Power your mission with efficient data management and retrieval without high cloud costs.
Benefits
Minimize cloud storage costs
You shouldn’t face a tradeoff between storing data and paying exorbitant cloud bills. Complex hardware requires retention of high frequency data at full fidelity for years. Sift’s storage is object-based, empowering users to selectively keep essential data with precise granular retention and downsampling policies. This ensures efficient data management that allows you to store all the data you need indefinitely — without excessive cloud costs.
Optimized for massive queries
Managing fleets of machines often entails sifting through large volumes of data across multiple telemetry channels, overwhelming most traditional backend storage tools. Sift gives you a fully optimized storage layer that’s built to support advanced exploratory analysis and searching long duration, high cardinality data. Sift overcomes the limitations of traditional options with efficient querying and retrieval that doesn’t compromise on performance.
Integration ready
Telemetry data is valuable and should power all of your team’s operations, and your storage layer should support multiple hardware data types. Sift stores telemetry in an open source format and provides user-friendly APIs to ensure that each customer can hook up their internal tools and algorithms to compressed, high-performance single source of truth.
Features
Request dozens of telemetry channels without performance degradation
Tune storage costs by deciding what telemetry to retain and for how long.
Connect other tools to Sift’s backend to train your AI/ML models or feed your existing workflows
Send Sift primitives, as well as strings, enums, and bitfields.
Sift Observability Platform
Missions are complex. Reviewing machine data should be simple.
Ingestion
Scale both the number and complexity of your machines without worrying about ingestion rates.
Storage
Query a backend that's purpose-built for long duration, high cardinality searches. Add granular retention policies to save only the data you need.
Visualization
Explore, share, and collaborate with your team — no Python scripts or SQL queries needed. Unify all your data sources in a single window.
Review
Automatically review simulation, testbed, and operational data. Ingrain your expertise in rules that flag anomalies real-time.
Reporting
With one-click, generate certification reports for stakeholders. Provide precise access to your data with Role-Based Access Controls.
FAQs
The data catalog is stored in PostgreSQL and contains metadata like runs, assets, and channels. Data is persisted to Parquet in batches every few minutes, and real time data is written to the cache for immediate availability. Parquet files are compacted into their optimal size on a recurring basis and as data reaches the end of its retention period.
We currently support 10 different data types. Standard ints, floats, and bools, but also strings and enums.
Our goal is to meet customers where they are. If you want to be deployed to a certain cloud environment or have an on-premises deployment, that’s something we can support.