When you configure Azure Stream Analytics you only have 2 levers;
Streaming Units (SU) – Each SU is a blend of compute, memory and throughput between 1 and 48 (or more by contacting support). The factors that impact SU are query complexity, latency, and volume of data. SU can be used to scale out a job to achieve higher throughput. Depending on query complexity and throughput required, more SU units may be necessary to achieve your performance requirements. A level of SU6 assigns an entire Stream Analytics node. For our test we wont change SU
SQL Query Design – Queries are expressed in a SQL-like query language. These queries are documented in the query language reference guide and includes several common query patterns. The design of the query can greatly affect the job throughput, in particular if and/or how the PARTITION BY clause is used.
Rolf tests along three margins: 2 versus 16 input partitions, 2 versus 16 output partitions, and whether to partition the data or not. Read on to see which combination was fastest.