Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Add_workflow_for_techpubs
AUTH2UserResourceIdentifier{userKey=8a00a02355cd1c2f0155cd26cb8305e9, userName='null'}
AUTH1UserResourceIdentifier{userKey=8a00a02355cd1c2f0155cd26cd5909df, userName='null'}
JIRAIDAUTHSBX-91521
REV5UserResourceIdentifier{userKey=8a00a02355cd1c2f0155cd26cd5909df, userName='null'}
REV6UserResourceIdentifier{userKey=8a00a02355cd1c2f0155cd26cd5909df, userName='null'}
REV3UserResourceIdentifier{userKey=8a00a02355cd1c2f0155cd26ca500415, userName='null'}
REV4REV1

internal_display_only

UserResourceIdentifier{userKey=8a00a02355cd1c2f0155cd26ca7c0455, userName='null'}
REV1UserResourceIdentifier{userKey=8a00a0c86268f97a016270f846210010, userName='null'}



Panel

In this section:

Table of Contents
maxLevel3


AWS Benchmarking Setup


Caption
0Figure
1AWS Benchmarking Setup
3AWS Benchmarking Setup

Benchmarking Setup Details

  1. Benchmarking is performed on an SBC HA with HFE.
    1. The call flow includes 9 signaling messages with media.
    2. Calls are tested from Public to Private network
  2. The SBC HA with HFE and PKTART are running on the default tenant.
  3. The PKTART Tool flavor uses c5.2xlarge to benchmark all of the SBC with HFE Instances except c5.4xlarge.
    1. For the c5.4xlarge Instance, the PKTART flavor used is the same as the SBC.
  4. The performance test is done for one hour.
  5. Perflogger is used for Logs collection.
  6. There is a 2,000 sessions System Limit for M5.xlarge for default profile. 
  7. TLS/SRTP enabled at Public side traffic. Separate TLS connection (Port) used for each call. 
  8. SRTP Calls are executed from PKTART.

Instance Types

S. NO.Instance TypeIOPSRAM (MB)CPU

HDD (GB)

1M5.xlarge50016384465
2M5.2xlarge50032768865
3C5.2xlarge50016384865
4C5.4xlarge500327681665
5P3.2xlarge50062464865


Hide

Columns to remove from SVT table when importing to this page:

  • Alias
  • Total Bandwidth
  • Log Path
  • Average Packet Loss Public/Private
  • Observations

Reword:

  • Max Memory Utilization in GB > Max Memory Utilization %
  • Swe_Profile & Limiting Factor > SWe Profile


CPU Instances

Div
classpdf6pttext


Call ScenarioInstance TypeCPS/CHTSimultaneous
Sessions
Worker CoreSignaling CoreUXPAD CoreAverage CPU
Utilization

Max Memory
Utilization %

SWe Profile

wrkr core#CPU%Sig core #CPU%UXPAD Core #CPU%
G711u-UDP-passthrough

M5.xlarge

20/100 2000Core 215.17%Core 115.06%Core 30.56%10.32% 43.22%

Default


G729A-UDP-passthrough

M5.xlarge

20/100 2000Core 218.83%Core 114.35%Core 30.55%9.70%43.26%

Default


G711u-Passthrough-TLS-SRTP enabled

M5.xlarge

20/1002000

Core 2

40.84%

Core 1 (shared)

Core 3


22.93%

25.09%

NANA22.61%42.49%

Standard Passthrough Profile


G711u-OPUS transcoded

M5.xlarge

2/4896Core 23.94%Core 110.65%Core 339.49%15.07%38.81%Standard Transcoding Profile
G711-UDP passthrough

M5.2xlarge

40/1506000

Core 2

Core 3

32.32%

32.14%


Core 1 (Shared)

Core 6

Core 7

24.71%

22.99%

24.03%

Core 4

Core 5

0.39%

0.70%

18.06%28.77%

Default


G729A-UDP passthrough

M5.2xlarge

45/1506750

Core 2

Core 3

37.36%

36.38%


Core 1 (Shared)

