The PhD thesis addresses the problem of predicting real time malfunctions in large distributed information
systems (LDIS). Such systems are composed of thousands of servers and hundreds of applications
used by thousands (if not millions) of users. LDIS are everywhere. Each of us has everyday
experience of retail corporations or public bodies information systems, just to cite a few of
them. The pressure on the systems operating in professional environment of the economic sector is
very strong: 7-day 24-hour availability is required.
Predicting that an adverse event, failure or breakdown of some part of the system, will occur before
a defined date would make it possible avoiding the incident by palliative actions or circumventing it
by detour routes. While non critical anomalies are commonplace in LDIS, availability anomalies
(critical anomalies) are rare events. Because of this asymmetry in the statistics, predicting the latter
is difficult. The first aim is achieving reliable prediction of the anomalies by machine learning applied
to the continuous monitoring of the LDIS. The second aim is providing automated aid to diagnosis
allowing more reliable and faster management of the information system's incidents.
Most AI models ask for massive amounts of good quality labeled data. Here the monitoring is on
network traffic, server operations and malfunctions, service availability and breakdowns, etc. These
highly strategic data are kept confidential by both the company that monitors the system and the
customer company whose activity depends on the system availability.
Groupe HN is a french leading company in the monitoring of LDIS. For the needs of this research it
has set up a consortium of its customer companies (large banks, insurance companies, retail corporations)
that will provide the research work with the data: logbooks, metrics, description of the system
and metrics, diagnoses and actions taken by the operators with the results of these actions. As
well, the consortium will provide the support of its engineers in charge of the monitoring.
Group HN -- http://www.groupehn.com -- is an innovative technology company providing a wide
range of services including software & product development, consulting, staffing and training, to
the major players in finance, insurance, telecommunication, manufacturing and services industry. A
branch of Groupe HN is dedicated to the monitoring of large distributed information systems.
LISSI (EA 3956) -- http://www.lissi.fr/home/ -- is a laboratory of University Paris-Est Créteil. It
develops multidisciplinary, theoretical and applied research activities in the field of Information and
Communication Sciences and Technologies in particular in artificial intelligence. LISSI is part of
Paris-Est Graduate School ``Mathématiques & STIC'' -- http://bit.ly/2mc6lh4 --
ESIEE-Paris -- https://www.esiee.fr/en# is a school for engineering, member of University Paris-
Est -- https://www.univ-paris-est.fr -- Its expertise is in Computer Sciences & Technologies.
Candidate's cursus: MSc in Computer Sciences or school of engineering with strong CS track.
Starting date: 02/2020, duration: 3 years.
Funding: Groupe HN and ANR (French national research agency) or Groupe HN and ANRT
(French national technological research agency, CIFRE contract).
Location: ESIEE-Paris Computer Science department. 25 minutes to the very center of Paris, 20
minutes to Groupe HN (Charenton-le-Pont), 30 minutes to LISSI (Vitry-sur-Seine).
Contacts: CV and short motivation letter to Patrick Siarry (UPEC) Jacob Ouanounou (Groupe HN)
Arben Çela - René Natowicz (ESIEE)
firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com