Sections of the Webinar:
- What are time series?Time series is where the pattern of values tracked over time are the metrics you are looking to track. A single value usually doesn’t matter.What happens when you want to track the humidity in the local region with 10,000 IoT devices taking measurements every 10 seconds over a period of 6 months?This involves dealing with a huge amount of data. You can have tens of thousands or even millions of data points for a single time series.
- How to Model Time Series in RavenDBRavenDB is a document database. See how it works when you want to store time series database as a document. You can absolutely store data points in documents.
- The structure of a time StampA single entity that happens at a single time frame. RavenDB gives you millisecond precision. Tags – you have free text – small set of value expected. Values – IEEE double precision float, 1-32 items.
- Querying over Time Series:First, select the document(s) that match: Then, project the time series using aggregation, time slicing, etc.
- From the Client APIEverything is transactional. In RavenDB you have support for Bulk Insert.
- Complex Server-Side ProcessingYou can do really interesting things on the server side. You can compare two different time series on the same document. Queries are always on a single time series.
- Rolling up DataRavenDB has support for Automatic roll up policies. You get transparent queries across roll ups. Every X time, you can roll up totals for the last Y time. For example, every hour, you can automatically generate tallies for the last six hours, 12 hours, 24 hours. Every day, you can get totals for the last week, month, 365 days. Each range can keep your minimum, maximum, sum, and count values.
- Estimating the Storage Requirement for your Time Series DataThe storage requirements can depend on how predictable your data is. RavenDB does a great job at optimizing the space you have available.
- Time Series in a Distributed EnvironmentRavenDB is multi-master. You can read and write to multiple nodes so you can accept data from many locations to a single time series. You have automatic merge. Conflicts are resolved in favor of the highest value.