In the test surrounded by AWS Graviton 4C8G.2XLARGE with 8 VCPU in the Valkey 8.1.1, it was possible to achieve a performance of 999.8 thousand SET checks per second, while Redis 8.0 reached a level of 729.4 thousand requests per second. In general, Valkey throughput was 37% higher than Redis for SET operations and 16% for GET. At the same time, compared with Redis, the Valkey project demonstrated a decrease in delays when processing requests by 30% for SET operations and 60% for Get operations.
Separately analyzed changes in the throughput and delays, depending on the number of parallel processors performed in multi-flow input/output processing mode. Up to 3 streams Valkey and Redis show approximately equal results, but then Valkey breaks forward. At 6 streams on the system with 8 VCPU, VALKEY productivity amounted to 678 thousand SET queries per second, and Redis – 563 thousand requests per second in a limit in 256 simultaneous connections. With an increase in connections to 400, ValKey performance increased to 832 thousand SET queries per second.
After optimizing interrupt processing in the system to reduce the number of context switching in Valkey, it was possible to raise performance to 999.8 thousand SET queries per second. The essence of the optimization was reduced to the allocation of 2 VCPU for processing interruptions and binding 6 of the remaining VCPU to the input/output processing flows to exclude the migration of handlers between the CPU.
Sudo ethtool -l ENS34 COMBINED 2 # Limit up to the 2nd number of IRQ GREP ENS34 /ProC /Interrupts # We