Service initialisation times are extremely high with nested caches

[alfmarcua]

Context

When executing the mapproxy-util serve-develop {conf-file}.yaml command, the startup time is increased by the number of caches that have other caches as a source and also by the number of caches that are part of the source.

This issue is similar to #70 but in this case the problem relies on nested caches.

Expected Behavior

Execution time should not be increased by containing nested caches.

Actual Behavior

Extensive analyses have been performed comparing mapproxy service startup times as a function of two factors:

1. Number of caches that contain caches as sources
2. Number of subcaches that make up those sources.

The results are as follows:

Layers	Time	Config file
200 regular layers	1.5 secs	200_0_0.yaml
200 regular layers + 100 layers with 1 nested cache	8 secs	200_100_1.yaml
200 regular layers + 100 layers with 5 nested caches	36 secs	200_100_5.yaml
200 regular layers + 100 layers with 10 nested caches	1 min 11 secs	200_100_10.yaml
200 regular layers + 200 layers with 1 nested cache	15 secs	200_200_1.yaml
200 regular layers + 200 layers with 5 nested caches	1 min 14 secs	200_200_5.yaml
200 regular layers + 200 layers with 10 nested caches	2 min 22 secs	200_200_10.yaml

Steps to Reproduce

1. Use the attached configuration files
2. Execute the command "mapproxy-util serve-develop {conf-file}.yaml" for each conf-file
3. Observe the log and check startup time

Context

The goal is that the startup time is not affected by the number of caches containing other caches as a source, since being proportional to both the number of such caches and nested caches on each cache, the multiplicative factor can easily become very high.

Your Environment

• Version used: MapProxy 1.16
• Environment name and version: Python 3.10
• Server type and version:
• Operating System and version: Ubuntu