A Service Adaptation Middleware for Delay Tolerant Network based on HTTP Simple Queue Service

Hao Zhuang, Hervé Ntareme, Zhonghong Ou, Björn PehrsonNSDR 2012 · OpenReview

TL;DR

This paper introduces DSAM, a middleware that simplifies communication between diverse applications and Delay Tolerant Network (DTN) daemons. By abstracting DTN complexities and providing a unified interface, DSAM enables developers to build DTN applications more easily across different platforms, improving system flexibility and fault tolerance.

What Problem Does This Paper Address?

In many parts of the world, traditional internet infrastructure is scarce, creating "communication-challenged areas." Delay Tolerant Networks (DTNs) offer a solution by enabling data transfer even with intermittent connectivity, often using mobile devices as "digital couriers." However, developing DTN applications is complex because they need to interact with DTN daemons, which can be implemented on various platforms (Linux, Android) and written in different programming languages (Java, Python, C/C++, PHP). This heterogeneity makes it difficult to create applications that can seamlessly communicate with DTN services. The paper addresses this by proposing a middleware that bridges this gap, allowing applications to interact with DTN daemons without needing to understand the underlying DTN protocols or platform specifics.

Methods

The core of the proposed solution is the DTN Service Adaptation Middleware (DSAM). DSAM acts as a communication layer, insulating application developers from the intricacies of DTN bundle protocols, operating systems, and network interfaces. Its architecture consists of three main components:

  1. HTTPSQS (HTTP Simple Queue Service): This lightweight service provides a simple message queuing mechanism based on HTTP GET/POST protocols. It's designed to be efficient and handle many concurrent connections, making it suitable for resource-constrained devices like ALIX boards.
  2. Client APIs: These APIs allow applications to send and receive messages to/from HTTPSQS. The paper highlights the availability of cross-language client APIs for Java, PHP, Python, and C/C++, supporting both text and XML message formats.
  3. HTTPSQS Message Processor (HMP): This daemon asynchronously processes messages stored in HTTPSQS queues. It analyzes incoming messages, dispatches them to the appropriate applications based on their logic, and handles serialization/deserialization between local application representations and external formats (text/XML). HMP also implements a request-reply mechanism with retry and dead-queue handling for robustness.

The middleware is built upon DTN-specific protocols but presents a unified interface to applications.

Datasets

The paper does not explicitly mention specific datasets used for training. However, it details performance evaluations using various file sizes (1MB to 100MB) for DTN bundle transmission and a large number of HTTPSQS messages (10,000) with different sizes (32 bytes to 512 bytes) for DSAM throughput testing.

Key Insights and Ideas

  • Abstraction is Key: The fundamental insight is that abstracting the complexity of DTN protocols and platform differences is crucial for enabling wider adoption and development of DTN applications.
  • HTTP as a Unifying Protocol: Leveraging HTTP for the queuing service (HTTPSQS) provides a familiar and widely supported protocol for inter-application communication, simplifying integration.
  • Cross-Platform Compatibility: The provision of cross-language client APIs is a significant enabler, allowing applications written in different languages on different operating systems to interact with the DTN.
  • Decoupling for Resilience: By making application modules interact solely through DSAM, the system achieves decoupling. This means that if one component fails, others can continue to operate, enhancing fault tolerance and reliability.
  • Optimizing for Efficiency: The evaluation highlights that choosing optimal message sizes (e.g., 20MB for DTN bundles, 256 bytes for HTTPSQS messages) can significantly improve throughput and reduce power consumption, which is critical in resource-constrained environments.

Key Results

The paper demonstrates the effectiveness of DSAM through two supporting applications: a DTN2 Network Management Tool (DNMT) and a Wireless Sensor Network (WSN) Environmental Monitoring Application (WEMA).

  • DNMT: DSAM enabled a user-friendly PHP web interface to interact with the Python-based DNMT, which in turn managed the DTN2 daemon. This showcased DSAM's ability to bridge different platforms and languages for management tasks.
  • WEMA: DSAM facilitated communication between a PHP web client, a Java client on SunSPOTs, and a C/C++ DTN2 client, allowing for environmental data collection and transmission in communication-challenged areas.
  • Throughput Optimization: Performance evaluations showed that selecting optimal file sizes for DTN bundles (around 20MB) yielded the highest throughput (6.65 MB/s). Similarly, for HTTPSQS messages, a size of 256 bytes provided the best throughput.
  • Power Consumption: The paper suggests that higher throughput directly correlates with lower energy consumption, as data is transmitted faster. Optimizing file and message sizes leads to significant energy savings.

Limitations and Future Work

Limitations:

  • The paper acknowledges that the performance evaluation of power consumption for file transmission was based on the assumption that power remains constant during the process, which might be an oversimplification due to meter limitations.
  • The evaluation of HTTPSQS message throughput showed a sharp decrease for larger message sizes (512 bytes), indicating a potential bottleneck or inefficiency at that scale.

Future Work:

  • Enhanced Queue Management: The authors plan to add features like queue management and event-driven mechanisms to further improve queuing latency.
  • DTN-Specific Applications: Developing more DTN-specific applications, such as one for connecting DTN regions with social networks like Twitter, is a future direction.
  • Trade-off Evaluation: Further evaluation of the trade-offs introduced by the extra middleware layer is planned.

Summary updated

BibTeX
@inproceedings{zhuang2012service,
  author = {Hao Zhuang and Hervé Ntareme and Zhonghong Ou and Björn Pehrson},
  title = {{A Service Adaptation Middleware for Delay Tolerant Network based on HTTP Simple Queue Service}},
  booktitle = {NSDR},
  year = {2012},
  url = {https://openreview.net/forum?id=DTUQe1j4Cv},
}
Figure