet lengths as low as 16 bytes. These
packets can be transmitted in less time, and allow more
users to access the network with low latency. In
addition, EV-DO Rev. A includes support for multi-user
packets. This capability allows a long physical layer
packet (with its associated overhead) to be addressed to
separate users, again reducing air link overhead as well
as per-user delay.
Support for header compression
In current networks the overhead (non-voice information)
required to direct traffic from source to destination is
typically 40 bytes and results in an inefficient
overhead to payload ratio of 2:1. For example, 20 msec
of speech from an 8 kbps voice coder is 160 bits, or 20
bytes, while the associated IP, UDP, and RTP headers
used for VoIP routing and control are 40 bytes. Such
inefficiency affects the capacity of the network to
handle voice traffic. EV-DO Rev. A supports compression
of these headers, from 40 bytes down to approximately 2
bytes, thus enabling high VoIP capacity.
Enhanced Idle State
Protocol support for faster, variable, paging
While a delay of several seconds in connecting to a
server is not noticeable with most data applications,
real-time applications such as voice are quite different.
Paging channels in digital cellular networks, and EV-DO
networks in particular, are logical control channels
that allow the devices to receive control messages.
Slow paging channels would manifest themselves as slow
setup times, slow reaction to dialed digits, etc. during
a voice call. Rev. A has addressed this issue by
creating a variable, faster paging cycle.
Voice over IP is the natural progression for mobile
networks as operators begin to expand their deployments
of packet data systems. Airvana is embracing EV-DO Rev.
A technology to deliver an end-to-end application-level
QoS to the wireless network, enabling service providers
to deliver high-quality voice services alongside data
and multimedia applications. Airvana is the first
company to demonstrate VoIP over EV-DO with QoS.