![Page 1: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/1.jpg)
Path Scheduling on Digital Microfluidic Biochips
Dan Grissom and Philip Brisk
University of California, Riverside
Design Automation ConferenceSan Francisco, CA, USA, June 5
![Page 2: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/2.jpg)
2
Digital Microfluidic Biochips (DMFB) 101
![Page 3: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/3.jpg)
3
Resource-constrained scheduling of operations into time-stepsTime-step ~ scheduling unit, usually 1s or 2s
Placement of operations during each time-step into modulesModule ~ 2D group of cells where operation takes place for 1+ time-steps
Routing of droplets between operations between time-steps
DMFB Synthesis
![Page 4: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/4.jpg)
4
Goal: Online Synthesis
Why: Programmability, control-flow, live-feedback
Problem: Past, optimized methods are too complex
Solution: Synthesis with good results in little time
High Level Motivation
[Luo et al., DATE 2012]
![Page 5: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/5.jpg)
5
DMFBs have limited storageUnlike computers, same resources used for operations and storage
Storage takes up space that can be used by useful operations
Scheduling Goal: Minimize storageIntroduce droplets to system ALAP
Use stored droplets ASAP to free space on array
DMFB Scheduling Problem
VS.
Extra Resource
![Page 6: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/6.jpg)
6
A path is a series of nodes that terminates with a mix/output
Paths are not unique
Colorimetric Protein Assay PCR Mixing Stage Assay
Schedule By Path
![Page 7: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/7.jpg)
7
Defined as the cumulative number of droplets being output in a node’s fan-out
IPP helps determine which node/path to schedule next
Colorimetric Protein Assay PCR Mixing Stage Assay
Independent Path Priority (IPP)
![Page 8: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/8.jpg)
8
1
1 5
7
1
1
1
1
1
1
1
2
2
2
2
2
2
2
3
3
3
3
3
3
3
4
4
4
4
4
4
4
5
5
5
5
5
5
5
6
6
6
6
6
6
6
7
7
7
7
7
7
7
8
8
8
8
8
8
8
531
Independent Path PrioritiesLive Example
Legend
Path leader (not on current path)
Potential next-node-on-path
Fully scheduled/submitted path
Scheduling Example
![Page 9: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/9.jpg)
9
Solution Quality Computational Complexity Storage Usage
Lower is better for all metrics# Assays Concurrent Scheduled # Assays Concurrent Scheduled # Assays Concurrent Scheduled
Protein Results
Implemented 3 schedulers in C++Path Scheduler
List-schedulerDecreasing critical-path priorities (worse)
Increasing critical-path priorities (better)
![Page 10: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/10.jpg)
10MLS_DEC
252 storage nodesMLS_INC
119 storage nodesPS
54 storage nodes
Black node = droplet being stored for a # of time-steps in 1 module
Placed Protein DAG Showing Storage
![Page 11: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/11.jpg)
11
More results in the paper/poster
Excels at assays with high fan-out and limited resourcesPath scheduler saves 100’s of seconds
Performs well on assays w/o high fan-outSimilar schedule quality
Slightly shorter computation time
Viable option for online synthesis
Conclusion
![Page 12: Path Scheduling on Digital Microfluidic Biochips Dan Grissom and Philip Brisk University of California, Riverside Design Automation Conference San Francisco,](https://reader035.vdocument.in/reader035/viewer/2022062308/56649eeb5503460f94bfd21f/html5/thumbnails/12.jpg)
12
Thank You