O slideshow foi denunciado.
Seu SlideShare está sendo baixado. ×

How a Time Series Database Contributes to a Decentralized Cloud Object Storage Platform

Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio
Anúncio

Confira estes a seguir

1 de 40 Anúncio

How a Time Series Database Contributes to a Decentralized Cloud Object Storage Platform

Baixar para ler offline

In this presentation, you'll learn how InfluxDB is a component to Storj’s Tardigrade service and workflows. John Gleeson and Ben Sirb of Storj Lab will Storj’s redefinition of a cloud object storage network, how InfluxData fits into Storj’s Open Source Partner Program, and how to collect and manage high-volume, real-time telemetry data from a distributed network.

In this presentation, you'll learn how InfluxDB is a component to Storj’s Tardigrade service and workflows. John Gleeson and Ben Sirb of Storj Lab will Storj’s redefinition of a cloud object storage network, how InfluxData fits into Storj’s Open Source Partner Program, and how to collect and manage high-volume, real-time telemetry data from a distributed network.

Anúncio
Anúncio

Mais Conteúdo rRelacionado

Diapositivos para si (20)

Semelhante a How a Time Series Database Contributes to a Decentralized Cloud Object Storage Platform (20)

Anúncio

Mais de InfluxData (20)

Mais recentes (20)

Anúncio

How a Time Series Database Contributes to a Decentralized Cloud Object Storage Platform

  1. 1. How a Time Series Database Contributes to a Decentralized Cloud Object Storage Platform InfluxData Time Series Webinar February, 25th 2020
  2. 2. Our Presenters John Gleeson | VP Operations @ Storj Ben Sirb | Senior Data Scientist @ Storj
  3. 3. Overview A brief primer on Storj Business need for time series data We succeed or fail based on the data What success looks like Q&A Our analytics approach
  4. 4. A Primer on Storj
  5. 5. Our Goal To create the world’s largest and most secure, resilient, performant, and economical cloud storage service - without owning or operating a data center. Our Goal
  6. 6. Storj: A Decentralized Cloud Storage Network
  7. 7. How it works System ViewHow it Works | System View
  8. 8. Commitment to Open Source If the dominant model for monetizing open source becomes “using open source as a loss leader for infrastructure,” and… Only a few large companies have the scale to operate centralized clouds and traditional infrastructure, and... Those same large companies also get a disproportionate amount of traffic, data, talent, economic returns…. ...how will we drive economic growth and innovation from new open source projects?
  9. 9. An Answer: Decentralized Infrastructure Fundamentally different economic model for delivering infrastructure Creates new model for monetizing open source
  10. 10. The Business Need
  11. 11. Need for time series data Two-sided market: Supply + Demand Durability: Function of Storage Node attrition over time Retrievability: Function of SNO uptime at any point in time Repair: Expensive business function to ensure Durability and Retrievability
  12. 12. Our GoalMathematics behind functions is data intense Storj V3 Whitepaper
  13. 13. Our GoalIf You Can’t Measure It You Can’t Improve It ● We have a list of key metrics that directly impacts roadmap progression - Qualification Gates. ● Exposing network behavior to the entire organization lets us move faster ○ Everyone on the team needs access to quality data to do their job successfully ● And some of these things can make or break the business
  14. 14. Succeed or fail based on the data
  15. 15. Our GoalBoiled it All Down to a Simple ETL Basic Processing and Serialization (We call it Statreceiver) InfluxDB Redash Monkit Collection
  16. 16. Bulk Data. Done!
  17. 17. Our GoalFocus on Optimizing Design ● The faster we start generating key data the faster we improve ● Don’t have resources for a lot of maintenance overhead ● Needs flexibility as the network grows and data needs change. ● Assume that non-Storj folks will use and run at some point ● Last but not least: Needs to Scale with Network
  18. 18. The Approach
  19. 19. Our Goalmonkit: Metrics, Monitoring, and More https://github.com/spacemonkeygo/monkit ● Captures event data (meters, rates, counts etc) right in the code ● Collected directly on network (satellites, nodes, and uplinks). Sent to separate service for processing.
  20. 20. Our Goal ● With simple config, takes UDP packets in and sends time series data out to InfluxDB via graphite protocol ● Some basic filtering and processing in between Statreceiver
  21. 21. Our GoalInfluxDB Supports Graphite Protocol Graphite Protocol Time Series Filter + template one Filter + template two Filter + template three InfluxDB Measurement InfluxDB Measurement InfluxDB Measurement Allowed us to build useful additional structure.
  22. 22. Our GoalInfluxDb - Redash BI Super easy to learn if already familiar with SQL.
  23. 23. The Challenge: Managing Series Cardinality
  24. 24. Framing the project Save All Data Goal: Maximum Visibility Result: Series Explosion Database Bloat
  25. 25. Framing the project ● Launched with goal of maximum visibility ● Which meant saving all the data ● Which led to massive database bloat ● Series explosion ● Start filtering out certain measurements to prevent massive DB size ● ID’s as tags rather than fields
  26. 26. Succeeding Based on the Data
  27. 27. What Influx lets us do Enabling segment decay forecast essential to repair forecast Monitoring growth rate Control complexity - number of objects on the network Tracking bandwidth usage for gets, puts, repairs audits - critical in bandwidth constrained environment Tracking vetting success rate over time - indication of health of network growth rate
  28. 28. Our GoalState of the Network: Data Driven Guidance ● Internal dashboards for the benefit of entire organization ● Root cause analysis of any anomalies ● Review current trajectory for key metrics
  29. 29. Our GoalMeasuring Model Parameters in Real Time
  30. 30. The latest and greatest With the release of monkit v3 we eliminated the need for the graphite layer monkit v3 can natively output measurements with appropriate tags and names No more need to “lock” measurements (before locking, code changes resulted in broken measurements; not with monkit v3)
  31. 31. Future Roadmap ● Full TICK Stack? ● Integrate statistical modeling directly into architecture ● Decentralized network eventually needs decentralized analytics
  32. 32. Open Source Partner Program 1.) The people who bring supply to the network (“SNOs”) should be fairly and transparently compensated---otherwise they will leave the network 2.) The people who bring demand to the network should be fairly and transparently compensated--so they help to build the network ...If open source is the biggest driver of cloud usage, why not let open source projects get paid by the network for helping it to grow?
  33. 33. How does the program work? Delivering significant benefits for open source projects and their users. Customers get secure, private, reliable cloud storage that is half the price of the big cloud providers and drive revenue to open source projects they love. Open source projects get a meaningful share of subscription revenue when their end users store data on the Tardigrade platform.
  34. 34. Network generates revenue for OSS Network benefits from increased demand OSS generates demand for network OSS innovates Virtuous Cycle
  35. 35. Are you an OSS project that generates demand for S3-compatible storage? Build a connector that gives users option to store on the network Network tracks usage, and returns meaningful portion of revenue that your users generate to you Sign up now (storj.io/partners) Storj Partner Program
  36. 36. Questions?
  37. 37. Join our community at forum.storj.io
  38. 38. Thank you!
  39. 39. © 2019 InfluxData. All rights reserved.40 We are looking for great speakers for InfluxDays London 2020. If you would like to speak, please submit your speaker application. 8-10 June, 2020 etc.venues St. Paul’s / London, UK Register now using code WEBINAR and save 20% www.influxdays.com

×