Support of Strong Consistency on Fog Applications

Diogo Lima, Hugo Miranda, François Taı̈ani2019 · DOI

TL;DR

This paper tackles the challenge of maintaining strong consistency in mobile applications deployed on Fog Computing architectures. They propose using a novel Light-Weight Group (LWG) abstraction, built upon virtual synchrony, to create fine-grained, dynamic groups of edge servers. This approach aims to balance the need for strong consistency with the scalability demands of large-scale distributed mobile applications.

What Problem Does This Paper Address?

Modern mobile applications, like collaborative games or smart city platforms, often require all users to see a consistent view of shared data. Traditionally, this is achieved by centralizing data in the Cloud. However, this introduces significant latency and jitter for mobile users due to geographical distance. Fog Computing offers a solution by deploying edge servers closer to users. Yet, this shift creates a new challenge: managing distributed transactions across these edge servers while ensuring strong consistency. Simply replicating data isn't enough; applications need to guarantee that all participants observe the same state, even when data is distributed and frequently updated by mobile users. This paper addresses how to achieve this strong consistency in a scalable manner within a Fog Computing environment.

Methods

The core of the proposed solution lies in leveraging virtual synchrony and introducing a new abstraction called Light-Weight Groups (LWGs). Virtual synchrony provides a framework for processes to communicate and maintain synchronized views within groups. The authors extend this by proposing LWGs, which are fine-grained, dynamically formed groups of edge servers (surrogates).

Here's how it works:

  • Geo-aware State Deployment (GeoDS): A background service (GeoDS) monitors application usage patterns and intelligently deploys application state fragments to the most convenient edge servers. This aims to maximize local transactions and minimize latency.
  • Light-Weight Groups (LWGs): When regional transactions (involving data across multiple surrogates) are identified as frequent and important by the GeoDS, LWGs are dynamically created between the involved surrogates. These LWGs are designed to be small, with minimal members, reducing the overhead of group membership management and message delivery.
  • Two-Tier Group Architecture: LWGs form the "partial membership" layer. For global transactions (involving many surrogates), a "Heavy-Weight Group" (HWG) encompassing all surrogates is used, leveraging the underlying virtual synchrony support. This two-tier approach aims to provide strong consistency for both regional and global transactions while maintaining scalability.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The Trade-off Between Consistency and Latency in Fog: The paper highlights that while Fog Computing brings data closer to users, it doesn't automatically solve consistency issues. In fact, distributed transactions on the edge can exacerbate these challenges.
  • Dynamic, Fine-Grained Grouping is Key: The core innovation is the idea of creating small, ad-hoc groups (LWGs) of edge servers specifically for transactions that frequently involve their data. This is a departure from traditional, static group management.
  • GeoDS as an Enabler for LWGs: The GeoDS is crucial for identifying the need for LWGs. Without intelligent analysis of data access patterns, the creation of LWGs would be arbitrary and potentially ineffective.
  • Scalability Through Granularity: By making LWGs small, the authors aim to circumvent the inherent scalability limitations of traditional virtual synchrony, which struggles with large group sizes.

Key Results

The paper presents a conceptual framework and a proposed architecture. It does not include empirical results or quantitative performance evaluations. The key "results" are the architectural design and the theoretical benefits of the LWG approach:

  • Improved Scalability: The LWG abstraction is theorized to provide better scalability for strong consistency in Fog environments compared to traditional approaches.
  • Reduced Latency for Regional Transactions: By forming specific groups for frequent regional transactions, the system aims to reduce the overhead and latency associated with these operations.
  • Maintained Strong Consistency: The core promise is to deliver strong consistency guarantees, which are critical for many mobile applications, even in a distributed Fog setting.

Limitations and Future Work

The authors acknowledge that the paper presents a conceptual solution and outlines future work:

  • Lack of Empirical Validation: The proposed solution has not been implemented or evaluated with real-world performance metrics.
  • Future Work: The immediate next step is to materialize the proposed solution and validate its benefits through evaluation with a concrete large-scale distributed mobile application. This would involve implementing the GeoDS and the LWG mechanism and measuring its performance against existing approaches.

Summary updated

Abstract

International audience
BibTeX
@misc{lima2019support,
  author = {Diogo Lima and Hugo Miranda and François Taı̈ani},
  title = {{Support of Strong Consistency on Fog Applications}},
  year = {2019},
  doi = {10.1145/3301419.3323969},
}
Figure