6.15.1.2.4. RabbitMQ performance (Client -> Slave-1, Server -> Slave-2), HA queues enabled

This report contains results of Message Queue Performance execution with Oslo.messaging Simulator. Simulator client and simulator server are connected to different slave nodes. RabbitMQ HA queues are enabled.

Simulator is configured with eventlet executor running in 10 threads. The overall number of threads is calculated as multiplication of eventlet threads, number of processes and number of used nodes.

6.15.1.2.4.1. Test Case 1: RPC CALL Throughput Test

6.15.1.2.4.1.1. Message processing

Messages are collected at 3 points: sent - messages sent by the client, received - messages received by the server, round-trip - replies received by the client. Also the number of lost messages is calculated. Sizes of messages is based on the distribution of messages collected on the 100-node cloud.

RPC CALL Message count

threads	sent, msg	received, msg	round-trip, msg	lost, msg
50	286874	286874	286874	0
100	335274	335274	335274	0
250	227804	227830	227804	0
350	238445	238445	238445	0
500	168374	167229	168374	0

6.15.1.2.4.1.2. The throughput, latency and RabbitMQ CPU utilization

The chart shows the throughput, latency and CPU utilization by RabbitMQ server depending on number of concurrent threads.

RPC CALL throughput, latency and RabbitMQ CPU utilization depending on thread count

threads	throughput, msg/sec	latency, ms	RabbitMQ CPU, %
50	2814.8	15.1	1229.4
100	3275.9	29.3	1546.0
250	2253.5	87.7	1981.4
350	2358.5	146.7	1901.1
500	1650.2	238.2	2260.0

6.15.1.2.4.1.3. Detailed RabbitMQ CPU consumption

Thus chart shows statistics on RabbitMQ CPU consumption per nodes.

RabbitMQ nodes CPU consumption during RPC CALL load test

threads	Master total, %	Slave 1 total, %	Slave 2 total, %	Master sys, %	Slave 1 sys, %	Slave 2 sys, %
50	210.9	659.8	358.8	57.5	170.0	90.0
100	201.4	944.7	399.9	50.9	243.7	95.8
250	182.7	1229.5	569.2	45.2	333.2	142.5
350	172.0	1297.6	431.6	44.0	367.9	107.5
500	157.0	1369.8	733.1	39.6	391.3	195.2

6.15.1.2.4.2. Test Case 2: RPC CAST Throughput Test

6.15.1.2.4.2.1. Message processing

Messages are collected at 2 points: sent - messages sent by the client and received - messages received by the server. Also the number of lost messages is calculated. Sizes of messages is based on the distribution of messages collected on the 100-node cloud.

RPC CAST Message count

threads	sent, msg	received, msg	lost, msg
50	481721	481721	0
100	566495	566495	0
250	557957	557957	0
350	420309	420296	13
500	497506	497506	0

6.15.1.2.4.2.2. The throughput, latency and RabbitMQ CPU utilization

The chart shows the throughput, latency and CPU utilization by RabbitMQ server depending on number of concurrent threads.

RPC CAST throughput, latency and RabbitMQ CPU utilization depending on thread count

threads	throughput, msg/sec	latency, ms	RabbitMQ CPU consumption, %
50	4760.2	5.5	896.5
100	5602.8	12.3	1124.2
250	5520.7	12.8	1443.0
350	4157.7	7.5	1448.2
500	4922.2	11.8	1450.0

6.15.1.2.4.2.3. Detailed RabbitMQ CPU consumption

Thus chart shows statistics on RabbitMQ CPU consumption per nodes.

RabbitMQ nodes CPU consumption during RPC CAST load test

threads	Master total, %	Slave 1 total, %	Slave 2 total, %	Master sys, %	Slave 1 sys, %	Slave 2 sys, %
50	177.5	549.1	169.9	53.8	137.0	44.9
100	177.8	727.4	218.9	50.3	187.1	56.1
250	171.7	1061.9	209.3	46.8	292.8	55.7
350	163.9	1074.6	209.6	43.9	302.0	57.1
500	148.8	1050.9	250.3	39.8	297.5	67.8

6.15.1.2.4.3. Test Case 3: Notification Throughput Test

6.15.1.2.4.3.1. Message processing

Messages are collected at 2 points: sent - messages sent by the client and received - messages received by the server. Also the number of lost messages is calculated. Sizes of messages is based on the distribution of messages collected on the 100-node cloud.

NOTIFY Message count

threads	sent, msg	received, msg	lost, msg
50	488810	488810	0
100	481277	481277	0
250	541306	541306	0
350	537552	537552	0
500	483554	483554	0

6.15.1.2.4.3.2. The throughput, latency and RabbitMQ CPU utilization

The chart shows the throughput, latency and CPU utilization by RabbitMQ server depending on number of concurrent threads.

NOTIFY throughput, latency and RabbitMQ CPU utilization depending on thread count

threads	throughput, msg/sec	latency, ms	RabbitMQ CPU consumption, %
50	4830.8	9.7	931.2
100	4761.0	20.4	804.5
250	5356.1	39.0	1249.1
350	5319.3	51.8	1298.8
500	4785.1	38.1	1420.2

6.15.1.2.4.3.3. Detailed RabbitMQ CPU consumption

Thus chart shows statistics on RabbitMQ CPU consumption per nodes.

RabbitMQ nodes CPU consumption during NOTIFY load test

threads	Master total, %	Slave 1 total, %	Slave 2 total, %	Master sys, %	Slave 1 sys, %	Slave 2 sys, %
50	155.6	542.5	233.1	44.1	138.1	61.5
100	154.8	441.1	208.6	43.1	111.6	53.0
250	152.4	848.6	248.1	40.7	226.1	61.3
350	150.5	902.3	246.0	40.4	240.5	62.5
500	156.4	1020.9	242.9	41.9	287.9	61.0