A Beginners Guide to Building a RAG App Using Open Source Milvus
Processing Real-Time Volcano Seismic Measurements Through Redis: David Chaves
1. PRESENTED BY
Processing Real-time Volcano
Seismic Measurements Through
Redis
David Chaves and Elzbieta Malinowski
Dept. Computer Science, University of Costa Rica
2. PRESENTED BY
1 Background
A short presentation about this work context
2 Scientific measurements using Redis
An overview of one solution to volcanic monitoring
3 Advantages and future work
Identified advantages of this implementation and upcoming work
Agenda:
4. PRESENTED BY
A research team in NoSQL databases:
• Interest to apply technologies in atypical scenarios
– Applied study cases
• Interdisciplinary work:
– Provides solutions to other research teams
– Currently working with Geophysics and Geography research groups
Team overview
5. PRESENTED BY
Costa Rica has many volcanoes:
• Around twenty
• Five are currently active
People live in the influence area:
• Two-thirds of the population around
three active ones
Monitoring is critical:
• Ash alerts: airport services and air
quality
• Eruptions: people, animals, crops,
and farms
Context
From: Costa Rica aérea
6. PRESENTED BY
RSN: Red Sismológica Nacional (National Seismic Network)
• Monitors volcanoes activities every single moment
• Every volcano with many stations monitoring its seismic activity using three
sensors
• Big amount of sensor data, only some portion is real-time processed
– Valuable calculations not available
• Requires a system that processes sensor data in real-time:
– Scientific calculations for volcanic activity
– Improving the monitor process
– Alerting population in case of emergency
Monitoring Volcanic Seismic Data
8. PRESENTED BY
In-memory storage means faster processing:
• Makes calculations highly precise
– Get updated seismic measures every second
• Improves monitoring work
– Experts with more volcanic activity information
– Authorities with better support for early warning of natural hazards
On-time calculations improve:
• Identification of particular events for future analyses
• Algorithm efficiency using a continuous stream
Redis in scientific scenarios
9. PRESENTED BY
Our Solution
Seedlinks
Store accordingly
to calculations
Pre-Processing
-StationID
-Axis
-Timestamp
-Hashes
-Lists
-Matrices (ML
Module)
-Streams?
{
Complete
calculations
Update results
each second
10. PRESENTED BY
• Based on Murray and Endo (1989)
• Used to visualize the seismic strain release rate
• Challenges:
– Stream packages may arrive in not strictly order
– Square operations over measure need to be completed every millisecond
• Solved using an auto-increment hash:
– One hash by each station using a key for each second
– Possible stream packages delays easily identified
– A square root is applied every second for each key
• These results are:
– Plotted every second
– Important for implementing localization algorithms
Real-time Seismic Energy Measurement (RSEM)
12. PRESENTED BY
• Proposed by Rogers and Stephens (1995)
• Helps to identify if the signal comes from the volcano and no from other
places
– For example, cows around the area could alter the measures
• Additional challenges:
– Fourier transform for time-defined windows
• Processing the data:
– Insert into a Redis list
– Apply fast Fourier transform for a time window
– Update the spectrogram each second
• An extension using Redis Streams is under consideration
Seismic Spectral Amplitude Measurement (SSAM)
15. PRESENTED BY
• Using several stations to determine
localization and depth
• Currently under development
• Based on Jurkevics (1988) and
Taisne et al. (2011)
• Require matrix calculations:
– Available through Redis ML Module
• Current challenges:
– Partial insertions are not possible in a
matrix
– Matrix multiplication
– High dependency on Python coding
Polarization and localization
From: Department of Computer Science and
Engineering. Michigan State University
17. PRESENTED BY
• High-speed processing:
– Continuously updated plots and immediate calculations
– Observation of evolution and dynamics of the volcanic activities
• Real-time calculations become faster than other approaches
– For example, batch processing of events
• Applicable for unconventional scenarios
– Other cases are under consideration, such as floods alerts
Advantages of using Redis
18. PRESENTED BY
• Complete polarization model
• Implement different localization algorithms
– Be able to make comparisons between them
– Build them over current data structures
• Flood alert model
– Population around rivers influenced areas
Future work