The MALLOC behavior trigger a memory allocation behavior on the tron, and this memory is held permanently. This can be used to simulate a memory leak. The header value must be use to control how much memory should be allocated.
The value for the malloc behavior has the following format: [MIN,MAX]
The MIN is an integer value representing the number of killobytes to allocate at the minimum.
The MAX is an integer value representing the number of killobytes to allocate at the maximum.
The tron picks a random value between MIN and MAX (inclusive). The random follows a normal distribution (bell curve), using the Box-Muller transform.
This causes a memory leak on the tron with an id of javatron1 for about 1MB
curl -i "http://localhost:8081/api/inventory/2" -H "x-demo-malloc-pre-javatron1, [800,1200]"This causes a memory leak on all trons for about 50KB (interval of 40KB and 60KB)
curl -i "http://localhost:8081/api/inventory/2" -H "x-demo-malloc-post-javatron1, [40,60]"While each tron does the same amount of memory allocation, each language has a different amount of overhead inherent to each object created. Here are some measurements observed for the various languages:
- Java, input of 60KB/min, measurement of 125KB/min, overhead ~ *2.08
- Node, input of 10MB/min, measurement of 85MB/min, overhead ~ *8
- Python, input of 2.5MB/min, measurement of 40MB/min, overhead ~ *16