Transcript Mobile-CORD-August-28-2015x

Mobile CORD and
Proposed Use Cases
August 2015 ON.Lab , SK Telecom, and AT&T
1
Mobile Operators Wishes
 Reduce content delivery cost over mobile networks
 Open CPRI model: White-box, Disaggregated eNB, vRAN
 Distributed Core: EPC disaggregation
 Mobile resource slicing: local MVNO and enterprise
 Revenue cooperation with over the top
 Real-time analytics for network aware applications
 Improve QoE and drive service innovation
 Customized service chaining and policies
2
CORD and Target Domains of Use
Infra Slices
Residential
Virtual infra +
ONOS + vOLT,
vCPE, vBNG,
vCDN
5G
Virtual infra +
ONOS +
mobile edge
over multi-RAT
4G
Virtual infra +
ONOS +
mobile edge
over multi-RAT
Enterprise
Virtual infra +
ONOS + VPN,
TE, vCDN
…
ONOS (Virtualization, Slicing) + OpenStack (Multi Domain) + XOS
Leaf-Spine
Fabric
BBUs
(Multi
-RATs)
PON
OLT
MACs
Commodity Servers, Storage, Switches, and I/O
ROADM
(Core)
Enterprise
Metro
Ethernet
3
Mobile CORD: Big Architecture Picture
Operator’s specific Applications
Radio
Resource
Scheduling
Network
Slices
LTE,5G
IoT
Enterprise
MBB
MVNO
Testbed
Real-time
Analytics
Service
Chaining
…
Orchestration
View
Mobile Resource & Service Orchestration CORD
Platform (ONOS + XOS + OpenStack)
Networking
Fabric
Video
Caching
SGW-D
XOS
vBBU
DNS
PGW-D
OpenStack
vBBU
Security
ONOS
CPRI (E)
Disaggregated Edge Service
BBUs
SGW-D, PGW-D: Data plane of SGW/PGW
Functions
Distributed
EPC
SGW-C/ PGW-C: Control plane of SGW/PGW
CORD
Platform
MME
PCRF
SGW-C
SGW-D
Backbone
Switch
PGW-D
PGW-C
Control
Apps.
Topology
View
Front-haul
Optical
Switch
vBBU
Centralized
EPC
4
POC
For Mobile CORD
5
POC Goals of Mobile CORD
1. Back-End:
2. Front-End :
Bring service functionality of mobile core to the edge
Realize the Values of eNB disaggregation
•
•
•
•
•
•
L4-L7 Service chaining at the edge
Cache @ Mobile Edge
Disaggregated EPC @ Mobile Edge
Localized Service customization
QOS
Eliminate Inefficiencies
•
•
•
•
•
Capex, Opex
Data Plane Programmability
Increased Radio Resource Optimization
Unified QOS support
Increased RAN Performance
Realize all Benefits of CORD
• Data Center Economics
• Agility of Cloud
6
POC #1
POC #1
Back-end POC
:Services at Mobile Edge
• By bringing service functionality and disaggregated EPC to mobile edge under the
framework of CORD,
• Improve operators’ agility, efficiency of network utilization and enhance users’ QoE
7
Problem Statements
POC #1
 All traffics are gathered to the center of mobile core, which cause
▫ load on backhaul, backbone transport and core systems like SGW/PGW
▫ waste of network resources from operators’ perspective
▫ deterioration on QoE from users’ perspective
 Operators over-provision their infrastructure to cope with peak traffic
▫ It’s inefficient to have surplus systems to handle peak traffic
▫ Deploying new infrastructure takes long time; a couple of months
▫ Traffic surge varies with time and location
▶ Efficiency, Agility and Scalability are essential for operators to keep up with
rapid growth and dynamic characteristics of mobile traffic
8
Solution and benefits
POC #1
▶ Services at mobile edge
- Certain contents like streaming of same video can be served at mobile edge without traversing
from and to the mobile core
▷Benefits :
▫ Decrease burden on backhaul and mobile core
▫ Increase QoE of users
Non-local Traffic
Local Traffic
▶ On-demand provisioning
- Virtualized infrastructures can be deployed when and where they are to be consumed on demand
▷Benefits
