Transcript WR - Computer Science and Engineering

What do you expect from a
Cellular service Provider
- LOW SUBSCRIPTION FEE
- HIGH QUALITY CONNECTION
7/23/2002
1
Outline
•
•
•
•
•
•
•
Review of the wireless network operation
Limitations of the current network architecture
Description of the new architecture and benefits
Backhaul cost reduction: Analysis and results
Increased availability: Analysis and results
Conclusions
Further research topics
7/23/2002
2
A Simple Wireless Network
Mobile Data Set
Base Station
Controller
(BSC)
Mobile
Switching
Center (MSC)
PSTN
Mobile
Voice Unit
Packet
Network
Base Transceiver
System (BTS)
Packet InterWorking Function
Challenge
7/23/2002 is to keep connection and not3loose any data during handoff operation
The Components
• BTS
– BTS consists of one or more transceivers placed at a single location.
The BTS terminates the radio path on the network side.
• BSC
– Provides allocation and management of radio resources.
– SDU: Selection and distribution unit. Also responsible for handoff
coordination
• MSC
– Provides and controls mobile access to the PSTN. Interprets the
dialed number, routes and switches call to destination number. Also
manages mobile’s supplementary services. Maintains a register of
visitors operating within the coverage area of the MSC’s connected
BTSs.
• PDSN: Packet data service node is basically a packet router.
7/23/2002
4
Current Wireless Network Architecture
MSC
PDSN
TDM
channels
Packets
BSC
(SDU)
BSC
(SDU)
BSC
BSC
(SDU)
TDM
channels
in the central office
BTS
7/23/2002
24xDs0
in T1
- Backhaul cost is by $$$/mile
- 10-100 miles between BTS and BSC
- Voice or data use one DS0 channel at a time
- BTSs are located in the tower
- BSC and MSCs are located
BTS
BTS
BTS
BTS
5
BTS
Soft Handoff between two BTS
Handoff: A handoff mechanism is needed to maintain connectivity as devices
move, while minimizing disruptions to ongoing calls. This mechanism
should exhibit low latency, incur little or no data loss, and scale to a large
network.”
7/23/2002
Handoffs == ( Hard
|| Soft )
6
SDU and soft handoff
- 3 to 6 BTSs involved in soft handoff
- SDU changeover due to weak signal
from primary BTS
- BTS forwards even corrupted
radio frames to the SDU for selection
SDU
-2
WR-B
WR-A
BTS-2
SDU
-2
SDU
SDU-1
BTS-1
7/23/2002
-1
BTS-3
7
Problems with the current
architecture
•
Duplicate traffic on the links: Frame selection is done at the BSC (One frame
is generated for each soft handoff leg.) This results in duplicate traffic flow at the
backhaul
•
•
•
No traffic aggregation: Each call is allocated DS0 capacity. Even when there is
no activity on the call, the DS0 is reserved. At this rate, currently each BTS can
support only around 20 calls per sector (normally 3 sectors per BTS). So, no traffic
aggregation – does not utilize statistical multiplexing (results in inefficient backhaul
link provisioning)
If six BTS, then more overhead: Seventy percent (70%) of wireless
operators expenditure is on RAN. Around 30% expenses are backhaul cost. Only
15% of the BTS-BSC traffic is payload and rest is overhead. If six BTS are involved in
the soft handoff then the overhead will be lot higher.
Carries dead payload: BTS forwards even error frames to BSC. Because the
selection is done at the BSC. This means we are carrying dead payload to BSC.
7/23/2002
8
Problems with the current
architecture (cont..)
• Uneven utilization of links: For data services as well as
voice, IP
networks overlay on top of current wireless networks. This is very
inefficient, not cost effective, and very difficult to deploy the new services.
• Performance: Less propagation delay. Currently, for some of
the BTS-BSC configurations, the links run 100 miles, it means
several milliseconds of delay. This creates problem during soft
handoff.
• Availability: Less availability due to single point of failure (in
terms of base stations, base station controllers and links
connecting between them)
7/23/2002
9
Review
•
•
•
•
•
Simple wireless network operation
Different components in the network
Wireless network topology
Mobility and soft handoff
Problems with the existing architecture
7/23/2002
10
-85.2
34
33.9
highway
Rural
Urban
33.8 Typi
US c
33.7 BTS
30x3
33.6
mile
33.5
33.4
33.3
33.2
33.1
7/23/2002
-85
-84.8
11
-84.6
-84.4
-84.2
-84
Solution: Distributed Control
Wireless Router is an IP router
with RF termination. Functions
include BTS, SDU, power
control, and handoff control
WR
Gateway Router is a IP router
connected to the internet core
WR
WR
GR
GR
Gateway Router
Gateway Router
Each packet
Contains several
radio frames.
