Beijing University of Posts and Telecommunications
OAI Based Wireless Distributed Storage Systems
06/21/2018Jianbin Chuan, Jiajia Wu, and Ge Wang
High Performance Computingand Networking Lab
Team leader:
Prof. Li Wang
Research Lab--HPCNHigh-Performance Computing and Networking Lab (HPCN) • Currently: 3 Ph.D. , 1 joint Ph.D (Peking University), and 12 Master Students• Alumni: 16 Master, 1 joint Ph.D. and 20 joint Master Students• Funding: 2 NSFC funding, China Mobile, Intel, Ministry of Education
Research Interests• Heterogeneous Networks• D2D Communications• Massive MIMO• Physical Layer Security• Social Networks• Distributed storage Systems• Computation offloading
Approaches and Tools• Matching theory• Graph theory• Contract theory• Convex optimization• Fractional Programming• Integer and Combinatorial Optimization
1/21
Outline
Theoretical Achievements
Project Results (WDSS)
Future Works
1
3
5
2
4
Challenges and Problems
Background and Motivations
Challenges:
Potential techniques:
Improve communication quality Increase wireless data demand Download duplicated content
D2D communications
– mobile devices can
communicate directly
• Increase network
capacity
• Improve spectral and
energy efficiency
• Extend coverage area
Computational offloading
– offload computational tasks
• Lower transmission latency
• Save energy for mobile devices
• Enable computation-intensive
and latency-critical applications
at the resource-limited mobile
devices
Distributed storage– cache popular content
items
• Reduce transmission
delay
• Improve reliability for
content sharing
• Offload the traffic of BS
Background and Motivations2/21
Background and Motivations
Computational offloading:
Mobile Edge Computing (MEC)
• Offload computational tasks to deployed MEC servers
Fog Computing
• Offload computational tasks to close-by devices
Base station
Mobile user
MEC server
Mobile user
Mobile user
Computational tasksOffload to MEC server
Computational results returning
back by MEC server
Offload to mobile users
Computational results returning
back by mobile users
• Advantages:• Lower latency
• Energy saving
• Device function extending
• …
3/21
Distributed Storage:
Non-coded caching• Straightforward replication of the data in multiple storage nodes
Coded caching • Content items are coded and stored in different devices
Background and Motivations
A1
A2A3
Coded caching
• Low redundancy
B B
B
B A
BS
A1 A2 A3
• High secrecy• …
• Low latency • High quality• Low energy
consumption• …
storage
Non-coded caching
4/21
(𝒏, 𝒌)MDS code:
• Principle: A content file with a size of 𝑴 is
partitioned into 𝒌 pieces, encoded and stored in 𝒏Content;
• The original file can be recovered by connecting
any set of 𝒌 CPs (for both content download and
content repair);
• Advantages: Comparing to MBR and MSR,
MDS is the best simplest erasure codes and
can provide an optimal tradeoff between
redundancy and complexity.
coded caching: Erasure Code
• MDS (Maximum Distance Separable Codes)
Redundancy Reliability
Tradeoff
Regenerate Codes• MBR(Minimum Bandwidth Regenerate Codes)
• MSR(Minimum Storage Regenerate Codes)
Repair Process
M New CPMCR
Download Process
M M/kM/k M/k…
Divide into k pieces
M/kM/k M/k M/k…1 2 3 n
Coded into n pieces
• Storage: each CP stores 𝜶𝑴𝑫𝑺=𝑴
kbits;
• Repair: the amount of data that needs to be
retrieved from the network to repair a failed node
𝜸𝑴𝑫𝑺 = 𝑴 bits;
Background and Motivations5/21
Outline
Theoretical Achievements
Project Results (WDSS)
Future Works
3
5
2
4
Challenges and Problems
1 Background and Motivations
Theoretical achievement
Published 90+ ; SCI journals 32 ; best papers in international academic conferences 3.
