Effective Buffer and Storage Management in DTN Nodes

1. Effective Buffer and Storage Management in DTN Nodes Stylianos Dimitriou, and Vassilis Tsaoussidis Dept. of Electrical and Computer Engineering Democritus University of Thrace Xanthi, Greece E-DTN ’09 – St. Petersburg, Russia

2. Why storage management (1/2) • Contact opportunities in DTN networks can be either deterministic or stochastic. • The time space between two data transmissions may be long and usually varies during the connection. • Persistent storage is used in DTN nodes to accommodate data in cases of intermittent connectivity. E-DTN ’09 – St. Petersburg, Russia 2/11

3. Why storage management (2/2) • Due to node mobility and, occasionally, service disruption, the communication between two nodes, may change from high-delay to earth-like. • In earth-like communications, the need for persistent storage is trivial. • A DTN node should be able to adjust its storage management based on communication characteristics. • With efficient storage and buffer management, we can achieve better resource exploitation and service differentiation. E-DTN ’09 – St. Petersburg, Russia 3/11

4. Solution • We propose a mechanism that decreases the delay in cases of frequently communicating nodes via better storage management. • The proposed mechanism maintains data that will be transmitted soon in the buffer, and moves the rest of the data in persistent storage. • Aims to decrease the delay inflicted by unnecessarily moving packets from buffer to storage and back. E-DTN ’09 – St. Petersburg, Russia 4/11

5. Node mobility • In DTN networks, usually delays are long. • Queuing and processing delay is only a small part of the total delay, and thus negligible. • The aim is to achieve better bandwidth usage. • Due to node mobility, two nodes may engage in constant low-delay transmissions. • In this case, the delay required to move packets from buffer to storage and back becomes a significant part of the total delay. E-DTN ’09 – St. Petersburg, Russia 5/11

6. Proposed model E-DTN ’09 – St. Petersburg, Russia 6/11

7. Allowed packet transfers • 1. LDT buffer  Persistent Storage: • New HD packets / no communicating opportunity. • Old LD packets / full LDT buffer. • 2. Persistent Storage  HDT buffer: • Old HD packets / communicating opportunity occurs. • Old LD packets / communicating opportunity occurs. • 3. LDT buffer  HDT buffer: • New HD packets / communicating opportunity occurs. E-DTN ’09 – St. Petersburg, Russia 7/11

8. Distinguishing LD and HD traffic • We calculate the average storage delay of all flows. • Flows with bigger average storage delay than the average, are classified as HD. E-DTN ’09 – St. Petersburg, Russia 8/11

9. Conclusions • DTN networks involve node mobility which may vary significantly the propagation delay throughout a connection. • Nodes may involve in low-delay communication which resembles wired communications. • Managing buffer and storage is necessary to maintain high transmission rates in these cases. • The proposed scheme aims to resolve the lack of flexibility of Bundle Protocol to adjust to node mobility. E-DTN ’09 – St. Petersburg, Russia 9/11

10. Future work • Determine the optimal sizes for LDT and HDT. • Review the WFQ scheme. • Review the algorithm that updates the average delay. • Evaluate the mechanism, both analytically and experimentally. E-DTN ’09 – St. Petersburg, Russia 10/11

11. Questions? • Thank you E-DTN ’09 – St. Petersburg, Russia 11/11