How Bevi Uses InfluxDB and Grafana to Improve Predictive Maintenance and Reduce Carbon Footprint
1 like257 views
Bevi leverages InfluxDB and Grafana to enhance predictive maintenance and reduce its carbon footprint in the beverage industry, which faces significant environmental challenges. By utilizing a distributed beverage production system, Bevi not only cuts costs for customers but also promotes sustainability by minimizing single-use plastic bottles. The company successfully navigated technical challenges with InfluxDB 2.0 and has improved its performance through strategic event handling and batching.
How Bevi Uses InfluxDB and Grafana to Improve Predictive Maintenance and Reduce Carbon Footprint
- 1. How Bevi Uses InfluxDB and Grafana to Improve Predictive Maintenance and Reduce Carbon Footprint Spencer Gagnon Lead Software Engineer August 2023
- 2. Bevi is making the beverage industry sustainable
- 3. ©2023 Bevi Confidential The Beverage Industry Today • >100 billion plastic bottles are produced each year • Less than 1/3rd of bottles are recycled • Nearly as much energy is used to transport beverage bottles as to produce them
- 4. ©2023 Bevi Confidential Most bottled beverages are 97-100% filtered tap water
- 5. ©2023 Bevi Confidential Centralized Beverage Production Beverage Ingredients Packaging Materials Tap Water Bottling Plant Wholesale Facility Retail Location Cost $0.15 $0.20 $0.55 $1.20
- 6. ©2023 Bevi Confidential Distributed Beverage Production Beverage Ingredients Bevi Tap Water Cost $0.15 $0.20
- 7. Residential Commercial Water Only Sparkling & flavored water ? First Target Market: Offices
- 8. ©2023 Bevi Confidential Product Line-Up Machines Up to 8 flavors & enhancements, multiple carbonation levels, multiple temperatures Flavors & Enhancements 14 natural flavors 3 natural enhancements Software Cloud logistics platform Touchscreen app Machine sensors & controls
- 9. Software powers the Bevi experience ● Inventory tracking and forecasting of concentrates, CO2, and filters to enable just in time service ● Remote troubleshooting capabilities ● Touchless dispensing and individual user tracking via web app
- 10. ©2023 Bevi Confidential Benefits to Commercial Customers Time Savings No one has to spend time stocking refrigerators with bottles and cans. Personalization We make it fun to stay hydrated by letting users mix & match high-quality ingredients. Environmental Responsibility We help customers avoid thousands of single-use bottles and cans. Affordability We cut customers’ beverage budgets by more than 50%.
- 11. ©2023 Bevi Confidential Hospitality Technology Pharma / Biotech Property Management Financial Services Some of our 5,000+ Customers Healthcare
- 12. A Brief History 2014: Initial InfluxDB usage - version 0.8 2017: First InfluxDB x Bevi webinar (6 years ago!) 2021 Q1: Upgraded to InfluxDB 2.0 OSS 2021 Q4: Began batching InfluxDB persistence w/ Redpanda + Telegraf
- 18. By The Numbers Bevi Machines send Event POST requests: ~150-200/second Events: ~2500/second All of which flow through Redpanda + Telegraf before ending up in InfluxDB
- 21. Challenges Problem Bevi Events are truly untyped “schemaless” key value pairs. New fields are frequently added in the Dispense App InfluxDB requires types (number, boolean) to be declared in advance for query aggregation and visualization via grafana, and will fail to persist wrong-type columns in the same shard Solution Schema for all measurements/fields is defined by our backend server, non-matching fields are persisted in _overflow_data field as raw key/value pairs Prevents accidental persisting of improper types, enforces well defined schema for queries
- 22. Challenges Problem InfluxDB 2.0 does not handle high throughput small batch writes well Thousands of Bevi machines sending events regularly (around 200 requests/second with only 10 rows per request) and persisting to InfluxDB synchronously was causing performance issues, thread overuse, and unexpected crashes Solution Publish events to Redpanda in sync with Bevi requests, Telegraf subscribes and batches events from all Bevi machines Resulted in reduced load on Influx, reduced thread usage on the Bevi backend, faster response times for Bevis, and no more crashes Increased trust and stronger relationship with the InfluxDB support team
- 23. Looking Forward No Complaints! Bevi + InfluxDB 2.0 have been running without issue since late 2021 Potential Improvements InfluxDB 3.0 and Managed options Data lifecycle configuration: Currently storing all event data for all Bevis since 2014 in InfluxDB, most is unused and will never be used
- 24. Q + A