this is for holding javascript data
Brandon Holt edited outline.md
about 9 years ago
Commit id: 20f50c0d489818ef3fe5a2ea2b6c4e06229414f3
deletions | additions
diff --git a/outline.md b/outline.md
index 49cbf98..bacf9e1 100644
--- a/outline.md
+++ b/outline.md
...
- Evaluation / Case Study
- *maybe pitch this as a case study? it's certainly not going to feel like much of an "evaluation"*
- explain Retwis application, txns:
- `new_user:5%`, `follow:10%`, `unfollow:5%`, `post:30%`, `read_timeline:50%`
- uniform vs powerlaw graph
- uniform vs zipfian workload generation
- evaluate with commutativity "on" and "off"
- *evaluate with multiple levels of commutativity? something like comm. lattice eval?*
- *evaluate against non-transactional?*
- could just run the same thing without transactions or with no aborting txns as a proxy
- but should we quantify the number of resulting conflicts/inconsistencies somehow?
- *evaluate against eventual consistency somehow?*
- Future work
- other potential uses of ADTs and commutativity
- batching and combining
- contention/hot-spot avoidance via splitting (phase reconciliation)
- replication/fault tolerance
## Questions:
- where to work in contention / naturally skewed workload? use as motivation? or just in case study?