Benchmarks
Last updated: 8th March, 2025
We benchmarked DiceDB using our membench benchmark utility with various configurations and various machine types, here are the details
Results Observed
On a Hetzner CCX13 machine with 2 vCPU and 8GB RAM here are our numbers around throughput and GET/SET latencies.
| num-clients | Throughput (ops/sec) | GET p50 (ms) | GET p90 (ms) | SET p50 (ms) | SET p90 (ms) |
|---|---|---|---|---|---|
| 1 | 3159 | 0.288767 | 0.354303 | 0.294911 | 0.364543 |
| 2 | 5415 | 0.325631 | 0.473087 | 0.327679 | 0.473087 |
| 4 | 9628 | 0.309247 | 0.518143 | 0.315391 | 0.528383 |
| 8 | 13030 | 0.344063 | 0.757759 | 0.350207 | 0.765951 |
On a Hetzner CCX23 machine with 4 vCPU and 16GB RAM here are our numbers around throughput and GET/SET latencies.
| num-clients | Throughput (ops/sec) | GET p50 (ms) | GET p90 (ms) | SET p50 (ms) | SET p90 (ms) |
|---|---|---|---|---|---|
| 1 | 3694 | 0.258047 | 0.309247 | 0.264191 | 0.315391 |
| 2 | 7310 | 0.251903 | 0.323583 | 0.254975 | 0.325631 |
| 4 | 15655 | 0.227327 | 0.337919 | 0.230399 | 0.339967 |
| 8 | 26459 | 0.248831 | 0.475135 | 0.250879 | 0.483327 |
Comparing with other in-mem databases
On a Hetzner CCX23 machine with 4 vCPU and 16GB RAM here are our numbers around
throughput and GET/SET latencies when compared with other in-memory databases
for num_clients = 4.
| Metric | DiceDB | Redis |
|---|---|---|
| Throughput (ops/sec) | 15655 | 12267 |
| GET p50 (ms) | 0.227327 | 0.270335 |
| GET p90 (ms) | 0.337919 | 0.329727 |
| SET p50 (ms) | 0.230399 | 0.272383 |
| SET p90 (ms) | 0.339967 | 0.331775 |
Execution Configuration
We ran DiceDB with default configuration on one VM and membench on another.
$ ./dicedb
██████╗ ██╗ ██████╗███████╗██████╗ ██████╗ ██╔══██╗██║██╔════╝██╔════╝██╔══██╗██╔══██╗ ██║ ██║██║██║ █████╗ ██║ ██║██████╔╝ ██║ ██║██║██║ ██╔══╝ ██║ ██║██╔══██╗ ██████╔╝██║╚██████╗███████╗██████╔╝██████╔╝ ╚═════╝ ╚═╝ ╚═════╝╚══════╝╚═════╝ ╚═════╝
2025-03-09T16:46:03+05:30 INF starting DiceDB version=0.1.02025-03-09T16:46:03+05:30 INF running with total_commands=212025-03-09T16:46:03+05:30 INF running with engine=ironhawk2025-03-09T16:46:03+05:30 INF running with port=73792025-03-09T16:46:03+05:30 INF running on cores=42025-03-09T16:46:03+05:30 INF running with shards=4membench was run with the following configuration.
The following command creates a load of num-clients concurrent clients
creating one connection each on dicedb (running on host and port). Each client fires
num-requests operations on the database.
$ ./membench benchmark \ --database dicedb \ --host xx.xx.xx.xx \ --port 7379 \ --num-requests 100000 \ --num-clients 4We iterated num_clients and underlying DiceDB VM and the numbers are as shown above.