Currently, the size of the scratch region is set when a snapshot is first created from a binary, and it can never be changed. This is unfortunate, because many workloads use a lot of memory during an early initialisation phase, and then very little memory in repeated runtime phases that occur after another snapshot. It would be nice to offer some API that changed the size of the scratch region on a snapshot so that we could avoid this situation, since the size of the scratch region is directly linked to performance, especially on certain kernel/hypervisor combinations.
Currently, the size of the scratch region is set when a snapshot is first created from a binary, and it can never be changed. This is unfortunate, because many workloads use a lot of memory during an early initialisation phase, and then very little memory in repeated runtime phases that occur after another snapshot. It would be nice to offer some API that changed the size of the scratch region on a snapshot so that we could avoid this situation, since the size of the scratch region is directly linked to performance, especially on certain kernel/hypervisor combinations.