Cristian Lumezanu

HybNET: network manager for a hybrid network infrastructure

The emergence of Software-Defined Networking(SDN) has led to a paradigm shift in network management. SDN has the capability to provide clear and easy management of complex operational challenges in large scale networks. However, most of the existing work in SDN network management assumes a full deployment of SDN enabled network switches. Due to both practical and financial limitation real implementations are likely to transition through a partial deployment. In this paper, we describe our experience in the design of HybNET a framework for automated network management of a hybrid network infrastructure (both SDN and legacy network infrastructure). We discuss some of the challenges we encountered, and provide a best-effort solution in providing compatibility between legacy and SDN switches while retaining some of the advantages and flexibility of SDN enabled switches. We have tested our tool on small hybrid network infrastructure, and applied it to manage the OpenStack Neutron interface a well known open-source IaaS provider.

Hybrid network management

We describe our collaborative efforts towards the design and implementation of a next generation integrated network management system for hybrid networks (INMS/HN). We describe the overall software architecture of the system at its current stage of development. This network management system is specifically designed to address issues relevant for complex heterogeneous networks consisting of seamlessly interoperable terrestrial and satellite networks. Network management systems are a key element for interoperability in such networks. We describe the integration of configuration management and performance management. The next step in this integration is fault management. In particular we describe the object model, issues of the Graphical User Interface (GUI), browsing tools and performance data graphical widget displays, management information database (MIB) organization issues. Several components of the system are being commercialized by Hughes Network Systems.

Enabling layer 2 pathlet tracing through context encoding in software-defined networking

Troubleshooting Software-Defined Networks requires a structured approach to detect mistranslations between high-level intent (policy) and low-level forwarding behavior, and a flexible on-demand packet tracing tool is highly desirable on the data plane. In this paper, we introduce a Layer 2 path tracing utility named PathletTracer. PathletTracer offers an interface for users to specify multiple Layer 2 paths to inspect. Based on the Layer 2 paths of interests, PathletTracer then accounts paths with identifiable IDs, and installs a set of flow table entries into switches to imprint path IDs on the packets going through. PathletTracer re-uses some fields in packet headers such as the ToS octet for recording path IDs. To efficiently carry imprints using limited bits, PathletTracer uses an encoding algorithm motivated by the calling context encoding scheme in the software engineering domain. With k bits for encoding, PathletTracer is able to trace more than 2k paths simultaneously.

Observing common spam in Twitter and email

Spam is pervasive across many types of electronic communication, including email, instant messaging, and social networks. To reach more users and increase financial gain, many spammers now use multiple content-sharing platforms---including online social networks---to disseminate spam. In this paper, we perform a joint analysis of spam in email and social networks. We use spam data from Yahoos web-based email service and from Twitter to characterize the publishing behavior and effectiveness of spam advertised across both platforms. We show that email spammers that also advertise on Twitter tend to send more email spam than those advertising exclusively through email. Further, we use DNS lookup information to show that sending spam on both email and Twitter correlates with a significant increase in coverage: spam domains appearing on both platforms are looked up by an order of magnitude more networks than domains using just one of the two platforms.

Generating consistent updates for software-defined network configurations

This paper addresses the problem of consistently updating network configurations in a software-defined network. We are interested in generating an update sequence ordering that guarantees per-packet consistency. We present a procedure that computes a safe update sequence by generating an add-before rule dependency graph. Nodes in the graph correspond to rules to be installed and edges capture dependency relations among them.

An analytics approach to traffic analysis in network virtualization

Network virtualization has been propounded as a diversifying attribute of the future inter-networking paradigm. However, monitoring and troubleshooting operational virtual networks can be a daunting task, due to their size, distributed state, and additional complexity introduced by network virtualization. We propose an analytics approach for the analysis of network traces collected across hypervisors and switches. To re-organize individual trace events into path-wise slices that represent the life-cycle of individual packets, we first present a trace slicing scheme. Then, we develop a path characterization scheme to extract feature matrices from those trace slices. Using those feature metrics, we develop a set of trace analysis algorithms to cluster, rank, query, and verify packet traces. We have developed the analytics approach in a SDN network management tool, and presented evaluation results to show how it can enable visibility and effective problem diagnosis in a SDN network.

Deep Learning IP Network Representations

We present DIP, a deep learning based framework to learn structural properties of the Internet, such as node clustering or distance between nodes. Existing embedding-based approaches use linear algorithms on a single source of data, such as latency or hop count information, to approximate the position of a node in the Internet. In contrast, DIP computes low-dimensional representations of nodes that preserve structural properties and non-linear relationships across multiple, heterogeneous sources of structural information, such as IP, routing, and distance information. Using a large real-world data set, we show that DIP learns representations that preserve the real-world clustering of the associated nodes and predicts distance between them more than 30% better than a mean-based approach. Furthermore, DIP accurately imputes hop count distance to unknown hosts (i.e., not used in training) given only their IP addresses and routable prefixes. Our framework is extensible to new data sources and applicable to a wide range of problems in network monitoring and security.