Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The development of applications has evolved. Several years ago, the trend was to create some monolithic applications on a mainframe. Now, it is the micro-services architecture in the cloud. The requirements of the users have also changed. They want a secured application highly available at any time. These constraints transform the IT sector and the application development. The workflow must be quick and automatised.

Kubernetes is a potential solution to support the development of cloud native application. This master thesis will explore how it can be used to create an environment to fit the new constraints. It begins by the creation of a cluster of 6 instances in the cloud (3 masters and 3 workers) and the development of a web application to represent a real use case. After that, a comparison is made between some tools dedicated to the continuous integration and deployment. It is the basis of the devOps and gitOps principles which ensure an efficient workflow. The communication between the services is performed by Istio which is a service mesh tool. It is responsible of the routing and the enforcement of policy rules. A micro-services architecture has some weaknesses in term of security at the level of the cluster itself, the Kubernetes API, the images used by the container, the running containers and the services. All these elements can be configured to mitigate or reduce the consequences of an attack. The performances of Kubernetes and the added tools are analysed to determine in which cases Kubernetes is interesting. In the majority of the cases, it is a efficient choice if the application receives a variable load and must be highly available and secured. The conclusion and the future works close this master thesis.

container --- Docker --- Kubernetes --- cluster --- service mesh --- security --- Ingénierie, informatique & technologie > Sciences informatiques

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Message logging is the tool of choice to stay informed about the health of a machine or application. These messages, called logs, are used for various purposes, including system management, performance optimization, investigation of suspicious activities, and more generally analysis and debugging. Operations that demand a level of reliability at least equivalent to the emphasis placed on them during their use. However, the syslog protocol was originally designed to work exclusively over UDP. Traditional applications, which have not benefited from the a postorio additions such as TCP, are forced to communicate over a network that is not suitable for them (corrupted or lost messages, reordering, or unreachable server) and over which they have no control. The objective of this work is to develop a resilient syslog relay that will operate downstream of applications, collect their syslog messages and send them to a central syslog server. Several mechanisms such as the use of the TCP protocol and the retention of messages in case of connection loss guarantee reliability. Topics related to message ordering and strategies in case of an overload are also discussed and several approaches are presented to either mitigate or regulate their impact. The implementation, in the form of a prototype, is deployed inside a router running the Cisco IOx environment and features the modern syslog message engine, rsyslog. The model is evaluated on the basis of its functionality and performance in a test environment with network quality such as 3G cellular and EDGE. Several configurations are proposed depending on the type of usage involved. Although the solution does not cover all possible and imaginable problems, such as router outages, the evaluations demonstrate the efficiency and scalability of the proposed solution, which can for example easily handle several tens of thousands of messages per second with a very low resource footprint.

docker --- cisco --- iox --- syslog --- log --- rsyslog --- aarch64 --- linux --- event --- relay --- reliable --- store and forward --- kibana --- elasticsearch --- message generator --- python --- router --- cellular --- network --- TCP --- UDP --- TLS --- configuration --- ioxclient --- chronological order --- timestamp --- 3G --- EDGE --- performance --- testbed --- wanem --- downtime, --- recover --- permanent storage --- open source --- central server --- server --- low footprint --- Ingénierie, informatique & technologie > Sciences informatiques