Mobility Management Compromise Proposal
Download
Report
Transcript Mobility Management Compromise Proposal
Why Has Integration Between Cellular and the
Internet Been So Difficult?
James Kempf
[email protected]
October 23, 2007
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
Outline
•
•
•
•
•
Current state of the Internet on Cellular
Business/Economic Problems
Technical Problems
What Can CleanSlate Do?
Summary
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
2
Current State of the Internet on Cellular
What is Available Today?
• Survey of cellular Internet service from 4 large cellular providers
• Match wired service as closely as possible
– Unlimited service as opposed to per month limit on amount of data
– Available on a smartphone handset as opposed to on a PC
card/laptop
• Price, performance, service offering, etc.
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
4
Survey Results
T-Mobile
Service
Type
Phone Internet
only (Web
browsing, Email)
GSM/GPRS
Sprint
Verizon
Phone as modem,
Blackberry email,
Web browsing
Laptop card Web
browsing and
email, no voice
Phone Web browsing
and email, phone as
modem
EV-DO Rev A
EV-DO Rev A
GSM/GPRS
EDGE
Cost
Maximum
Bandwidth
(kbps)
Comment
ATT Mobility
EDGE
$39.99/mo for
unlimited access,
voice extra
$49.99/mo for
unlimited access,
voice is $0.02 per
minute
$59.99/mo for
unlimited access,
no voice
$59.99/mo for
unlimited access US,
$0.015/kb Canada,
$0.0195 kb
international, no
voice
GPRS: 80d/40u
2450d/1800u
(theoretical)
600-1400d
500-800u
(advertised)
GPRS: 80d/40u
No Web service
available on
smartphone
HSDPA available as
laptop card but only
in selected US cities
EDGE: 236d/118u
(theoretical)
National WLAN
hotspot and home
WLAN network
also available
Rev B is 14700d
Also WiMax network
in buildout
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
EDGE: 236d/118u
(theoretical)
7/18/2015
5
Comparison with Wired Service
•
More expensive
– Wired service typically goes for $20-30/mo for about 1000 kbps
•
Bandwidth is lower and typically more asymmetrical
– Wired service has from 800-3000 kbps
– ADSL is also asymmetrical in wired service but cable isn’t
•
Complex collection of plans and service makes choosing difficult
– Some services are available on handsets, some aren’t
– Wired service have basic services like email but its not a big feature
•
Some operators restrict what services you can use or charge extra
– Verizon charges $6/mo for location info
– Who’s going to pay that when you might use it a couple times a month?
•
End user terminals are typically subsidized and service comes with a two
year contract
– But unlocked phones allow unrestricted access to applications
• GMS networks only
– Would you buy a wired service just because the operator subsidizes your PC?
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
6
But if you use it...
• Downloads are slow even though advertised network bandwidth
looks reasonable
– Images in Web pages are especially a problem
• Few Web pages are optimized for mobile
– Small screen, tiny keyboard
– Difficult navigation
• Lack of integration between apps
– Just try planning a vacation on your smartphone!
• Forget about video
– Most operators block it
– Mobile broadcast TV products are currently in test but they’ll only give
you what the operators want you to see
• If you have a locked phone, apps might be limited
– iPhone!
What’s the problem?
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
7
Business/Economic Problems
Costs of Deploying a Cellular Data Network
•
Spectrum acquisition costs
– 700 MHz C block auction (20 Mhz) in January reserve price is $4.5 billion
• $225 million/Mhz!!!
– That’s just the starting price!
•
Radio network build out costs
– About 1/3 of access network deployment cost is in radio equipment
– Google estimated $12 billion to acquire spectrum and build out C block
– Probably less for existing carriers since they have some infrastructure in place
•
Network equipment costs
– Higher than for wired networks due to specialized equipment
– Many companies have IPR on the radio equipment
• Some companies business model consists of charging for their IPR
•
More expensive than for a wired network
– Fiber network deployment costs are of similar scale, but no spectrum
acquisition cost
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
9
Costs of Managing a Cellular Data Network
• Coverage holes lead to expensive maintenance
– Sometimes difficult to find tower sites to plug hole
– Providing last 10% of coverage is 90% of cost
• Large parts of US are low density
– Covering low density areas for high bandwidth is expensive
– In Japan, coverage is more uniform
• Less low density areas in Japan relative to population size
• More technical expertise needed to run a cellular radio access
network
• Constant stream of upgrades makes maintenance expensive
– Cdma2000, EV-DO Rev A, EV-DO Rev B, WiMax, ...