▫ Avoid inefficient provision of infrastructure
▫ Agile action for dynamic traffic request
9
POC Scenario
POC #1
 Local video streaming service at mobile edge
▫
▫
▫
▫
On-demand provisioning of vBBUs at cell sites near big sport match
On-demand provisioning of video caching application(VM) for local video caching service
Functions like DNS and DPI also need to be deployed locally for traffic classification
Other traffic of spectators is treated same as before; travers from and to the Centralized Core
 Local communications hosted by distributed EPC
▫ Virtualized EPC can also be deployed to host local and internal communications
- Communication between security staffs
- Remote monitoring of Security CAM
Local Traffic
vBBU Pool
Edge Service
EPC
Mobile CORD
(Distributed DC)
RRH
10
POC Implementation






POC #1
BBU: Servers hosting VMs of BBU function
Edge Service Functions: Servers hosting VMs of service functions like DPI, DNS and Video caching
Distributed EPC: Data plane of disaggregated EPC (possibly white boxes)
Control Applications: Control plane of disaggregated EPC as northbound applications of ONOS
Mobile CORD Platform: Servers hosting control S/Ws like XOS, OpenStack and ONOS
Mobile CORD Fabric: White-box switches networking mobile edge components
Networking
Fabric
Front-haul
Optical
Switch
vBBU
Video
Caching
SGW-D
XOS
vBBU
DNS
PGW-D
OpenStack
vBBU
Security
ONOS
CPRI (E)
Disaggregated Edge Service
BBUs
SGW-D, PGW-D: Data plane of SGW/PGW
Functions
Distributed
EPC
SGW-C/ PGW-C: Control plane of SGW/PGW
CORD
Platform
MME
PCRF
SGW-C
SGW-D
Backbone
Switch
PGW-D
PGW-C
Control
Apps.
Centralized
EPC
11
POC Implementation (S/W Architecture)
POC #1
 XOS
▫ Orchestrate infrastructure services (by OpenStack) and control services (by ONOS)
 OpenStack
▫ Responsible for provisioning VMs and virtual networks
 ONOS
▫ Manages switching fabric
▫ Forwarding control:
‘video traffic’ to local edge service servers and distributed EPC
and other traffic to centralized EPC
 Applications
▫
▫
▫
▫
Video content caching
DNS
DPI
MME, PCRF, SGW-C, PGW-C
12
POC Plan
POC #1
 POC Target: March 2016 (ONS Summit)
 Detailed Plan: TBD
POC
Demo
Lab
Trial
OCT. 2015
Design
Dec. 2015
Solution Sourcing
Implementation
with partners
Jan. 2016
Integration
March 2016
Test
XOS modification
ONOS modification (OF wireless extensions)
OpenStack modification
vBBU implementation
Disaggregated EPC implementation
Application modification
13
POC #2
POC #2
Front-end POC
:Realize the Values of eNB disaggregation
14
Front-end Problem Statements
POC #2
 Today eNBs are proprietary and non-interoperable among vendors and expensive resources such
as radio is at the mercy of proprietary vendor algorithms
 Certain Control Functions such as schedulers are vendor specific
 New Mobility Services introduction are not agile
 Localization and adaptability to zones, time of day , and traffic behaviors are limited
Proprietary
Control Plane
Data Plane
Ciphering
Mobility
Control
RLC
Handover
forwarding
PDCP
Compression
Flow QoS
Segmentation
Retransmission
HARQ
MCS
MIMO
MAC & PHY
Antenna &
Resource
Scheduling
Security
Configuration
Load Balancing
Scheduling
CoMP
Power Control
15
eNB Disaggregation Model
POC #2
CORD Applications
(Mobility Control, higher layer scheduler & traffic characterization, load Balancing & QoS functions, Handoff and
Radio Resource Optimization, Congestion avoidance & Customer Care)
CORD Interface ( OF Wireless Extensions)
BBU, RRU Resource
Abstraction
Protocols, Tunnel Management ….
