Mimicking FogDirector Application Management Stefano Fort orti, Ahmad Ibrahim and Antonio Brogi Service-oriented, Cloud and Fog Computing Research Group Department of Computer Science University of Pisa, Italy 12th Symposium and Summer School On Service-Oriented Computing, June 25 – June 29, 2018 in Crete, Greece 1
4
IoT Deployment Models Cloud • Not sufficient per se to support the IoT mo momentu tum Internet alone. • There is a need for filt lterin ing and pr proc ocessin ing before the Cloud. M2M/ LAN/WAN • Processing should occur wherever it is best-pla laced ed for any given IoT application IoT+ T+Edge IoT+Clo loud • Low latencies, but • Huge computing power, but • Limited capabilities, • Mandatory connectivity, • Difficulties in sharing data • High latencies, • Bandwidth bottleneck. 5
Fog Computing Cloud Fog computing is a system-level horizontal architecture that distributes resources and services of computing, storage, control and networking anywhere along the continuum from Cloud to Things , thereby accelerating the velocity of decision making . Fog -centric architecture serves a specific subset of business problems that cannot be successfully Fog implemented using only traditional cloud based architectures or solely intelligent edge devices . [ OpenFog Reference Architecture , 2016.] 6
Fog Director A single pane of glass to manage ap appli lication lif lifec ecycle le on Fog devices. GUI GU A P I # management.py app FogDirector # management.py infrastructure # management.py admin publish(id,A) n = choose(get_info_N()) publish(id,A) d = deploy(id,n) n = choose(get_info_N()) publish(id,A) start(d) d = deploy(id,n) n = choose(get_info_N()) start(d) d = deploy(id,n) on alert do start(d) PI REST API on alert do stop(d) undeploy(d); on alert do stop(d) m = choose(get_info_N())) undeploy(d); stop(d) deploy(id, m) m = choose(get_info_N())) undeploy(d); deploy(id, m) m = choose(get_info_N())) RE deploy(id, m) 7
The App SmartBuilding Steve, App Admin 8
Problems #1 (quickly) #1 #2 #2 write un understand corr orrect and Fog ogDir irector effect ctive functioning management 9
Problem #1 10 https://www.cisco.com/c/en/us/td/docs/routers/access/800/software/guides/iox/fog-director/reference-guide/1-6/fog_director_ref_guide.html
Solution #1 • Ope peratio ional l sem semantic ics of all ba basi sic fun functionalit itie ies of FogDirector. • Com Compact and un unambiguous reference. 11
Anatomy of a rule C LIENT C ONDITIONS FOR O PERATION AND M ANAGEMENT E RROR C ODE THE OPERATION TO P ARAMETERS P ROGRAM BE SUCCESSFUL I NFRASTRUCTURE M ANAGED A PPS S TATE 12
Problem #2 • Corr Correctness can be verified by using the semantics. • Effectiveness involves considering variations in: • Fog node res esources and • QoS of communication links write corr orrect and • Wh effective What the then? management 13
Problem #2 • Corr Correctness can be verified by using the semantics. • Effectiveness involves considering variations in: • Fog node res esources and • QoS of communication links write corr orrect and • Wh effective What the then? management 14
Solution #2 FogDirMime is the core of a simulator: - Infr frastructure mgmt - App mgmt - Monitoring & Aler erts (A2T, resource) # management.py FogDirMime publish(id,A) n = choose(get_info_N()) d = deploy(id,n) A start(d) P on alert do I app stop(d) monitored undeploy(d); admin m = choose(get_info_N())) infrastructure data deploy(id, m) https://github.com/di-unipi-socc/FogDirMime 15
The Big Picture varying network app admin A QoS Qo S and workload P I conditions # management.py FogDirector infrastructure publish(id,A) n = choose(get_info_N()) d = deploy(id,n) start(d) on alert do stop(d) undeploy(d); m = choose(get_info_N())) deploy(id, m) FogDirMime probabilistic sampling of f A latency & bandwidth P I and of f available node monitored infrastructure data res esources 16
A (simple) example Infr fras astru ruct cture App pp 17
A (simple) example FogDirMime A P I 18
Conclusions con oncise and validation of perf perform rmance pr pred ediction management scripts at una unambiguous s reference and tuni tuning of desi esign time for FogDirector management 19
Future Work implement a full- include othe other fledged sim simulatio ion fun functionalit itie ies and en environment for QoS QoS-aware FogDirector management consider sc scali ling and study other recent osmotic com osm omputing tools such as ry TM for mul ulti-component EdgeX Fou oundry appli ap lications 20
Q&A → Poster Session 21
Mimicking FogDirector Application Management Stefano Fort orti, Ahmad Ibrahim and Antonio Brogi Service-oriented, Cloud and Fog Computing Research Group Department of Computer Science University of Pisa, Italy 12th Symposium and Summer School On Service-Oriented Computing, June 25 – June 29, 2018 in Crete, Greece 22
Recommend
More recommend