Default value for max_rate_per_instance for balancing mode of utilization

[gowtham b.]

Hello Everyone,

  I was wondering how does the attribute max_rate_per_instance work when the balancing mode is Utilization and this attribute is not specified. 

1. Would it use a default value? What is this value?

2. if it does not do use any default value and does not apply anything how I do I simulate this behavior by providing a value? like if I give a really high value like 9999 would it be similar to not specifying it? 

Thanks!

[Digil (Google Cloud Platform Support)]

As per my understanding, the max-rate-per-instance target capacity works along with the RATE balancing mode. However, you could use it along with a possible balancing mode combinations. As per the help center article about 'Utilization balancing mode' , not every load balancer support 'max-rate-per-instance' target capacity with the 'Utilization' balancing mode. 

I am not sure whether there is any default value available for  the 'max-rate-per-instance' target capacity, if it using along with the 'Balancing mode combinations' .  But as per my knowledge, if you would like to receive the optimum result, the 'max-rate-per-instance' must be configured as per the table provided in the 'rate balancing mode'.

[Serdar Yildirim]

Hi,

I really wonder what RPS means in the backend service. I can't find a satisfying explanation on GCP documentation. For example when I check the metrics I see some numbers like 0.5, 0.7 or 3.2 RPS. How a client can make a request like 0.2? If I make one request it should count as 1 but I always see some decimal numbers. What is the calculation methods for RPS or what these numbers mean actually? As I understand if I choice RPS for a backend like 300 it actually doesnt mean a 300 request per second. How should we interpret these metric numbers in stackdriver?

[Pedro Moreno]

Hi,

About RPS, It's the backends maximum number of HTTP requests per second, in which backends[].maxRate is a integer. You may find a better view of these concepts and calculations in the Capacity scaler section from documentation. About interpreting these metrics numbers in HTTP(S) Load Balancing logging and monitoring please review it from documentation, as a few different metrics could be used.