Dai Jiao

Llama: Distributed Multi-versioned Model Data Management for SCADA

In Supervisory Control and Data Acquisition (SCADA) systems, it is of great significance maintaining all Model versions for tracking, diagnosing and recovering from the system shutdown. With the Model data increasing rapidly, traditional relational databases are not capable to cover the scalability issue, thus there comes up the more scalable NoSQL stores. However, integration to NoSQL brings up several new challenges: firstly NoSQL have little capacity on schema management of structured data, secondly Model has many versions but changes little, which lead to much unwanted redundancy, lastly, SQL supported on NoSQL could lead to a decline in access performance. We propose Llama, a distributed multi-versioned Model data management architecture to fix these problems. Llama provides general schema management to manage the schemas and attributes of structured data in NoSQL. Incremental version maintenance and time aggregation compression strategies are applied to reduce the average storage overhead. Furthermore, Llama reaches the current hot spot version Model with low-latency, with it cached in memory and write penetration implemented. Experiments show that the average storage overhead is reduced to 20% in Llama, and the read speed of hot spot is more than 10 times faster.