Queries Performance

Query execution time and resource consumption depends on several factors:

• Number and size of virtual tables defined by query's context references. For example, in a 1000-devices installation CPU and memory load implicated by users.*.devices.*:temperature context reference will be 1000 times higher than one of users.admin.devices.meter1:temperature references.
• Nature of variables or functions of server contexts referred by the query. See variables performance and functions performance for more information.
• Complexity of the query (i.e. usage of JOINs, functions, etc.)

Debugging a query is the way to find out what stages of the query execution process are slow.

Similarly to any other SQL queries, Iotellect queries performance dramatically depends on existence of indexes.

Indexes are automatically created on-the-fly during query execution for any key fields of query source variables or function results.

See Key Fields section for more information.

Queries are specifically optimized for working in distributed architecture. Once a source table is being build from a context reference, data from all referred contexts connected via distributed architecture is queried concurrently. If multiple distributed contexts are located on different servers, all servers will prepare and send their part of data in parallel.

Operation of queries in distributed architecture is very similar to MapReduce concept:

• Remote servers prepare partial data by performing loading, filtering, sorting and aggregation of data on their side
• Original query running on the consumer server unions coming pieces of data and provides additional sorting, filtering and aggregation
• Moreover, the query may join resulting dataset with other datasets coming either from the local server or via distributed architecture

Was this page helpful?

Tell us more

Send

Thank you!