Infrastructure for Scientific Instruments Lifetime Connectivity PI: - - PowerPoint PPT Presentation

CC*Integration: BRACELET: Robust Cloudlet Infrastructure for Scientific Instruments’ Lifetime Connectivity

PI: Klara Nahrstedt University of Illinois at Urbana-Champaign klara@Illinois.edu

Current situation in campus scientific instruments cyber-infrastructure

Building L2 high speed switch

Campus Network Building L2 high speed switch Building L2 high-speed switch Building L2 high speed switch

Private Cloud New instruments

Older instruments (offline)

New instruments

Goal of BRACELET: Connect older Scientific Instruments to Private Cloud

• Allow connections from Windows NT, Windows XP,

Windows 2000 PCs to Private Cloud

• Allow upload of data from old instruments to the

private cloud

• Allow some curation functions of data close to
• lder instruments

Challenges

• Performance mismatch: Older

instruments’ Windows NT or XP runs network protocols at lower bandwidth speeds (10Mbps or 100Mbps)

• Obsolete security: Older devices and their

OS systems cannot be patched, hence being vulnerable & taken offline

Approach: Putting edge device between

• lder instruments and private cloud

Building L2 high speed switch Campus Network Building L2 high speed switch L2 slow speed switch Cloudlet Building L2 high-speed switch Cloudlet

Building L2 high speed switch

…. ….

Private Cloud New instruments New instruments

L2 slow speed switch

Approach: Putting edge device between

• lder instruments and private cloud
• Performance component:
• Have two network interfaces configured at

different speeds

• Perform traffic shaping from slower to faster

network

• Security component:
• User & instrument registration
• Data encryption during upload
• Firewall to protect against external threats

Co-design of performance and security components at the cloudlet

Cloudlet & private cloud integration with resource monitoring & adapation

Task.A.Cloud Task.B.Cloud Task A’s cloud workers Task B’s cloud workers

Lookup task dependencies

• f workflow

Scientific instruments Scientific instruments

Task A’s edge workers Proxy Task.A.Edge /uploader

Campus network

4CeeD Curator Edge controller Cloud controller

Host cluster (Cloud) Edge node

A B 4CeeD Uploader

Task monitor

Kapacitor

System identification Control algorithm

System model

System states

Task monitor Workflow monitor TDS monitor TDS monitor Offloading protocol

• Neural network

system model

• Using GAN to

reduce training effort

• MPC for

controller design

• Centralized vs.

hierarchical control model

Lookup task dependencies

• f workflow

Scientific instruments

Current status

• Finalized preliminary design of BRACELET’s cloudlet

& its integration to private cloud architecture

• Setup BRACELET’s cloud & edge testbed
• Worked with Engineering IT to allocate & configure

isolated networks for testing BRACELET in real environment