Transport and IP Layers (TCP/UDP, IP, RTP, …)
Scheduler Parameters: QCI Assignment, Buffers, …
CORD BBU
MAC Layer
(ARQ, Burst Allocation, FEC, Probes, Converged Channel, CPRI-End-Points)
Physical Layer 5G, 4G, LTE, WiFi
(Coding, Antenna, MIMO, OFDM, dark silicon, Power Mgmt, Probes, CPRI-End-Point …) Plug-in cards
16
POC Solutions & Benefits
•
•
•
•
•
POC #2
CORD to enable dynamic CPRI switch for mobility dimensioning
RAN abstraction and small cell splitting with auto interference control & discovery
Intelligent access network selection and routing (multi-RATs)
Vehicular secured VPN service
Elastic Fixed LTE Service (Wireless fixed BW on demand with 8x8 MIMO )
BBU Racks
RRH
RRH
Wi-Fi
hotspots
Compression
Ciphering
CPRI
Future(IQ over Ethernet)
Segmentation
Retransmission
HARQ
RRH
Wi-Fi
hotspots
CORD
Handover
forwarding
Antenna &
Resource
Scheduling
X1 over
OF Interface
Apps
Mobility Control
Security Config
Flow QoS
Load Balancing
Scheduling
CoMP
Power Control
MCS
MIMO
17
Hybrid, Programmable eNB Data Plane
DSP
RRH
Re-TX
CORD BBU
POC #2
CORD-APP
HARQ
Resource
Scheduling
MCS
Controller
MIMO
DSP
RRH
RRH
Mobility Control
Re-TX
CORD
interface
Security
Configuration
HARQ
Compression
Resource
Scheduling
Handover
forwarding
MCS
Ciphering
Load Balancing
MIMO
Segmentation
Scheduling
DSP
Re-TX
HARQ
Resource
Scheduling
COTS
Flow QoS
CoMP
Power Control
Virtualization
Layer
MCS
MIMO
18
Implementation Approach
Control Apps
Closed
Serv. Apps
POC #2
Mgmt Apps
MOBILE CORD PLATFORM
Data Plane Func.
Open OS
Plug-Ins
BaseBand
BBU Servers
19
Implementation
POC #2
 BBU: disaggregated control function from data plane Edge
 Mobile CORD Platform: Mobility Treatment and traffic functions among
 vBBUs Mobile CORD Fabric: White-boxes; LTE, 5G BBU signaling stack and RRC functional components
Operator’s specific Mobility Management Applications
Scheduler Policies, QOS, Security, IOT, Load & congestion Policies, SON
Mobile Resource & Service
Management CORD Platform
CPRI (E)
FPRF & FPGA
Lime-micro
Altera( CPRI-End-Point)
FrontHaul
Optical
Switch
Mobile
Backhauls
RRH
Leaf-Spine
Fabric
Disaggregated
BBUs
CPRI-O/E
CPRI-End-Point
L4-L7
Services
ONOS, XOS,
Open Stack
eNB Control /
Optimization
Applications
20
RAN + EPC + Mobile NFV Virtualization
POC #2
Mobile Resource & Service Management
Leaf-Spine
Fabric
BBUs
Macro
RRU
CPRI/ETHER
CPRI-O/E
CPRI-End-Point
EPC
Services
ONOS, XOS,
Open Stack
Small
Cells
21
CPRI Switching Scenario
POC #2
Mobile Edge Virtualized RAN/BBU sharing RRUs, and Micro Cells
22
Longer Term Mobility Management Wish list
POC #2
• Autonomous Load balancing among eNBs/BBUs
• Control eNB/BBUs treatment for different traffic types
• Controlled handover for best utilization of radio resources
• Minimize # of tunnels for mobile services
• Demonstrate opportunity for Mobility on Demand
• Demonstrate Multi-homing instead of single home
• Opportunities for different kinds of eNBs that are not bearer specific
• Demonstrate on demand analytics and instrumentation of probes for
autonomous control, optimization, and customer care
23
POC Plan
POC #2
 POC Target: `2Q, 16
 Detained Plan: TBD
Component
Disaggregated
BBU
ONOS
Applications
2015
3Q
Solution
Sourcing
2016
4Q
1H
2H
Develop
Modification / Develop
Integration
Test
Modification / Develop
24
CORD Realization Benefits Summary
1. Data Center economy
2. Agility of Cloud
3. MVNO Model ( third Party)
4. Mobile Edge architecture maps nicely to CORD architecture
5. OF enabled eNBs can be architected like vCPE & vOLTe
6. Netconf enabled eNB can accept configuration and policy updates in real time
7. Edge Cloud can support emerging services such as IOT, Big data, social & internet
8. XOS Instrumentation as a service maps perfectly to Real-time network data
analytics to Improve infrastructure’s efficiency, with more intelligent and
optimized networks.
25
Backups
26
Mobile CORD: Mapping to Future Infrastructure
Control Platform
& Applications
DB
Distributed DC
.
