Google Meet

This section provides the performance summary of the Google Meet service managed through CCKM on the CipherTrust Manager server.

Requests Per Second

The following sections list the wrap requests per second, approximate latency, and the number of virtual users for different deployment scenarios.

Note

To provide a good user experience, Google recommends a maximum latency of 200 ms (for 99% of the requests). Therefore, the performance numbers on this page are based on a latency of around 200 ms.

• Google Cloud

• AWS Cloud

• Physical Appliance

Google Cloud

Simulated the wrap requests for Google Meet on the CipherTrust Manager deployed on Google Cloud Platform using the k6 tool. The following table shows the handled number of requests per second (RPS), with approximate latency, and the number of virtual users for different data samples on a standalone CipherTrust Manager and a two-node CipherTrust Manager cluster connected with a load balancer.

Click a tab to view performance numbers based on two different specifications.

AWS Cloud

Simulated the wrap requests for Google Meet on the CipherTrust Manager deployed on AWS cloud using the k6 tool. The following table shows the handled number of requests per second (RPS), with approximate latency, and the number of virtual users for different data samples on a standalone CipherTrust Manager and a two-node CipherTrust Manager cluster connected with a load balancer.

Click a tab to view performance numbers based on two different specifications.

=== "Standalone"

    Users | Latency (in ms) | Requests/Second
    ---|---| ---
    10 | 40.74 | 16.462869
    20 | 41.405 | 32.643829
    30 | 42.07 | 48.824789
    40 | 42.99 | 64.9318095
    50 | 43.91 | 81.03883
    60 | 45.02 | 96.9078885
    70 | 46.13 | 112.776947
    80 | 50.8 | 128.259757
    90 | 55.47 | 143.742567
    100 | 59.32 | 159.43998
    110 | 63.17 | 175.137393
    120 | 86.17 | 187.6379445
    130 | 109.17 | 200.138496
    140 | 132.855 | 212.1405135
    150 | 156.54 | 224.142531
    160 | 171.395 | 235.5833245
    170 | 186.25 | 247.024118
    180 | 233.91 | 252.1894955

Physical Appliance

CipherTrust Manager Configuration

Simulated the wrap requests for Google Meet on the CipherTrust Manager deployed on a physical appliance using the k6 tool. The following table shows the handled number of requests per second (RPS), with approximate latency, and the number of virtual users for different data sample for three runs.

Comparison Graphs

Physical Appliance vs AWS Cloud vs Google Cloud

Recommendations

Note

When an encrypted Google Meet call is started, a wrap request is initiated. Similarly, when a participant joins the call, an unwrap request is initiated.

Google Cloud

1. Response time compliance of around 200 ms was met for a maximum throughput of 144.66 operations per second with a standalone CipherTrust Manager k170v instance with 4 CPUs and 16 GB RAM.

2. Response time compliance of around 200 ms was met for a maximum throughput of 279.02 operations per second with a two-node CipherTrust Manager k170v cluster (each node with 4 CPUs and 16 GB RAM) connected with a Google Cloud load balancer.

3. Response time compliance of around 200 ms was met for a maximum throughput of 330.899 operations per second with a standalone CipherTrust Manager k470v instance with 8 CPUs and 64 GB RAM.

4. Response time compliance of around 200 ms was met for a maximum throughput of 755.37 operations per second with a two-node CipherTrust Manager k470v cluster (each node with 8 CPUs and 64 GB RAM) connected with a Google Cloud load balancer.

AWS Cloud

1. Response time compliance of around 200 ms was met for a maximum throughput of 252.19 operations per second with a standalone CipherTrust Manager k170v instance with 4 CPUs and 16 GB RAM.

2. Response time compliance of around 200 ms was met for a maximum throughput of 442.50 operations per second with a two-node CipherTrust Manager k170v cluster (each node with 4 CPUs and 16 GB RAM) connected with an AWS load balancer.

3. Response time compliance of around 200 ms was met for a maximum throughput of 559.30 operations per second with a standalone CipherTrust Manager k470v instance with 8 CPUs and 64 GB RAM.

4. Response time compliance of around 200 ms was met for a maximum throughput of 1059.78 operations per second with a two-node CipherTrust Manager k470v cluster (each node with 8 CPUs and 64 GB RAM) connected with an AWS load balancer.

Conclusion

The number of users and throughput almost doubles up on moving from CipherTrust Manager k170v to k470v. Moreover, adding an additional node to the cluster also doubles up the throughput. Overall, a performance gain of 400 percent is achieved by moving from a standalone CipherTrust Manager k170v to a two-node CipherTrust Manager k470v cluster.