4.2 Performance Benchmarking with Kafka Feed

The performance benchmarking test is performed on OCNADD with 100K MPS Kafka feed which includes 55K MPS SCP traffic (27.5K TPS with 2 traffic copy trigger points on SCP), 15K MPS NRF traffic (7.5K TPS with 2 traffic copy trigger points on NRF) and 30K MPS SEPP traffic (15K TPS with 2 traffic copy trigger points on SEPP). The benchmarking results are provided below:

Note:

One Ingress message from a NF is "1" MPS for OCNADD.

Benchmark Test Environment

The test environment specifications are listed below:
  • Ingress Traffic Rate: 100K MPS
  • The latency, resource utilization, and MPS is observed for the period of 3 hrs or more.
  • OCNADD single site deployment with ASM disabled
  • OCNADD Release 23.3.0 deployed over CNE (Bare metal) 22.3.0 version
  • Kafka Replication Factor (RF): 1
  • OCNADD Data Feed: HTTP2/TLS
  • OCNADD Kafka PVC: 400GB

    For more information about PVC requirements, see Kafka PVC Storage Requirements.

  • 3rd Party Application: 1 Perfgo Server with 2 End-points
  • cnDBTier Release 23.3.0
  • NRF Release 23.3.0
  • SCP Release 23.3.0
  • SEPP Release 23.3.0
  • SASL/SSL, enabled between NRF/SCP/SEPP and OCNADD
  • CNCC Release 23.3.0
  • Message Size: 3580 Bytes
  • Environment: CNE (Bare metal)
  • Execution time: 20 Hours

Benchmark Testcase Specifications

The testcase parameters are as follows:
  • OCNADD
    • A feed is configured using the OCNADD Console with aggregation rules set for SCP, NRF and SEPP.
    • Message Ingestion Rate: 100K MPS
    Resource Specifications:

    Table 4-4 OCNADD Resource Specifications

    Services CPU Request Per Pod CPU Limit Per Pod Memory Request Per Pod (Gi) Memory Limit Per Pod (Gi) min Replicas max Replicas Kafka Topic Partitions and Retention
    ocnaddconfiguration 1 1 1 1 1 1  
    ocnaddalarm 1 1 1 1 1 1  
    ocnaddadmin 1 1 1 1 1 1  
    ocnaddhealthmonitoring 1 1 1 1 1 1  
    ocnadduirouter 2 2 1 1 1 1  
    ocnaddscpaggregation 2 3 2 2 2 2

    12(SCP)/

    retention.ms = 300000
    ocnaddnrfaggregation 2 3 2 2 1 1

    6 (NRF)/

    retention.ms = 300000
    ocnaddseppaggregation 2 3 2 2 2 2

    12 (SEPP)/

    retention.ms = 300000
                 

    144 (MAIN)/

    retention.ms = 300000
    ocnaddkafka 5 8 64 64 4 4  
    zookeeper 1 1 2 2 3 3  
    ocnaddgui 2 2 1 1 1 1  
    ocnaddcache 1 1 24 24 2 2  
  • SCP

    SCP Traffic: 55K TPS using two trigger points.

  • NRF

    NRF Traffic: 15K TPS using two trigger points.

  • SEPP

    SEPP Traffic: 30K TPS using two trigger points.

Benchmark Test Results

Traffic Feed Details

Table 4-5 Traffic Feed Details

NF NF Traffic Copy Trigger Points Traffic Rate Duration in hours E2E Traffic Feed Average Latency E2E Traffic Feed Success Rate
OCNADD NA 100K MPS 20 NA

Feed-1: 99.93%

Feed-2: 99.91%
NRF 2 - IGW/EGW (Request and Response) 15K MPS 20   99.94%
SCP 2 (Request Ingress/Request Egress) 55K MPS 20   99.89%
SEPP 2 (PLMN IGW) 30K MPS 20   99.90%

CPU and Memory Utilization

The following table describes the OCNADD CPU and memory utilization:

Table 4-6 OCNADD CPU and Memory Utilization

Microservice/ Container Pod Count CPU Utilization (%) Memory Utilization (%)
ocnaddadminservice 1 0.16% 28.50%
ocnaddalarm 1 0.33% 19.60%
ocnaddcache 1 0.17% 84.90%
ocnaddconfiguration 1 0.10% 32.50%
ocnaddgui 1 0.04% 6.43%
ocnaddhealthmonitoring 1 0.17% 31.10%
ocnaddnrfaggregation 1 91.70% 54.00%
ocnaddscpaggregation 2 97.30% 79.00%
ocnaddseppaggregation 2 88.40% 61.90%
ocnaddthirdpartyconsumer 16 22.30% 80.70%
ocnadduirouter 1 0.07% 27.30%
kafka-broker 4 95.00% 57.20%
zookeeper 3 0.15% 18.50%