thatDot streaming graph was designed for real-time event pipelines. Combining a graph data model that scales to ingest millions of events per second with an actor-based compute model means Quine can find complex patterns in high-volume categorical data.
In addition to incorporating the best properties of graph databases and event stream processing systems, thatDot streaming graph has some important and unique features that offer significant benefits for event pipelines and event-driven applications:
A data model optimized for complex pattern matching - A property graph structure encodes relationships as edges, effectively pre-computing joins. This allows for highly expressive yet high performance N-hop queries.
Parallelized compute for fast writes (ingest and updates) - An asynchronous, actor-based graph computation model provides very high throughput when processing complex events. It also makes possible our killer feature: standing queries.
Out-of-order data handling - Standing queries easily handle out-of-order data and late-arriving data. A standing query is a special query type that persists in the graph, incrementally updating as new data arrives.
No time windows - Query historical data, incoming data, and even use a standing query to watch for matches in the future, all in real-time.
thatDot Streaming Graph eliminates time window tradeoffs. For instance, unlike Flink and ksqlDB, streaming graph is not limited by time windows. In fact, it is designed to find patterns in events spread our over days, months, and even years. Late-arriving and out-of-order data are no problem for Quine.
And if processing high volume event streams is critical to your business, thatDot Streaming Graph provides clustering for both high availability and scale. Process millions of events per second reliably with a clustering system built on the Apache Pekko framework.
Categorical data is both incredibly useful and almost totally ignored in modern event processing applications. Quine allows you to express the relationships between common objects and information as a graph and look for critical patterns.
No other event processing system lets you extract and transform categorical data from event streams in real time and without lossy encoding.
Use categorical data from your real-time event pipeline for financial and authentication fraud detection, XDR/EDR applications, network observability, and recommendation engines.
Streaming Graph incorporates the throughput and scalability of event stream processing systems like Flink and ksqlDB with the rich graph query syntax found in traditional databases like Neo4J and TigerGraph.
Streaming Graph runs inline with existing real-time event pipelines. Use Cypher to extract data from Kafka or Kinesis streams or data warehouses and transform that data into graph nodes, properties, and relationships or edges.
Pluggable storage options include Cassandra for high-volume workloads or extra resiliency matter. When paired with thatDot Streaming Graph's clustering for resiliency and horizontal scaling, you can process millions of events per second without the limitation of time windows.
Standing queries make thatDot Streaming Graph unique, allowing you to monitor event streams for specific patterns of importance to you -- for instance, patterns suggesting fraud, impending system failures, or opportunities to make recommendations that increase revenue -- and take action on them the instant a match is made.
With traditional databases, you must continuously query to see if a match has occurred. This is inefficient when microseconds matter.
Standing queries maintain partial matches over long time intervals, allowing Streaming Graph to handle out-of-order data and eliminate time windows.
Once a standing query match is made, the streaming graph can execute arbitrary code, write to a Kafka topic, update the graph, post an alert message to an API, or whatever else your business logic demands. All in real time.