WR
WR
WR
WR
WR
WR
WR
WR
WR
WR
WR
WR
WR
7/23/2002
Radio frames
Embedded in
IP packets
WR
WR
12
Solution (cont.)
GR-GR links
GR
Gateway Router
Transit traffic management is
handled by IP routing, QoS
GR
Gateway Router
WR-GR links
WR/BTS
WR/BTS
WR/BTS
WR-WR
Links
7/23/2002
WR
WR/BTS
- Radio frame routing
using IP routing
WR/BTS
- Radio neighbors exchange
resource info. using IP
routing protocols
- Mobility is handled through
IP signaling protocols
- Radio resource management
is handled by IP traffic
management
WR/BTS
- WR assumes SDU function
WR/BTS
WR-WR
Links
13
WR
WR/BTS
Solution (cont.)
– Distributed SDU. Distributed bearer and control
– Radio Routing Protocol: IP routing merged with Radio frame
routing for soft handoff and mobility. Wireless extensions of OSPF
with radio neighbors.
– Radio Discovery Protocol: Discovering Radio Neighbors
– Radio Resource Management: IP traffic management
merged/enhanced with Radio Traffic Management. Radio Power
Control integrated and aligned with IP QoS
– RSVP extensions for Radio: IP resource management
merged/enhanced with Radio Resource Management. For example,
Radio Resource Management and soft hand off signaled with
RSVP.
– IP transport in Radio Access Network
Disruptive Technology
7/23/2002
14
Benefits
• Cost Reduction: Efficient use of backhaul links and
aggregation. Only 15% of the BTS-BSC traffic is payload
and rest is overhead. Around 30% expenses are backhaul
cost. Objective is to reduce the backhaul traffic and small
number of high speed backhaul links.
• Scalability:
– Separation of call processing and bearer paths
– Distributed SDU and Distributed Control
– Ability to provide coverage and capacity during peak hours
• Redundancy and Availability: Due to meshed architecture,
network is robust and works around the failures.
7/23/2002
15
Benefits (continued)
• Marriage of IP and wireless protocols: Seamless operation of
IP-Network-Layer with Radio Control.
• Reuse already deployed routers in the Central Offices.
• New wireless services: Automatic Reconfiguration of Radio Access
Network. Expand cell attributes to provide more capacity
• Performance: Less propagation delay. Currently, for some of the
BTS-BSC configurations, the links run 100 miles, it means several
milliseconds of delay. This creates problem during soft handoff. Due to
short lengths between base stations, the delay will be negligible.
7/23/2002
16
2G/3G RAN Network (Traditional)
Interoffice distance (costs per mile) cost + Fixed Cost
CO
CO
CO
CO
Channel
Termination
Cost
Channel
Termination
Cost
BSC
MSC
BTS
BTS
7/23/2002
BTS
BTS
BTS
BTS
17
What configuration is best in
terms of cost and availability
•
Cost Reduction: How and where to place wireless router in the
RAN network with respect to network-level backhaul cost and
availability. For example, existing WR can be part of GSR (high end
router) ? How close it to BTS
•
Higher Availability: Distributed IP-RAN for backhaul cost
savings and higher availability
Variables in analysis
–
–
–
–
7/23/2002
Different kinds of transport cost structures
Different types of links (T1/T3/Microwave etc.)
Different types of connectivity between BTS and wireless router
Number of carriers supported supported by BTS
18
Cost Model
•
•
•
•
•
•
•
Commercial BTS network topology. BTS are connected to BSC using the
chain of Central Offices.
Real ILEC cost structure is assumed for the T1 leased lines from urban and
rural areas to the BSC.
20-30 Node network in urban area and 10 BTS in rural areas.
Soft hand factor of 2.0: How many BTS are in soft handoff
1.5 T1 per BTS per carrier
BTS Network Regions: I) Dense Urban, ii) Urban iii) Suburban iv) Rural
Cost models
– Channel termination costs
– Interoffice fixed costs
– Per mile costs: Transport cost changes according to distance and the type of
transport (T1/T3/OC3)
7/23/2002
19
Configurations
Four BTS network configuration are considered:
1.
2.
3.
4.
