120111-impact.org 5.2 KB
Introduction
Content-Centric Networking (CCN)
A data-centric network architecture based on named content
- A unique content identifier is given to every content
- CCN router performs name-based routing with longest prefix matching
- CCN router has a buffer cache for storing forwarded contents
Virtual Content-Centric Networking (VCCN)
An extension to CCN, which enables group-based communication
- A group identifier is embedded in the content identifier
- CCN router is virtualized as multiple instances of VCCN routers
- Several VCCN router instances are logically connected over CCN network
Router Virtualization in VCCN
Several CCN router resources are shared among VCCN router instances
- ContentStore
- FIB (Forward Information Base)
- PIT (Pending Information Table)
Appropriate allocation of CCN router resources to VCCN router instances are crucial in terms of efficiency, fairness, scalability, and security
(Possible) Positive and Negative Impacts of Router Virtualization
-
Positive impact
- Router resource allocation to VCCN router instances increases locality of resource accesses, which should lead to more efficient resource usage
-
Negative impact
- Router resource allocation to VCCN router instances prevents statistical multiplexing effect, which should lead to less efficient resource usage
Motivation
Research interest:
- How does the router virtualization in VCCN affect the performance of VCCN?
Objectives
We evaluate the performance of VCCN (i.e., CCN with router virtualization) under several types of router resource (in particular, ContentStore) allocation methods, and try to answer the following questions.
Research questions:
- How is the overall performance (i.e., throughput, latency) of VCCNs affected by router virtualization?
- How is the individual performance (i.e., throughput, latency) of VCCN affected by router virtualization?
- How is the fairness among VCCNs affected by router virtualization?
- For a given scenario, how should router resources be allocated to VCCN router instances?
Related Works
Related Works
Tortelli et al.
- Tortelli11:Fairness
Psaras et al.
- Psaras11:Modelling
Carofiglio et al.
- Carofiglio11:Experimental
- Carofiglio11:Modeling
- Carofiglio11:Bandwidth
ȽƤ
Simulation
Experiment Setup (1/5)
-
Network topology
- aggregation network topology (binary tree network)
- core network topology (Abilene network)
- hybrid network topology (scale-free network?)
Фɤˤ?
-
Network parameters
- Link capacity: 10 [Gbit/s]
- Propagation delay: 10 [ms]
Experiment Setup: Aggregation Network Topology (2/5)
Experiment Setup: Core Network Topology (3/5)
Experiment Setup: Hybrid Network Topology (4/5)
Experiment Setup (5/5)
-
Contents
- Number of contents: XXX
- Content size distribution: XXX
-
Routers
- ContentStore size: XXXX - XXX
- Cache replacement policy: LRU
-
Workload
- Number of groups: XXX
- Number of users in a group: XXX
- Content request rate per user: XXX
Resource Allocation Methods
-
Shared
- ContentStore is shared by all VCCN router instances
-
Partially-Shared
- A fraction of ContentStore is shared by all VCCN router instances, and the rest is equally divided and assigned to every VCCN router instance
-
Seperated
- ContentStore is equally divided and assigned to every VCCN router instance
Metrics
- Throughput (mean and CV (coefficient of variation))
- Fairness index
- Content delivery delay (mean and CV (coefficient of variation))
- Link utilization (average and maximum)
Simulation Result: The size of assigned ContentStore v.s. Efficiency
ˤ碌ƺƹͤƤ ͽۥ (ǡXY٥뤬ä) Ƥ
Simulation Result: The effect of impact of a group in a given CCN network
ˤ碌ƺƹͤƤ ͽۥ (ǡXY٥뤬ä) Ƥ
ɬפ˱ƥ饤ɿ䤷Ƥ
Conclusion
ˤ碌Ƥ