A Scheduling-based Routing Network Architecture Omar Y. Tahboub & Javed I. Khan Multimedia & Communication Networks Research Lab (MediaNet) Kent State University
Outline Introduction The Scheduling-based Routing Network Architecture Case Study: An Institutional Remote Data Backup & Recovery Network Performance Evaluation. Conclusion and Future Work
Outline Introduction The Scheduling-based Routing Network Architecture Case Study: An Institutional Remote Data Backup & Recovery Network Performance Evaluation. Conclusion and Future Work
Introduction Bandwidth-intensity will be dominating aspect in future emergent Internet applications. Will pose network capacity demands beyond imagination reaching Gigabytes and yet Terabytes per day. Internet2 [1] model will likely be the reference architectural model for the next generation high-performance networks. The Internet2 Dynamic Circuit Networking (DCN) [2] will also be the key communication paradigm. Multi Protocol Labeling Switching (MPLS) [3] play a central role massive data flow routing, switching and forwarding
Introduction Finally, on the basis of the case study, we carried out a performance evaluation study: Demonstrated two simulation experiments. Compared the performance between the scheduled data backup transfer to the conventional non-scheduled. We first describe a scheduling-based routing network architecture namely [4,5]. Implements DCN operation at the MPLS level. Enables time-scheduled route (LSP) information to be disseminated into MPLS domains. Second, we present a case study focusing on remote backup and recovery networking application. Utilized the Ohio Super Computing Network OSCnet backbone. Connects 11 universities in the state of Ohio,
Outline Introduction The Scheduling-based Routing Network Architecture Case Study: An Institutional Remote Data Backup & Recovery Network Performance Evaluation Conclusion and Future Work
The Scheduling-based Routing Network Architecture Figure 1: The Network Architecture [4][5]
The Network Tier For each edge e i E, bw i denotes its bandwidth (bps) and l i denotes its propagation delay in seconds. Figure 2: The Network Tier Represented by G = (N, E). N = {n 1, n 2, …, n m } be the set of m label switch routers. E = {e 1, e 2, …, e n } be the set of edges (links), Each edge e i in E connects a pair of label switch routers (n u, n v ) N. For each switch router n i in N, c i : service rate in bits per second (bps) and b i : the available storage buffer in bits.
The Edge Tier Clients of this architecture are multi-disciplinary demanding various communication services: Telemedicine Content Distribution Distance Learning Figure 3: The Edge Tier Represents the user-groups requesting on-demand data flow transmissions via the network tier.
The Edge Tier A FEC is further presented by a task t defined by the tuple (u, v, o, dl, s), where u: the ingress LER. v: the egress LER. o: the task origination time in seconds. dl: the task completion time deadline in seconds. s: the task size in bits. Figure 4: The FEC as a Task
The Routing Tier The main task of the route scheduling tier is computing time-scheduled routes in the underlying network domain. Figure 5: The Routing Tier Consists of the route scheduling solver.
The Routing Tier Let set R T ={ r 1, r 2, …, r i,…, r n } defines a route schedule as a set of routes, where each task has a route (is committed to a task). Figure 6: The LSP Specifications Given a MPLS domain G = (N, E) Let T denote the set of n tasks Let the route (LSP) r i be a solution to task t i, defined as an ordered set of k node hops (switch routers) H i = {u i, n i,2,…, n i,j, …, n i,(k-1), v i }, or as k-1 link (edge) hops. L i = {e i,1, e i,2,.., e i,j, …, e i,k-1 }, where e i,j connects n i,(j-1) and n i,j.
The Routing Tier
The Scheduling Tier Figure 7: The Scheduling Tier This tier consists of three entities: Node Resource Information Base (NRIB) Link Resource Information Base (LRIB) and Router server.