D2D Communications1. L. Wang, H. Tang, H. Wu, and G. Stüber, “Resource Allocation for D2D Communications Underlay
in Rayleigh Fading Channels,” IEEE Transactions on Vehicular Technology, vol. 66, no. 2, pp.1159-1170, Feb. 2017. (2区,IF:4.066)
2. L. Wang, and G. L. Stuber, “Pairing for Resource Sharing in Cellular Device-to-Device Underlays,”IEEE Network, vol. 30, no. 2, pp. 122-128, Mar.-Apr. 2016. (2区,IF:7.230)
3. L. Wang, H. Tang, and M. Cierny, “Device-to-Device Link Admission Policy Based on SocialInteraction Information,” IEEE Transactions Vehicular Technology, vol. 64, no. 9, pp. 4180-4186,Sept. 2015. (2区,IF:2.642)
Distributed Caching1. L. Wang, H. Wu, Y. Ding, W. Chen, and H. Vincent Poor, “Hypergraph Based Wireless Distributed
Storage Optimization for Cellular D2D Underlays,” IEEE Journal on Selected Areas in
Communications (JSAC), vol. 34, no. 10, pp. 2650-2666, Oct. 2016. (1区,IF:8.085)2. L. Wang, H. Wu, Z. Han, P. Zhang, and H. V. Poor, “Multi-Hop Cooperative Caching in Social IoT
Using Matching Theory,” IEEE Transactions on Wireless Communications, vol. 17, no. 4, pp. 2127-2145, Dec. 2017.(2区,IF:4.951)
3. L. Wang, H. Wu, and Z. Han, “Wireless Distributed Storage in Socially Enabled D2DCommunications,” IEEE Access, vol. 4, pp. 1971-1984, Mar. 2016. (3区,IF:3.244)
Computational Offloading1. Y. Ai, L. Wang, Z. Han, P. Zhang, and L. Hanzo,"Social Networking and Caching Aided Collaborative
Computing for the Internet of Things", Communication Magazine, 2018, submitted.2. M. Guan, B. Bai, L. Wang, S. Jin, and Z. Han, “Joint Optimization for Computation Offloading and
Resource Allocation in Internet of Things,” in Proc. IEEE Vehicular Technology Conference-fall (VTC-fall), Toronto, Canada, Sept. 24-27, 2017.
6/21
Communicationresource
Computational resource
Cache (storage) resource
Low Delay & High quality
High Security & High reliability
High Energy efficiency
High Spectrum efficiency
D2D Communications
Distributed Cache (Storage)
Computational Offloading
6/21
Outline
Theoretical Achievements
Project Results (WDSS)
Future Works
3
5
2
4
Challenges and Problems
1 Background and Motivations
U4
U5
U6
U1
U2
WDSS
Server
U3
D2D link Fragment
Content
RepairContent
Request
RS(4,3) code
4 fragments
Database
Wireless AP
Decode
Encode
Content
OAI/Wi-Fi link
Repair
Control
3 fragments
2
1
2
3
4
leave
Scenario of wireless distributed storage
system (WDSS) implementation
Project Results (WDSS)
7/21
Project Results (WDSS)
Architecture design of WDSS
8/21
Project Results (WDSS)
Function of modules in WDSS
Content Management
User Management
Request Handler Repair HandlerWireless
Distributed
Storage
System
Server
Links FeedbackPower Adjustment
Handler Units
Content Management
User Management
Request Handler
Power Adjustment
Repair Handler
Links Feedback
Manage the storage and encoding of content
Manage the login and signoff of users
Handle the content request of users
Handle the content repair of users
Adjust the router power according to user needs
Feedback the Bluetooth RSSI of users
9/21
Project Results (WDSS)
Wi-Fi based WDSS (finished)
10/21
Project Results (WDSS)
OAI based WDSS
11/21
Project Results (WDSS)
WDSS sever GUI and application GUI
12/21
Outline
Project Results (WDSS)
Future Works
3
5
4
Challenges and Problems
Theoretical Achievements2
1 Background and Motivations
Future Work
Our finished work
Our problems:
1) Exhausting too much
signaling resources;
2) Inefficient and blocked
3) Invalid for the devices
with mobility
Potential solution:Using the Network Slicing to
provide corresponding telecommunication services and
network capabilities.Purple marked slice——high
speed, large data size, low delay:
transmit content from cloud center to
users.
Orange marked slice——large
scale, low delay, high reliability:
maintain the function of the
distributed wireless cache system.
Black marked slice——large
traffic, mobile broadband services:suitable for the mobility scenario
Slices and corresponding service requirements
13/21
Future Work
OAI function in Network Slicing supported WDSS
1、OAI supports the full protocol stack of 3GPP
standard. With OAI, the transceiver functionality is
realized via a software radio front end which has strong
programmability inherently.
2、Network Slicing allows multiple logical networks
with different demands to be created on top of a common
shared physical infrastructure. Thus, it can fulfil different
network service in our system.
3、OAI can support Network Slicing flexibly by using
FlexRAN.
14/21
Summary: OAI can provide corresponding
telecommunication services and network capabilities
for users by allocating 3C resources efficiently with the
help of Network Slicing technique.