Core 6

Core 7

27.30%

25.01%

26.82%

Core 4

Core 5

0.42%

0.84%

20.16%29.25%

Default


G729A-UDP passthrough

c5.2xlarge

40/1506000

Core 2

Core 3


32.93%

31.36%

Core 1 (Shared)

Core 6

Core 7

23.78%

22.30%

22.89%

Core 4

Core 5

0.33%

0.63%

17.52%55.69%

Default


G711-Passthrough TLS-SRTP enabled

c5.2xlarge

30/2006000

Core 2

Core 3

47.48%

52.18%


Core 1 (Shared)

Core 4

Core 5

Core 6

Core 7

10.18%

12.81%

10.05%

11.67%

11.77%

NANA20.40%55.38%

Standard Passthrough Profile


G711-OPUS

c5.2xlarge

7/60420Core 28.47%Core 116.63%

Core 3

Core 4

Core 5

Core 6

35.09%

38.19%

39.25%

38.15%

22.89%46.42%Standard Transcoding Profile
G711-UDP pass through

c5.2xlarge


40/1506000

Core 2

Core 3

30.31%

31.14%

Core 1 (Shared)

Core 6

Core 7

24.61%

23.37%

24.04%

Core 4

Core 5

0.34%

0.61%

17.59%55.87%

Default




G729-Passthrough-SRTP enabled

c5.2xlarge


30/1504500

Core 2

Core 3


47.92%

34.95%

Core 1 (Shared)

Core 4

Core 5

Core 6

Core 7

13.33%

15.64%

12.49%

13.89%

14.19%

NANA20.91%53.78%Standard Passthrough Profile
g711-g729-Transcoded

c5.2xLarge


10/1281280

Core 2


20.82%

Core 1

21.00%

Core 3

Core 4

Core 5

Core 6

71.82%

74.74%

81.97%

79.92%

45.63%47.86%

Standard Transcoding Profile

g729-Passthrough

c5.4xlarge

40/1506000

Core 4

Core 5

Core 6

Core 7

12.44%

13.30%

13.21%

13.33%

Core 1 (Shared)

Core 12

Core 13

Core 14

Core 15

15.04%

15.09%

14.02%

14.25%

14.44%

Core 8

Core 9

Core 10

Core 11

0.24%

0.33%

0.27%

0.47%

8.92%31.86%

Default


g711-Passthrough

c5.4xlarge

40/1506000

Core 4

Core 5

Core 6

Core 7

10.03%

09.92%

10.13%

10.07%

core1 (Shared)

Core 12

Core 13

Core 14

Core 15

13.86%

14.16%

12.95%

13.20%

13.29%

Core 8

Core 9

Core 10

Core 11

0.23%

0.32%

0.25%

0.47%

8.74%32.65%

Default


g711-passthrough-SRTP enabled

c5.4xlarge

30/1504500

Core 4

Core 5

Core 6

Core 7


12.81%

12.71%

12.70%

12.71%


Core 1 (Shared)

Core 12

Core 13

Core 14

Core 15

12.85%

12.75%

11.65%

11.86%

12.09


Core 8

Core 9

Core 10

Core 11

0.43%

0.31%

0.15%

0.66%

8.15%31.15%

Default


g711-g729A udp- transcoded

c5.4xlarge



10/1501500Core 221.56%

Core 1 (Shared)

Core 15

17.03%


14.51%

Core 3

Core 4

Core 5

Core 6

Core 7

Core 8

Core 9

Core 10

Core 11

Core 12

Core 13

33.54%

32.55%

33.48%

28.102%

31.12%

29.82%

31.28%

31.80%

33.82%

31.37%

32.76%

25.88%28.44%

Standard Transcoding Profile


opus-20ms-g711u-20ms

c5.4xLarge


3/142426Core 28.18%

Core 1 (Shared)

