Meeting notes 8/06/20 Strategy+Telemetry Sync

Attendees:

@Former user (Deleted) @Jess Douglas Muir (Deactivated) @Avery Chiu @Umar-Farooq Shahid Sheikh (Deactivated)

Questions for strategy

• Data storage requirements for strategy

• Data access requirements for strategy

• What’s shared between strategy and telemetry storage

• Anyone know anything about battery box?

Questions from strategy

• What’s the permanent solution?

• How does it look while racing?

• How does it work for future cars?

Napkin math for data ingress

• 1.2 kb from 150 fields, 8 bytes per field

• Send message every 5 seconds, 16 hours a day

• Equals 14 MB per day

• Given 5 GB of storage, we’re ok for 357 days

Napkin math for data egress

• 10-15 seconds call period x 3 fields

• Clarke data egress TBD

• For 8-10 hours of racing, ~2000 calls

• 2000 * 10 b = 20 kb per day = “more than safe” - emma wai

Database research: (free tiers only)

MongoDB

• 512 MB

• 10 GB in/out per week

Firestore

• 1 GB

• 50,000 reads per day

• 20,000 writes per day

• Network egress 10 GB per month

GCP Cloud storage

• 5 GB-months

• 5,000 writes / month

  • Batched writes every 10 mins?

• 50,000 reads / month

• 1 GB egress per month

DynamoDB

• 25 GB storage

• 2.5 million stream read requests

• 1 GB data transfer out / month

FRED Influx DB

• 50 mb size

• 2 day retention

Cosmo DB (Azure)

• 5 GB

• 400 ops/sec (reads, writes, updates, queries)

Decision:

• We’ll go with AWS DynamoDB for best future-proofing and for integration with backend for clients.

Strategy usage:

• Models running through flask server somewhere

  • Always re-evaluating, sending to telemetry / strategy clients

• Models run in preparation

  • Only run a few times, before the race. Don’t need to run them during the race.

• Clarke’s web app?

• Weather modeling?