Load Balancing Optimization of In-Memory Database for Massive Information Processing of Internet of Things (IoTs)

Based on the analysis of the key technologies of the Internet of Things service platform architecture, a load balancing optimization scheme of in-memory database based on the massive information processing of the Internet of Things service platform is proposed. This scheme firstly proposes a system model that can satisfy the mass sensor information processing under the open platform environment and designs several functional unit modules of the system. By combining these functional units, the service can be configured for thousands of services and tenants. This paper presents an adaptive strategy selection method, which can automatically select the optimization strategy by dividing the position and querying the selection rate to improve the efficiency of the adaptive index algorithm. The index structure is initialized by parallel sorting algorithm, and the query statement is executed and the index structure is optimized by thread level parallel and radix sort methods. An elastic pipeline technique is proposed, which includes an elastic iterator model and a dynamic scheduler. The elastic iterator model is an upgrade of the traditional iterator model, adding the characteristics of dynamic multicore execution. In the process of query processing, dynamic scheduler monitors the load of each node in real time and dynamically adjusts the parallelism, so as to realize the load balance of in-memory database and maximize the utilization of hardware resources. The elastic pipeline realizes the isolation of parallelism from query compilation to avoid inappropriate parallelism allocation caused by missing and insufficient information during query compilation.