Network Slicing Based WDSS
Future Work
• OAI: C-RAN architecture adopts the IF4p5
• FlexRAN:A Flexible and Programmable Platform for Software-Defined Radio
Access Networks
• Network Slicing: A Network Slicing Prototype for a Flexible Cloud Radio Access
Network[Foukas2016] X. Foukas, N. Nikaein, M. M. Kassem, M. K. Marina, and K. Kontovasilis, “FlexRAN: A Flexible and Programmable Platform for
Software-Defined Radio Access Networks.” in Proc. Conference on emerging Networking EXperiments and Technologies (CoNEXT '16), ACM,
New York, NY, USA, pp. 427-441.
[Costanzo2018] S. Costanzo, I. Fajjari, N. Aitsaadi and R. Langar, “A Network Slicing Prototype for A Flexible Cloud Radio Access Network,” in
Proc. IEEE Annual Consumer Communications & Networking Conference (CCNC), Las Vegas, NV, 2018, pp. 1-4.
15/21
Future Work
FlexRAN based network slicingSO
SDNC
SRM
Slices
SDB
PRV
PRM
PR
FlexRAN Agent
FlexRAN Protocol FlexRAN Protocol
FlexRA
N
Co
ntro
llerP
lane
No
rthb
ou
nd
AP
ISo
uth
bo
un
dA
PI
FlexRAN Master Controller
FlexRAN AgentFlexRAN Agent
Apps
FlexRAN Agent
FlexRAN Agent API
eNodeB data plane
FlexRAN Agent API
eNodeB data plane
SO: Slice OrchestratorSDNC: SDN Controller SRM: Slice Resource ManagerSDB: Slice Dedicated Bandwidth
PRV: Physical Resource Virtualization PRM: Physical Resource ManagementPR: Physical Resource
FlexRAN Agent
FlexRAN Protocol FlexRAN Protocol
FlexRA
N
Co
ntro
llerP
lane
No
rthb
ou
nd
AP
ISo
uth
bo
un
dA
PI
FlexRAN Master Controller
FlexRAN AgentFlexRAN Agent
Apps
FlexRAN Agent
FlexRAN Agent API
eNodeB data plane
FlexRAN Agent API
eNodeB data plane
16/21
Future Work
SO: instantiate the CN elements by using PNF or VNF and creat theCN instances
SDNC: 1) create slice instances for the services
2) specify the CN elements to the slices
3) manage the slices by usingSRM and SDB
SRM: schedule resources for UEs belongingto its Slice
SDB: enforce policy by the SO to the PRVwhen the slice is first created
APPs
FlexRANAgent
FlexRANMasterController
eNBDataPlane(RRU)
Paradigm of SDN-based network slicing
17/21
[Ksentini2017] A. Ksentini and N. Nikaein, “Toward Enforcing Network Slicing on RAN: Flexibility and Resources Abstraction,” IEEE
Communications Magazine, vol. 55, no. 6, pp. 102-108, 2017
Future Work
Network Slicing supported FlexRAN Agent
eNB FlexRAN Agent(RCC)
RCC
RRU
18/21
Future Work
Network Slicing supported FlexRAN Master Controller
Main functions:1) Information monitoring
of the RAN2) Task management
of users 3) Events notification service4) Virtual resource
management5) Slices management
19/21
Outline
Project Results (WDSS)
Future Works
3
5
4
Challenges and Problems
Theoretical Achievements2
1 Background and Motivations
The works we have finished
1) Published papers on related area
2) Developed a wireless distributed storage system based on both the Wi-Fi and the OAI (no FlexRAN )
3) Developed a FlexRAN system and realizing the static slicing
Challenges and Problems
The works on going
1) To develop a Networking Slicing supported FlexRAN system
2) To facilitate wireless distributed storage system using the Network Slicing supported FlexRAN system
EPC
RCC
RRU
20/21
Challenges and Problems
Problems in Network Slicing
supported FlexRAN
• Is it possible to control and manage 3C
resources through FlexRAN based
Network Slicing?
• Can OAI support these services? What
changes may be needed?
• It’s hard to maintain the dynamic
resources, e.g., RB, to multiple slices.
• It doesn’t performance well for the users
with mobility.
• The robustness of the isolation among
multiple slices is unfriendly.
21/21
Thank you !
Beijing University of Posts and Telecommunications
Contact: Li Wang [email protected]
High Performance Computing and Networking Lab (HPCN)School of Electronic Engineering, BUPT
116#, Beijing University of Posts & Telecom,10 Xitucheng Rd., Haidian Dist.,Beijing 100876, China