The Scheduling Tier The Node Resource Information Base (NRIB ) Node resources information includes: Available service capacity (bps) Total service capacity (bps) Total input/output buffers capacity (bytes) and Available input/output buffer capacities (bytes). Figure 8: The NRIB
The Scheduling Tier The Link Resource Information Base (NRIB ) Link resources information includes: Source LSR Destination LSR, Total link capacity (bps), and Propagation delay (seconds). Figure 9: The NRIB
The Scheduling Tier Network Resource Reservation Figure 11: Network Resource Reservation
The Scheduling Tier Route Schedule Dissemination Figure 12:Route (LSP) Schedule Dissemination
Outline Introduction The Scheduling-based Routing Network Architecture Case Study: An Institutional Remote Data Backup & Recovery Network Performance Evaluation. Conclusion and Future Work
Case Study: An Institutional Remote Data Backup & Recovery Network We utilize the Ohio Supercomputing Computing [6] network OSCnet as practical network backbone. Safeguarding data and information against all types of disasters is an urgency. Offline remote data backup & Recovery Networks serves an efficient solution. In organizational Information centers, data & information backup is performed in a daily, weekly and monthly basis.
Case Study: The OSCnet Network Backbone Figure 13: The OSCnet Network Backbone
Case Study: Backup Mirror Site Assignment Table 1: Backup Mirror Sites Assignment
Case Study: Data Backup Transfer Demands Table 2: Projected Daily Transfer Demands
Case Study: Critical Performance Challenges Figure 14: Sample Average Shortest Path Length Stable Chaotic Will Chock out other bandwidth contending Applications
Case Study: Critical Performance Challenges Figure 15: Sample Aggregate Shortest Path Load
Figure 16: The Four- Tier OSCnet-based Network Architecture Case Study: The Network Architecture
Outline Introduction The Scheduling-based Routing Network Architecture Case Study: An Institutional Remote Data Backup & Recovery Network Performance Evaluation. Conclusion and Future Work
Performance Evaluation To demonstrate the performance incentives of scheduled-based data transfer over the classical transfer scheme. Compares the performance achieved by scheduled backup data transfer to the classical unscheduled scheme. This study is conducted as a simulation study of the OSCnet network backbone shown by Figure 13.
Simulation Experiment-1 Setup Link capacity allocation: Unscheduled: Day = 100%, Night = 100% Scheduled: Day = 10%, Night = 90% Number of Tasks: 156. Performance Metrics: Average Shortest Path Length at Link Load 90% Aggregate Shortest Path Load at Link Load 90%
Simulation Experiment-2 Setup Link capacity allocation: Day = 100%, Night = 100% Number of Tasks: 156. Performance Metrics: Overall Task Schedulability Percentage The ration of number of tasks completed by their deadline to total of all tasks * 100%
Simulation Experiment-1 Results Figure 16: Average Shortest Path Length Unscheduled Scheduled
Simulation Experiment-1 Results Figure 17: Aggregate Shortest Path Load
Simulation Experiment-2 Results Figure 17: Overall Task Schedulability
Outline Introduction The Scheduling-based Routing Network Architecture Case Study: An Institutional Remote Data Backup & Recovery Network Performance Evaluation. Conclusion and Future Work
Conclusion and Future Work On the basis of the performance evaluation stud, it can be concluded that Scheduling-based routing significantly improves: The Average Shortest Path Length. The Aggregate Load of the Shortest Path. The Overall Task Schedulability. Presented a four-tier scheduling-based routing architecture namely Demonstrated a OSCnet-based remote data backup case study.
Conclusion and Future Work The Scheduling-based data backup and recovery Near-optimal mirror site exploration and Selection Heuristics. Hierarchical scheduling-based routing network architecture is a centralized architecture. MPLS & CR-LDP Protocol Extensions Timed Route Schedule Dissemination in MPLS networks Pathway Intermittency Route Scheduling in Physically/Logically Intermittent Networks
Thank You !
References [1] The Internet2, Wikipedia, url: [2] Internet2 Consortium, “Internet2’s Dynamic Circuit Network”, [3] E. Rosen, A. Viswanathan, and R. Callon, “Multiprotocol Label Switching Architecture”, RFC 3031, January, [4] O. Tahboub, A Network Architectural Model for Dynamic Circuit Networking at Multiple Protocol Label Switching”, TR , MediaNet Lab,, [5] Tahboub, O., Khan, J., A Network Architectural Model for Dynamic Circuit Networking at Multiple Protocol Label Switching”, The First International Workshop on Concurrent Communication ConCom 2009, Seattle, WA, [6] The Ohio Super Computing Network, url: