Timeseries graph modeling using TigerGraph

Hi,

I am newbie here and have been struggling to find examples of modeling timeseries graph using TigerGraphStudio. Tried to ask in slack almost 1 week ago no luck. Hope to find help with any similar examples here…

Like airports with many flights at same or different time from source to destination with different delay or transit hops of each flights, I have network traffic data (vertex as device IP and edge as socket connections), so each directed edge is source_IP:source_port → dest_IP:dest_port with delay metrics and timestamp of each event.

How do we model such graph (similar to airports, airlines and flights time tables) using TigerGraphStudio? If there is a way to model such timeseries graph in TigerGraph, how to query the graph so we can see what is all connections/path from given vertex A to remote vertex B between timestamp t1 to t2? and compare delay of each path in given time range?

I thought its typical scenario of graph like to book flights for travel with selected dates considering different time delay and cost, but unfortunately unable to find such modeling examples yet from graph DB so not sure if I missed something…

Appreciate for help and pointer of examples!
eric

1 Like

Hi @eric,

the best way to build timeseries schema using graph database would be to put the date part into a vertex (or multiple vertices, depending on your query preferences!) and to connect the data with date using edges. That way you will be able to make a very fast queries, much faster than filtering out by using where clause.

Please check slides 11 and 12 in this document:
https://drive.google.com/file/d/11YL497fWZwObR1J2qyPb7iIA55sHEiTR/view

Best,
Bruno

1 Like

Thanks a lot @Bruno for this doc. So looks like this approach is splitting “subgraph” snapshots into time range buckets, but seems hard to scale as query on graph can be any time range based on timestamp/epoch (like time series DB) and split into subgraph by time bucket could be lots of duplicate vertex and edges in time bucket.

The problem I am currently facing using TigerGraph is - how to model geo-temporal graph so can make time and geo-based query? i.e each of vertex has geographical location (like city, country) and each vertex and edges also changes in time with its own timestamp (like a vertex/edge exist in time T1 and not exist in time T2). There are numerous use cases like search flights, search uber/lyft rides, search social networks all has time and location based search in graph, but seems unable to find such example yet.

Is there good solution example of time and geo location based graph in TigerGraph where we can apply graph algorithms based on interested time range and geo location range?

Appreciate for help and pointers of examples!
eric

Here is a good example of a time series implementation using IoT sensor readings. Vertices are added to enable queries by day or hour in this case. https://www.youtube.com/watch?v=494lqlLsniE
There is a link to the Colab notebook in the comments that you can follow step-by-step.