- summarize
  - set summary has the problem that we don't close our figures and they will be reopened and the same allocators will be drawn multiple times.

- randomization
  - env randomization *easy*
  - benchmark order *relatively easy* warmup?
  - heap randomization *not possible* because we want to benchmark different allocators with deterministic workloads
  - other segments *difficult* temci build?

- fix mysqld termination with malt
- add mysql warmup in preallocator_hook

- rethink colors

- check dependencies

- implement requirements for allocators

- convert cmake build systems to make: espresso, cfrac

- merge measurements if facts match

- possible benchmarks:
   - https://github.com/GraphChi/graphchi-cpp
   - http://glaros.dtc.umn.edu/gkhome/metis/metis/overview

- TESTS!!