HSS
PGW-D
SGW-D
Centralized DC
27
Mobile CORD: Network Slicing example
Slice#1
Slice#2
Slice#3
Slice#4
Orchestration – CORD (XOS, OpenStack, ONOS)
Resource Recomposition
Mobile
Edge
HW/SW
Resources
vBBU,
vRAN, vEPC,
VNFs
Mobile
Edge
HW/SW
Resources
Centralized CORE HW/SW Resources
(VNFs, MME, SGW, PGW, PCRF, SON, Mobile Control Applications)
Distributed CORE HW/SW Resources
(SGW, PGW, Edge Services / MME, PCRF, SON, Mobile Control Applications)
RAN HW/SW Resources
(BBU, RF, RRH, Spectrum Pool, Backhauls…)
28
Open Source References
“Prior to the advent of software such as OpenBTS and OpenBSC, many would have deemed GSM to be too
complex and arcane a technology for a community of open source developers to tackle. However, those who
thought this to be the case have, like early Linux detractors, since been proved wrong thanks to a growing
community of talented engineers who are committed to open source.” https://myriadrf.org/blog/open-source-lte/
“OpenLTE is developed by Ben Wojtowicz and the first release was made in late 2011. At present GNU Octave
code is available for test and simulation of downlink transmit and receive functionality, along with uplink PRACH
transmit and receive. In addition, GNU Radio applications are available for downlink transmit and receive to and
from a file, and for receive use with a selection of SDR hardware. More recently, a simple eNodeB application
has been added.”
“The bladeRF is the first open source RF project to bring USB3.0 onto the board and combines the Lime FPRF
chip with an Altera Cyclone IV FPGA. This combination allows it to create exceptionally complex networks on
any mobile communications standard or frequency.
The $420 board has been designed for both the hobbyist and the professional developer and is also USB2.0
compatible, allowing it to connect directly to the Raspberry Pi and the Beagle board too.?”
29
Use case
Mobile CORD Use Case
1) Performance KPIs: anomaly detection, RCA,
and mitigation
Requirements
-
Real-time or near real-time nature. Data
should be available with low latency.
Granularity can be in seconds or minutes.
-
Identify anomalies in performance KPIs across time and space. Drill-down to specific rootcause. E.g., is only 1 eNB impacted or entire market? Is only one device class or service
type impacted? Or is it not even a network issue, but end-point (service provider) issue.
-
Take corrective action to mitigate or eliminate the problem if possible. E.g. load balancing,
power control, others.
-
Predict user performance/QoE before and after corrective action and choose best option.
E.g. will a user be better served by hand-off or by changing QoS parameters.
30
Use case
Autonomous Control
Requirements
Mobile CORD Use Case
2) HotSpot Solutions: Identification of
trouble-spots and solutions.
-
Real-time decisions to long-term
planning
Granularity can be in seconds or
hours/days.
-
Identify persistent hotspots with high usage/load based on historical trend.
-
Identify mitigation options for short-term (re-direct other base stations, load balancing, handoffs, etc.) and impact on user QoE.
-
Identify transient hotspots and mitigation options. E.g. base stations along a freeway may be
congested during commute times only. What is the cost-benefit analysis of adding new sites for
such scenarios?
-
How to densify the network based on this hotspot info: what technology may be appropriate
(small cells, DAS, WiFi, others), based on network and user information (user concentration,
data or voice heavy, user mobility, etc.). Cost-benefit analysis of different approaches if possible.
31
Use case
Mobile Cord Use Case
3) Admission control: Congestion detection & QoS
solutions.
Requirements
-
-
Real-time or near real-time nature. Data
should be available with low latency.
Granularity can be in seconds or minutes.
-
In a given scenario (during peak congestion times, in high-cap venues, or presence of
premium users), identify which apps/services can be given low priority based on service
parameters, user preferences, network settings, etc. E.g., should we take resources from
m2m users on the cell, or maybe delay email download or ftp session?
-
During congestion, identify which parameters can be tuned to better manage resources. E.g.,
priority based on ARP, resource allocation based on QCI, max speeds based on MBR, others.
Analyze impact of multiple options in real-time and select best approach.
-
Help avoid dimensioning for peak loads or busy hours with intelligent traffic management.
32
Use case
Big Data Analytics
Requirements
Mobile CORD Security Use Case
4) Security anomaly detection
-
Fine-grained data required.
