The US Department of Energy (DOE)s networking cyber infrastructures are experiencing a massive explosion of data volumes because of emerging e-science applications to support high-energy physics, astrophysics, genomics, climate modeling, large-scale science collaborations, etc. The same is true for other networks around the world to support smart devices, big data applications, social networks, triple play, etc. Novel paradigms such as cloud services, Network Functions Virtualization (NFV), Software Denied Networking (SDN), etc. are emerging to support new Internet applications. Traditional approaches for network management use many disjoint tools at dierent networking layers for monitoring, management, and planning, thereby increasing overall management complexity and costs. Network administrators often follow a cumbersome process of searching and correlating performance bottlenecks, and perform manual network and trac engineering to solve these issues. In todays heterogeneous, multi-domain, multi-layer networks, application demands vary widely during the day. Thus, optimizing service delivery through resource-constrained net- works is a complex task, and requires fast analytics and correlation of data from dierent layers/domains and better visualization to augment the network management process to satisfy application requirements. Realizing the above challenges, in this SBIR proposal, Ennetix will develop XNet, a cloud-based network management platform which will provide ecient means for collecting network data from various sources, including perfSONAR services. This Software-as-a-Service (SaaS) platform will leverage cloud computing resources to analyze and correlate management data and detect performance bottlenecks and hotspots in multi-domain/multi-layer networks (without needing to deploy boxes in hundreds to thousands of locations). XNet will also provide intuitive and meaningful web-based visualization of data and analysis, and will support agility for dynamic and fast network reconguration. The architecture is modular and extensible to generate and consume data, e.g., performance analytics, network ow analysis, trac engineering, etc.; and, in future, can be used to develop tools and services for other vertical business applications, e.g., security, capacity planning (i.e., eectively grow the network to meet future trac demands), energy and fault management, etc. XNet can work as a virtual collaboration environment to manage federated networks and present ubiquitous visibility and control of the network without expensive management appliances, thereby simplifying a network operators task of running a network and optimizing it by balancing performance and cost. In this Phase I eort, Ennetix will conduct a full requirements analysis and design of the XNet platform architecture. A working XNet prototype will be developed, and evaluation studies will be conducted to determine XNets eectiveness and scalability in actively managing large IT infrastructures. Ennetix will collaborate with UC Davis (a large enterprise) and LBNL/ESnet (DOEs nationwide network) to pursue initial performance and scalability studies of XNet. This work is part of a larger eort intended to lead to a Phase II pro ject which will focus on commercializable cloud-based network management services. The proposed solution will greatly benet network administrators at DOE and other government organi- zations via an innovative and collaborative network management platform. The wider benets of this eort will extend well beyond the immediate DOE scientic community, and on to other network operators and enterprises. In particular, many commercial network operators and enterprises can leverage the proposed SaaS services to build tools for actively managing their complex, multi-domain/multi-layer networks.
