Analysis of failures impact on operation of a software-defined network

Programski upravljana mreža (SDN) je nova paradigma koja zadovoljava zahtjeve za fleksibilnim upravljanjem mrežom. Mrežom upravlja udaljeni SDN upravljački uređaj, koji ima globalni pogled na mrežu te donosi odluke o konfiguraciji mrežnih uređaja. Kako se SDN upravljački uređaj ne bi preopteretio, u kontrolnom sloju postoji nekoliko instanci takvog uređaja koje zajedno upravljaju SDN mrežom. Prilikom mogućeg ispada pojedine instance, moraju postojati mehanizmi oporavka koji brzo i učinkovito raspodjele mrežne uređaje na preostale instance SDN upravljačkog uređaja. U ovom radu je dana analiza utjecaja kvarova instanci SDN upravljačkog uređaja ONOS na rad SDN mreže. Pokazano je da, uslijed kvarova nekih instanci, samo jedna od njih nije dovoljna kako bi SDN mreža nastavila s ispravnim radom.Software-Defined Network (SDN) is a new paradigm that meets the requirements of flexible network management. The network is managed by a remote SDN controller that has a global network view and makes decisions on configuration of network devices. In order to prevent SDN controller from overloading, there are several instances of such a controller in the control layer that manage the network together. In a likely event of one instance failing, recovery mechanisms must exist that quickly and efficiently distribute network devices among the existing SDN controller instances. This thesis analyses impact of failures with respect to instances of SDN controller ONOS on the SDN network operation. It is shown that, after some instances fail, one of them is not sufficient to continue supporting the correct operation of an SDN network