7/23/2002
Traditional BTS-BSC network (Config-1, fully star, call it
“Traditional”)
One Wireless Router supporting multiple BTSs (10-30). For
example, existing WR can be part of GSR (high end router) ? How
close it to BTS (Config-2, call it “star”)
Meshed Wireless Routers (Config-3, one wireless router per BTS,
full mesh within the central office region. Call this “WR”)
Meshed Wireless Routers (WR) and each WR connected to
multiple Gateway Routers for higher availability (Config-4, with
connection to all the nearby high end routers. Call this “WR-HA”)
20
2G/3G RAN Network (Traditional)
Interoffice distance (costs per mile) cost + Fixed Cost
CO
CO
CO
CO
Channel
Termination
Cost
Channel
Termination
Cost
BSC
MSC
BTS
BTS
7/23/2002
BTS
BTS
BTS
BTS
- Around 150 BTS per BSC
21
WR supporting multiple BTSs (Star Topology)
WR
WR
CO
CO
WR-BTS
links
WR-BTS
links
BTS
BTS
BTS
BTS
BTS
BTS
BTS
BTS
BTS
BTS
BTS
WR supports multiple BTS (10-20). Selection and distribution is done in the WR. WR collocated with CO
7/23/2002
22
BTS
WR collocated with BTS (WR)
GR
Gateway Router
GR-GR links
Central Office
GR
Gateway Router
Central Office
WR-GR
links
WR/BTS
WR/BTS
WR/BTS
WR/BTS
WR
WR/BTS
WR-GR
links
WR/BTS
WR/BTS
WR-WR
Links
WR
WR/BTS
WR-WR
Links
* WR-GR link (primarily) used for backhauling selected traffic to the destination
* WR-WR link (primarily) is used for selection and distribution traffic between
two wireless routers.
* GR-GR links are links between two IP routers. These routers do not distinguish between wireless and wireline traffic
7/23/2002
23
WR collocated with BTS with HA (WR-HA)
GR
Gateway Router
GR
Gateway Router
Central Office
Central Office
WR/BTS
WR-GR
links
WR/BTS
WR/BTS
WR/BTS
WR
WR/BTS
WR
WR/BTS
WR/BTS
WR-WR
Links
WR/BTS
WR-WR
Links
GR is collocated in the the closest CO
Connectivity to two (atleast) GRs is established for higher availability. The second GR is collocated in the
neighboring Central office.
7/23/2002
24
Traffic Models
data
stream
voice
stream
Overhead
/ stream
voice
stream
Overhead
/ stream
Overhead
/ stream
Overhead
/ stream
Segment.
Segment.
multiplexer
multiplexer
packetizer
Dt
packetizer
Dt
Traffic Mix
•
•
•
•
100% voice + 0% data
100% data, 14.4K, 64K and 144Kbps
80 % voice + 20% data, 14.4K, 64K and 144Kbps
20 % voice + 80% data, 14.4K, 64K and 144Kbps
7/23/2002
data
stream
scheduler
Overhead /
container
25
Cost Comparisons—Urban (30 node network)
$$$$$
800000
700000
600000
WR (one per BTS)
WR with HA
Traditional
500000
400000
300000
Star (10 BTS per WR)
200000
100000
0
No. of Carriers (increased bandwidth at BTS, 1carrier requires ~~2 Mpbs)
7/23/2002
26
Cost Comparison—Rural (around 10 nodes)
600000
500000
400000
WR
WRHA
Trad
Star
300000
200000
100000
$$$$
0
1
2
3
4
5
6
7
8
9
10
11
12
No. of Carriers (increased bandwidth at BTS, 1carrier requires ~~2 Mpbs)
7/23/2002
27
Why less cost ??
• Backhaul cost reduced with WR mesh architecture.
Customer saves $$$$ (per mile backhaul cost)
• Why is the backhaul much less with WR? What is
new with WR?
– Frame Selection is done at the WR. No duplicate traffic
after the selection is done
– The aggregation of voice and data traffic from multiple
WRs enables better Statistical multiplexing and reduces the
backhaul requirement. This also enables customer to use less
costly T3 and saves them more $$$$
– Flexible and more reliable traffic routing.
7/23/2002
28
Review and Conclusions
• Statistical multiplexing and compression techniques are not accounted
in the results described. If counted, more savings are realized
• Cost savings for one carrier are not much but substantial for multiple
carriers.
• Star at the near-by-CO with a high speed (e.g., DS3/OC3) uplink is the
optimum but without higher availability.
• Mesh is ideal for higher availability and cost savings
• All the WR cases win over traditional deployment
• Ongoing work: Different kinds of transport cost structures and
different types of links (T1/T3/Microwave etc.)
7/23/2002
29
Service Availability
CO
CO
CO
-Single point of failure at BTS
- Single point of failure at BSC
- No rerouting around congested nodes
possible.
CO
BSC
BTS
WR
MSC
BTS
BTS
BTS
BTS
BTS
- Rerouting around failed links/nodes
- Rerouting around congested links/nodes
- Wireless router is also IP router hence no
need to deploy full mesh
- If BTS fails, neighboring resources can be
used to complete the calls.
