Fog ecosystems differ in their needs from the usual cloud environments. Storing and accessing data in fog needs specialized data placement algorithms that are used in place of the database’s default partitioning and replication mechanisms.
To tackle the fog needs and challenges, we build upon the popular open-source distributed in-memory database Apache Ignite, creating a fog-tailored DBMS with fast I/O and persistency guarantees. Ignite is used as a “special” distributed database in the sense that only local data structures are used instead of the default partitioned & replicated choices that are common in distributed systems.
We implement two different data placement algorithms that tackle different fog challenges and are complementary with each other to move stored data efficiently and transparently from one instance to another. The first one aims to place data from unstable to stable fog nodes while the second one considers the network topology as well as the communication overheads when a distributed processing engine is continuously querying data.
- Anna-Valentini Michailidou
- George Vlahavas
- Theodoros Toliopoulos
- Anastasios Gounaris
- Vasileios Psomiadis
- Athina Vakali (Datalab director)