– Operators used to plan for 30 lifetime on network equipment, now
they are lucky to get 5 years
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
10
Irrational “Dumb Pipe” Fear
• “Dump pipe”
– Customer buys just IP service from you
– Other services are bought from other providers
• In wired services, this is the norm
– But wireless operators have an irrational fear of becoming like that
– This fear is fanned by equipment vendors and analysts for their own
reasons
• What the operators don’t realize is that they must become a dumb
pipe if the wireless Internet is ever to take off
– Imagine that you had to pay $2 every time you drove from your home
to school or work
– How often would you drive?
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
11
Technical Problems
Technical Problems Overview
• Cellular radio link channel structure
• Access network architecture
• Legacy circuit switched protocols
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
13
Cellular Radio Link Channel Structure
Optimal Link Layer Structure for IP
•
•
•
•
•
Single channel for data and signaling
Roughly symmetric bandwidth uplink/downlink
Minimal latency
Optimal wired link layer: Ethernet
Optimal wireless link layer: WiFi
– One channel multiplexed by collision detection
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
15
Typical Cellular Link Layer Structure
• Multiple channels for signaling and different kinds of data
• Different channel uplink/downlink multiplexing modes
– Frequency division duplex (FDD): one frequency for uplink, one for
downlink
– Time division duplex (TDD): uplink and downlink multiplex on same
frequency using separate time slots
• Asymmetric bandwidth uplink/downlink
– Uplink is typically much slower (up to half as fast)
– Power consumption on handset is controlling factor
• Latency typically much higher and very variable
– Up to 200 ms in HSDPA/3G networks with 60 ms standard deviation
– This is changing in “4G”
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
16
PHY Channel Structure for Frequency
Division Duplex WCDMA*
Name
Abbreviation
Direction
Function
Synchronization
SCH
Downlink
Cell search
Common pilot
CPICH
Downlink
Phase reference for SCH
Primary common control
P-CCPCH
Downlink
Control cell broadcast
Secondary common control
S-CCPCH
Downlink
Paging and small amounts of data
Downlink shared
PDSCH
Downlink
Shared channel for control or data
Paging indicator
PICH
Downlink
Paging indicator for data on paging channel
Acquisition indicator
AICH
Downlink
Signatures for random access procedure
CPCH access acquisition preamble
indicator
AP-AICH
Downlink
Control channel for bursty data
CPCH status indicator
CSICH
Downlink
Status channel for data status information
CPCH collision-detection/channel
assignment indicator
CD/CA-ICH
Downlink
Collision detection indicators if channel assignment
is not active
High speed downlink shared
HS-PDSCH
Downlink
Very high shared data channel
Shared control for HS-DSCH
HS-SCCH
Downlink
Control for HS-DSCH
Dedicated physical data
DPDCH
Downlink,Uplink
Single terminal data channel
Dedicated physical data control
DPDCH
Downlink,Uplink
Control channel for single terminal data channel
Physical random access
PRACH
Uplink
Contention-based, non-realtime control or data
Physical common packet
PCPCH
Uplink
Contention-based, for bursty data
Uplink dedicated control for HS-DSCH
HS-DPCCH
Uplink
High speed data access control
*There are also 12 channels available in time division duplex mode
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
17
Why So Many Channels?
• Physical layer control of various types
– Estimating and reporting on channel conditions
– Ranging and power control
– Paging (dormant mode location)
• Multiple data channel types are for data having different statistical
characteristics
– Constant bit rate
– Bursty data
• Net Effect
– Cellular protocols can handle faster moving terminals
• IP can’t make much use of so many channels though
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
18
Access Network Architecture
WiFi Access Network
Authentication,
Authorization, and
Accounting Server
Routed Access Network
First Hop
Router
AR
Wireless
Access Point
Laptop
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
20
WCDMA Access Network (UTRAN/GPRS)
Radio
Network
Controller
Class 5
Telephony
Switch
Authentication,
Authorization, and
Accounting Server
First Hop
Router
Wireless “Access
Points” (actually
Base Stations)
Overlay Network in
RAN*
Overlay
Router
GPRS Overlay
Network between
First hop router
and RAN
*May also be an ATM network
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
21
UTRAN/GPRS Stack
GPRS Overlay
Internet
RAN Overlay
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
22
Soft Handover
• Mobile terminal communicates with two or more base stations at a
time
– Special procedure for CDMA because all base stations use same
frequency
– Separate connections use different spreading codes
• Effect on network
– On downlink, RNC segments packets into radio frames transmits
each radio frame to more than one base station
– On uplink, RNC must resolve multiple radio frames into one and
reassemble a packet
• Called “macrodiversity resolution
• Newer cellular link protocols don’t do this
– Poor effect on access network performance
– OFDM (orthogonal frequency division multiplex) and MIMO (multiple
input/multiple output) achieve similar gains with less impact on access
network
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
23
Soft Handover Example
RNC
To the Internet
RAN
Mobile Terminal
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
24
Why the RAN Overlay?
