# Meeting with Sylvan
###### tags: `Meeting notes`

## Engineering difficultys and tasks
1. Ian and Hector suggested that we run Lotus and Visor and Sentinel as is, then run another database (or just new tables) to cache the queries that we run off Sentinel. My question is how to create caches. What queries are we caching? Fundamentally, we don't know what the better representation of our current models.
2. The more difficult approach is to modify Visor, Ian already mentioned that he can assist us with removing `Tasks` or `Models` from visor to reduce the type of data collected. The problem here is Ian doesn't know what data do we want to collect here.
3. Why does Ian think an 2000-Epochs dump is 27GB? One epoch is around 5KB as indicated by filfox, so 2000 epochs should be around 27(MB).

## Self notes
**Ian:**
<!-- 1. In lotus build, different configurations between different networks. -->
2. Indexer.go manages all the tasks.
Looks at the tipset and setup the tasks.
4. In the tasks package, all tasks are listed.
5. PL working on getting the internal messages that is not saved.
6. Look in the `commands` folder. `daemon`, `watch` are defined. `tasks` are the unit of tasks executed, `storage` and `schema` are infrastructure. `models` define the kind of data produces, data types mapped to the database tables.
7. You can use visor to output to csv.
8. If there is a specific query in mind, Ian can help trace the specific task that generate the raw data.