The DECENTER pro­ject meet­ing took place in the peri­od from 22–24 Octo­ber 2019 in Gren­oble. In a very lively meet­ing, the part­ners dis­cussed the review­ers’ com­ments and sug­ges­tions and made decisions on how to address them.

As a next step in the pro­cess, we are focused on elab­or­at­ing the use cases in the con­text of the DECENTER plat­form, in par­tic­u­lar, what DECENTER bene­fits from these use cases, and what the use cases bene­fit from DECENTER. The cur­rent status of imple­ment­a­tion of the use cases was dis­cussed and their innov­at­ive fea­tures.

From archi­tec­tur­al view­point, we dis­cussed in detail the nov­elty brought by the Digit­al Twin tech­no­logy, and how it is to be used in the con­text of the pro­ject, the advanced new orches­tra­tion mech­an­ism that will take into account the require­ments to dynam­ic­ally exchange AI meth­ods dur­ing the oper­a­tion of AI applic­a­tions, and the usage con­text of the Broker­age Plat­form that may oper­ate not only for com­put­ing resource, but also for the broker­age of AI mod­els.

Sev­er­al tasks of our AI-related work pack­age deal with the need to split AI mod­els, so that they can be used into sev­er­al com­put­ing tiers, for example, the front part of a Deep Neur­al Net­work can be deployed close to the Edge, and the rear part can be deployed »some­where« in the Fog/Cloud. The part­ners had an in-depth dis­cus­sion about the use of vari­ous (e.g. fed­er­ated) learn­ing mech­an­isms for AI con­sid­er­ing the pos­sible arrange­ment of applic­a­tions in mul­tiple com­put­ing tiers.

A DECENTER applic­a­tion can be arranged in two or more com­put­ing tiers, and our goals here are also to address the Qual­ity of Ser­vice require­ments when deployed in the Fog. This requires the imple­ment­a­tion of a spe­cif­ic mon­it­or­ing sys­tem. A very import­ant require­ment for the man­age­ment of Ser­vice-Level Agree­ments by the plat­form is that the Qual­ity of Ser­vice mod­el is cap­able of provid­ing assur­ances, rank­ing and veri­fic­a­tion of the applic­a­tion deploy­ment options.

Cross-bor­der data man­age­ment which includes the neces­sity to freely move AI mod­els from one coun­try to anoth­er while respect­ing vari­ous reg­u­la­tions, per­mis­sions, cer­ti­fic­a­tion and oth­er requre­ments was also dis­cussed dur­ing the meet­ing. The part­ners agreed to an inter­est­ing and innov­at­ive path to cross-bor­der data man­age­ment that includes the use of Block­chain, which will be imple­men­ted and revealed in the forth­com­ing peri­od.

Oth­er top­ics dis­cussed were the improve­ments of the Web site, the dis­sem­in­a­tion, stand­ard­isa­tion and exploit­a­tion activ­it­ies and man­age­ment. The meet­ing was led by our coordin­at­or Dr Domen­ico Siracusa in such an effect­ive way that all part­ners now have plenty of imme­di­ate duties to fol­low. The next pro­ject meet­ing will take place from Feb­ru­ary 11–13, 2020, in Slov­e­nia, a small, beau­ti­ful coun­try on the Sunny Side of the Alps. 

Categories: meetingnews