• IP Packets come into the RNC as 1500 byte (typical Ethernet)
• RNC downlink packet processing
–
–
–
–
Segment packets into 300 byte radio frame chunks
Cypher for radio/encrypt
Transmit to base stations with which terminal is in soft handover
Retransmit any packets that are lost
• But interferes with TCP
• RNC uplink packet processing
– Decypher/decrypt
– Resolve any macrodiversity
• Isochronous transmission for precise radio scheduling
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
25
Radio Access Network Management
Protocols and Their Function
• Sophisticated radio management protocols in the network
– NBAP: Between Node B and RNC
– RANAP: Between RNC and MSC
– RNSAP: Between one RNC and another
• Fine grained handover control
– Between RNCs
– Between SGSNs/MSCs
• Dormant mode location tracking and alerting
– Saves power in handset
• Handover is controlled by the network, not the terminal
– Terminal only provides information on SNR, visible base stations, etc.
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
26
What Benefit is Sophisticated Radio
Resource Management?
• Allows better capacity utilization among base stations
– Only the network can see the load in neighboring cells
• Better power utilization – more bits per watt
– Terminal uses high energy channels only when really transmitting or
receiving data
• Better spectral utilization – more bits per herz
– Get the most out of very expensive spectrum
• Better capacity utilization – more terminals per cell
– Get the most out of a very expensive radio network equipment
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
27
Why the GPRS Overlay?
• Support for IP mobility through tunneling
– Tunneling protocol is GPRS Tunneling Protocol (GTP)
• SGSNs maintain different IP subnets
– Mobile terminal never sees these
• When a mobile terminal moves to a different SGSN, IP address
stays the same
– SGSN and GGSN work together to move the GTP tunnel endpoint
from old SGSN to new SGSN
• Mobile terminal never sees a new subnet
– Drastically reduces the overhead in IP mobility
– No need for terminal to configure a new IP address
– Location of terminal is not revealed to correspondent nodes
• IETF is now developing a similar protocol
– Network-based Localized Mobility Management (NETLMM)
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
28
Legacy Circuit Switched Protocols
Legacy Circuit Switched Protocols
• Two basic interfaces from RAN to core network:
– Iu-PS : for “packet-switched” data (i.e. Internet)
– Iu-CS: for circuit switched data (i.e. legacy telephony)
• Protocol on Iu-CS interface is legacy SS-7 based circuit switched
protocol
– Mobile Application Part, legacy GSM protocol
• This protocol “invades” the Iu-PS part too
– Authentication, authorization, and accounting
• But little used IP AAA protocol Diameter can also be used
– Most IP systems use Radius
– Paging (Dormant Mode Location)
• Next generation 3GPP system is removing legacy protocols
– Core: All IP Network (AIPN)
– RAN: Long Term Evolution (LTE)
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
30
What Can CleanSlate Do?
For RAN and Radio...
•
Essentially, nothing
– Unless you don’t care about widespread deployment
•
Cellular radio adoption controlled by large international standards bodies
– ITU, 3GPP, 3GPP2, IEEE
– Heavy business involvement
• Billions at stake
• Large radio infrastructure vendors run heavy simulations to prove their technology
• If a company gets a key patent into the standard, they stand to win big time
– Best idea often isn’t the one that is accepted
• More emphasis on preserving large infrastructure vendors’ and wireless carriers’
business models
•
On the bright side, radio/RAN engineers are now sensitized to the needs
of packet-switching services
– Most current designs, including 3G, were primarily designed for ATM voice
frames
– LTE and 4G are being designed with IP networks in mind
• No legacy circuit switched interfaces
• All services, including voice/media, over IP
• Will this work for InternetNG?
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
32
For the Access Network...
• Virtualization might better support RAN mobility and IP mobility
– Simplify signaling and data planes
– But how to commuicate between virtualized slices?
• What runs in the slices?
• New routing and addressing architecture
– Network mobility integrated with routing
• No overlays needed
– Fixed identifier routing
• Terminal never changes its IP address
• Maybe different transport layer protocols
– Separate congestion control from reliability
– Cross-layer to allow the transport layer to make use of radio link
information
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
33
Summary
Summary
• Much of the problem is nontechnical
– Expensive spectrum acquisition and network buildout
– Irrational fears about deterioration in business models
• Much of the technical problem has to do with the influence of
legacy circuit switched networks on cellular radio design
– In the future, all new cellular protocols will assume IP
• WiMax is the first
• Network architecture may help but the real impact is really beyond
cellular
– Having network-layer mobility integrated with routing will help with
other types of wireless and nonwireless protocols too
– Having a better transport architecture will help with media transport
too
Copyright © 2007 DoCoMo Communications Laboratories USA, Inc. All Rights Reserved.
7/18/2015
35