Core 15


9.38%


8.42%

Core 3

Core 4

Core 5

Core 6

Core 7

Core 8

Core 9

Core 10

Core 11

Core 12

Core 13

12.38%

13.60%

14.37%

16.56%

14.68%

19.82%

16.59%

14.76%

15.05%

13.43%

12.60%

12.75%27.89%Standard Transcoding Profile



GPU Instances

Div
classpdf6pttext


Call ScenarioInstance Type

HFE Instance 
Type

CPS/CHTSimultaneous
Sessions
Worker CoreSignaling CoreAverage CPU
Utilization

Max Memory
Utilization %

SWe Profile


DSP Utilization
wrkr core#CPU%Sig core #CPU%
G711u-G729a transcodingp3.2xlargeM5.2xlarge38/1003800Core 255.11%

Core 1 (Shared)

Core 3

Core 4

Core 5

Core 6

Core 7

47.72%

50.47%

42.79%

41.99%

28.08%

33.57%

38.51%14.66%

Standard GPU I-SBC Profile

82%
G711u-G722 transcodingp3.2xlargeM5.2xlarge 36/1003600Core 246.17%

Core 1 (Shared)

Core 3

Core 4

Core 5

Core 6

Core 7

44.09%

46.62%

38.45%

34.60%

23.62%

31.68%

34.20%14.86%

Standard GPU I-SBC Profile

78%


Pagebreak

AWS SLB Performance Data

Deployment Types on AWS with SLB

  1. HFE/HA SBC behind HFE/HA SLB
  2. HFE/HA SBC behind HA SLB

AWS with SLB Performance Testbed

The diagram below depicts the HFE front-ending the SLB and SBC.

  • Pktarts are used as testing the SLB performance as UAC and UAS
  • 600pxFor Standalone SLB/SBC, replace HFE-SLB and HFE-SBC with standalone SLB and SBC
Caption
0Figure
1AWS with SLB Performance Testbed

Pagebreak

SLB Performance Data

Div
classpdf6pttext


SLB
Instance Type

SLB 
Sessions

SLB 
CPS/CHT

SLB
Average CPU Utilization %

SLB
Max Memory Utilization %

SLB
SWe Profile

SBC
Instance Type

Number
of SBCs

Sessions/SBC


SBC
Average  CPU Utilization %

SBC
Max Memory Utilization %

SBC
SWe Profile

M5.xlarge (HFE)

(IPv4 UDP)

50K500/10011.99%37.04%Standard SLB ProfileM5.2xlarge
(SA)
7

7K

8.56%29.3%Default
M5.xlarge (HFE) UDP50K

500/100

15.22%36.96%Standard SLB ProfileM5.2xlarge
(HFE HA)
68.3K14.67%31.04%Default

C5.2xlarge (HFE)

UDP

50K500/10010.18%42.6%Standard SLB ProfileC5.2xlarge
(SA)
68.3K17.04%62.5%Default

C5.2xlarge (HFE)

UDP

50K500/1009.76%42.8%Standard SLB ProfileC5.2xlarge
(HFE HA)
68.3K14.35%61.8%Default
C5.2xlarge (HFE) UDP36K500/7211.98%40.71%Standard SLB ProfileC5.2xlarge
(HFE and SA)
4
(2HFE and 2 SA instances)
9K20.65%62.4%Default

C5.2xlarge (HFE)

TLS

50K500/10010.26%41.9%Standard SLB ProfileC5.2xlarge
(HFE and SA)
6
(3 HFE and 3 SA instances)
8.3K19.29%63.5%Default



Info
titleNote

Test parameters:

  • Tested using two SLB HFE EIPs for two different clients each. SLB had two backend SBCs
  • Tested ~4,000 session load with media enabled on m5.xlarge SLB and SBC setups


Pagebreak