5) End-to-end flow detection and classification
-
How do we track a flow across multiple
servers/VMs?
Is there any impact on flow performance from
compute/network/storage performance?
What happens to user data when VMs go
down or fail? Impact on QoE?
-
33
ONOS Application Examples
 RAN management Application Examples:
▫ Load balancing
▫ Power Control
▫ Interference coordination
 Virtualized eNB ( Basebands) are programmed to send network events over X1/OF
▫ CSI and flow state records to ONOS controller
▫ Controller maintains network graph, link quality, CSI and flow demand tables
 RAN management applications to program the ONOS wireless graph abstractions (eNBs)
to coordinate the use of the shared radio resource and implement functions such as
CoMP, Interference mitigation, Radio optimization
34
ONOS Controller ↔ Data Plane Interface
 Interface is match/action, where:
▫ Match: Matches on UE IDs, GTP Tunnel IDs and Flow IDs
▫ Action: Is a packet processing pipeline that has to be executed for the
matched set of packets
•
Spans PDCP to L1
•
Specifies handover forwarding rules (if any), segmentation parameters, resource block
(spectrum) assignments, MIMO streams and the L1 stack to be used for processing the packet
•
Enables precise control over how the radio resource is allocated and used
Source: Sachin Kathy
35
SDN combined with SON
 SON & Fronthaul/Backhaul Control
•
•
Today, SON has been focused on the optimization of Remote Radio Head only
Intelligent control of Fronthaul/Backhaul could be achieved by making full use of analytics derived from SON
Main Features;
- Neighbor Mgmt.
- H/O Optimization
- Fault Mgmt.
- Diagnosis
 All those are for RRH Mgmt.
eNB Data
SON
Existing
Solution
SDN Controller
RRH
Fronthaul
Virtualized BBU
Pool
Distributed DC
Backhaul
New
Area
Useful Analytics
can be used for
Fronthaul/Backhaul control
Possible Features;
- BW Scheduling of Backhaul
- On-demand BW allocation
- Failure(of vBBU) Mgmt.
Back-Bone
Transport
36
Network Slicing
 Networks for the Customers’ needs (SLA)
•
•
Traditional: One network for all purpose with same architecture and configuration
Trend: Network Slices to meet specific demands of customers and services (Network as a Service)
[ Concept References ]
※ Reference: NGMN
• Key Enablers for Network Slicing
- Abstraction of network functions
- Programmability of Networking
- End to End Orchestration
※ Reference: SK Telecom
NFV
SDN
37
EPC disaggregation (UP/CP decoupling)
 CP/UP Separation
•
For the full Flexibility of Packet Core in handing ever-growing traffic, UPs must be separated from CPs
and White-Boxes may replace them.
[OpenFlow enabled EPC Architecture ]
[ Traditional EPC Architecture ]
Decoupled Control Plane of
SGW/PGW
HSS
MME
S6a
MME
PCRF
S1-MME
Gx
S11
eNB
(BBU)
SGW
SGW
S1-U
SGW
PGW
SGi
PCRF
OpenFlow*
Distributed
GW (SGW-U)
Distributed
GW (SGW-U)
Internet
Distributed
GW (PGW-U)
Internet
SGi
Distributed
GW (SGW-U)
S5
Control/User Plane
Integrated GW
•
•
•
•
eNB
(BBU)
SGW-C PGW-C
SDN Controller
User Plane of
Core GW (PGW)
(Dotted line)
User Plane of
Distributed GW (SGW)
Control I/F
EnhancedOpenFlow
(Solid line)
(Solid line)
User Traffic
User Traffic
Control plane is tightly coupled with Data plane(User plane)
Limited Scalability of Data plane
Distributed Management
High CapEx/OpEx
•
•
•
•
Disaggregate Control plane and User plane
Logically centralized control and Programmable
White-box User plane. Scalable with data traffic growth
Reduce CapEx/OpEx
38
Service Aware Chaining
 L4-L7 Service Aware Chaining
• Traffic is steered through the appropriate L4-L7 network functions for a given flow, based on service type, user profile, traffic
patterns, or other characteristics.
• 3GPP defined Sd interface between PCRF and TDF to detect service traffic and to apply policy
User Profile &
Policy Control
PCRF
Sd
TDF
L7 Traffic
Classification
Virtualization Platform
39