7/23/2002
WR
WR
30
WR
WR
WR
WR
WR
WR
Availability Model (example)
Backplane
Line Card
Control
Processor
SW
Line card
Control
Processor
SW
Calculate MTBF and MTTR of all the components in each element
WR
WR
GR
Tower
WR
WR
GR
7/23/2002
WR
WR
WR
WR
31
Annual Down Time Comparison
Link availability is varied from 0.99 to 0.99999 and downtime is computed
For traditional and WR-HA network topologies.
Link Avail .99
Traditional 174
(Hours)
.999
17.77
.9999
2.01
.99999
0.438
WR-HA
(Minutes)
5.25
5.25
5.25
5.25
BTS Availability: 0.99999
GR Availability: 0.99999
7/23/2002
32
Annual Down Time Comparison
Link availability is varied from 0.99 to 0.99999 and downtime is computed
For traditional (2G/3G) and WR-HA network topologies.
Link Avail
.99
.999
.9999
.99999
Traditional 174
(Hours)
17.77
2.01
0.438
WR-HA
(Minutes)
5.78
5.78
5.78
5.85
Assumptions:
BTS Availability: 0.99999
GR Availability: 0.99
7/23/2002
33
RESULTS
• In conventional case, when a link fails, the call level reliability is low
because the call is not redirected without dropping. However, in mesh
architecture, the call can still be maintained and the rerouting takes
place at the frame/packet level.
• Though GR’s availability is only 0.99, the overall service downtime is
not impacted due to the fact that there are multiple paths from BTS to
another GR.
7/23/2002
34
Review and Conclusions
• Distributed RAN architecture saves backhaul cost (less than half of the
cost of existing architecture )
• Distributed RAN architecture supports 99.999% service-level
availability compared to conventional network. In fact, full mesh is not
required for realizing the 99.999% availability. Even partial mesh can
achieve this level of availability
• Distributed architecture increases the complexity of control and
requires working prototype for understanding the new protocols.
• Disruptive technology: Required new protocols design and approval
from vendor and hence may take long time to get to the field (this is
weakness in this architecture).
7/23/2002
35
Further Work
7/23/2002
36
Content Based SDU
•
•
•
•
The solution in this patent describes how loss/delay/delay variation objectives are
met for different services in SDU. These SDUs cater to different loss, delay and jitter
budget requirements.
The SDU selection and its parameters are setup during the call setup. If the call
characteristics change during the duration of the call, the behavior of the SDU is
also changed.
A three dimensional grid, with each dimension representing loss, delay, jitter, is
defined. Next, each service is placed in this grid depending on how sensitive the
service is with respect to loss, delay or delay sensitivity.
Advantages
 Efficient traffic engineering
 Extension of the service level agreements to provide a comprehensive QoS
 Enhancement of end to end QoS
7/23/2002
37
Fast and Smooth SDU change over
• Fast and smooth SDU (anchor) handover between two wireless
routers.
 Fast means low latency and smooth means no packet loss.
 Frame-level handover between two wireless routers
 Handover mechanism should coexist with native IP or Mobil-IP
(either version 4 or version 6) protocol defined by IETF (at
network layer)
 Handover should also be transparent to the mobile.
 Handover mechanism should be applicable to UMTS/CDMA 2000.
7/23/2002
38
Wireless Content Processor (WCP)
Problem:
To date, the size and cost of Wireless equipments like BTS, BSC, MSC,
SGSN, GGSN, and PDSN are huge (cost and space) and there are different
technologies for different standards (2G, 2.5G, 3G, CDMA, GPRS,
UMTS). Vendors have separate boxes (examples listed above) for each of
these technologies. Wireless Service providers are asked to invest huge
amount for these overlapping technologies covering the same geographical
areas. In addition, these technologies and standards are changing very
rapidly. For cost-savings, space-savings, scalability, back-haul savings,
rapid deployment of new services, and efficient air coverage it is
imperative to introduce programmable wireless-content-specific network
processors.
7/23/2002
39
Wireless Content Process (WCP)
Proposed Solution:
Network processors being developed in the current industry are
specially designed for the wireline-data processing. To our knowledge,
there is no body developing the wireless-specific content processors.
This paper describes the WCP role in wireless-specific content
processing. Several new and existing services can be enabled by using
the WCP and these will be described in this document. In particular,
how WCP can process wireless protocols and control bearer paths,
associate bearer paths with signaling will be described in this paper. In
addition, WCP role in scaling PDSN and other wireless network
elements will be described.
7/23/2002
40