Report from Digitalocean managed Kubernetes cluster.

Cluster size: 5 Nodes CPU Optimized droplet - 8 vCPU 16 GB RAM

Estimated droplet cost's for cluster: $800/month

Mainflux services scaled to 5 instances

Fan-in Scenario

Publisher pool: 10 000

Subscriber pool: 1

Test durition: 5min

RPS: 1

QoS Level: 2

In this scenario, 10 000 publishers connets and sends 1 one message every second, one subscriber is receiving all those messages.

No message loss was detected.

NOTE: Graphs Time unit is microsecond.

We see total number of messages 3M

We see total number of messages per second (PRS) is 10 000

We see connection ACK latency

Useful facts:

Max latency was up to 20sec

95% Of clients had latency from 5sec up to 15sec under high load

Avarge latency was 5sec without pekas

We see Publish received ACK RPS

We see Total Publish received ACK Latency

Useful facts:

Max latency was up to 8sec

95% Of clients had latency from 0.5sec up to 6sec under high load

Avarge latency was ~2sec without pekas

Results metrics Results metrics are generated by mzbench tool

Digitalocean Kubernetes cluster node's insights during testing

TBA