How to name objects

When we have lots of accounts, experiments, production systems running at the same time, we need to standardize naming conventions.

This document specifies some standard nomenclature for the objects involved in deployment of a system, e.g.,

Airflow DAG names
Git branch names
Docker image names
Directory names
Notebook names
Telegram channel names
Invoke task definition
The name of each object needs to have reference to its important metadata
The schema and the values of each field should be listed and explained in this markdown
Note that we prefer to spell pre_prod rather than preprod, since it's more readable

Airflow DAGs

The schema is {stage}.{location}.{workload}.{subschema} where:
stage = {test, pre_prod, prod}
location = {tokyo, stockholm, ...}
- TODO(gp): It would be better to switch to the AWS region name
workload can be:
- scheduled_trading: a system trading actual capital
- TODO(gp): We should rename it to something better, e.g., real_trading. "Scheduled trading" is a bad name since we could trade not on a schedule with real money. "Prod trading" is a bad name since prod is a stage
- shadow_trading: a system running a model but saving trades, instead of sending them to the exchange
- broker_experiment: an experiment running the broker interface to collect information about market impact
- system_reconciliation
- scheduled_trading_system_observer: publish notebooks every N minutes to track the performance of a scheduled_trading or shadow_trading system
- TODO(gp): Maybe rename internal_monitoring?
- external_monitoring: publish notebooks every N minutes to represent the performance of a scheduled_trading or shadow_trading system for investors
subschema depends on the workload
- E.g., for *_trading can be {model}.{config}.[optional tag]
The path of the DAG in the source tree and in the deployment directory has the same name as the DAG itself

Branch names should follow the schema {stage}.{workload}.{model}.{config}.{tag}
E.g., pre_prod.shadow_trading.C11a.config1
If one branch is used for several values in a field (e.g., for all the workloads) we use all for the value of that field or just skip it
E.g., a Git branch used for all the models in pre_prod scheduled_trading is called pre_prod.scheduled_trading
We can add as tag a date in the format YYYYMMDD

The name should be cmamp.{stage}.{location}.{workload}
TODO(gp): Right now is cmamp-test-tokyo-full-system ```

i docker_create_candidate_image --task-definition "cmamp-test-tokyo-full-system" --user-tag "grisha" --region "ap-northeast-1" ```

The schema is {stage}.{location}.{workload}.{model}.{config}.{tag}
E.g., prod.tokyo.system_observer.C11a.config1.last_5minutes
Current links are like http://172.30.2.44/system_reconciliation/C11a.config1.prod.last_5minutes.html