Loading…
ONS2016 has ended

Be sure to register for Open Networking Summit 2016 - Register Now

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Track: Technology [clear filter]
Tuesday, March 15
 

3:05pm

The Need for Complex Analytics from Forwarding Pipelines - Nic Viljoen, Netronome; Tom Tofigh & Bryan Sullivan, AT&T
Given that the software paradigm leads to faster innovation, we have learned that emerging forwarding pipelines such as switches, Routers, and compute and storage elements are expected to be programmable and have the ability to support complex analytic measurements. To these switches to achieve both high performance packet processing and complex measurement, the pipelines needs to be programmable as suggested by P4. This talk will make an attempt to conceptualize the complexity of current overlay analytics networks and how SDN and P4 approach can be used to program the pipelines for adaptive advanced measurement.
In addition we will show how SDN based platform such as ONOS can be instrumental to achieve programmable pipelines that are friendly to complex measurements for small and large flows. This talk will be complementary to Analytic cord.

Speakers
BS

Bryan Sullivan

AT&T, Director of Service Standards
Bryan Sullivan is the Director of Service Standards of AT&T and leads a team representing AT&T in a diverse set of Open source platform architecture and development.
TT

Tom Tofigh

AT&T, Principal Member of Technical Staff
Tom Tofigh is Principal Member of Technical Staff in the AT&T’s Domain 2.0 architecture and planning Labs at AT&T. Currently involved in development activities at Open Networking Lab with various use cases including SDN/NFV, Mobile Cord and disaggregation of optical equipment's... Read More →
avatar for Nicolaas Viljoen

Nicolaas Viljoen

Associate Director Software Engineering, Netronome
Nic works on upstream solutions using SmartNICs, focusing on the interaction between SmartNICs and the Linux kernel using the upstream BPF JIT for the Netronome NFP processor.


Tuesday March 15, 2016 3:05pm - 3:25pm
Grand Ballroom F Santa Clara Convention Center

3:30pm

Interactive Programming for Dynamic Network Probe: A Use Case for POF Data Plane - Haoyu Song, Huawei Technologies
Network operators require better network observability in order to use the data acquired to better support their business needs such as improving QoE, customer care, and security. The network data plane needs to be interactively programmable to support such applications in real time and on demand. This is a further stretch of the open programmable data plane. We developed the protocol oblivious forwarding technology, protocol, and prototype to show how this is achieved in an efficient way. POF allow users to dynamically install probes at any places in the data plane and run customized actions to monitor the data plane states and events. We show our system could be the foundation of a networking big data analytics platform.

Speakers
avatar for Haoyu Song

Haoyu Song

Principal Architect, Huawei Technologies


Tuesday March 15, 2016 3:30pm - 3:50pm
Grand Ballroom F Santa Clara Convention Center

4:10pm

Container-based Dynamic Service Chain using Distributed SDN Pipeline Injection (Openstack, Dragonflow, Kuryr) - Eshed Gal-Or, Huawei
With NFV becoming a reality, OpenStack Cloud deployments grow in number and size, giving birth to issues of scale, performance and service flexibility.
Many struggle to minimize the effects of the virtual overlay on the physical network, while also looking to provide network services that are both efficient and flexible.
In this session, I will review two new Open Source projects based on Linux, Containers and OpenVirtualSwitch, that may hold the answer:
Dragonflow – a lean distributed SDN controller that fully implements OpenStack Neutron and provides several network services (e.g. DHCP, L2, L3)
Kuryr – a project that enables Containers networking to be managed in OpenStack Neutron.
Join this session to see how these projects can be used together to bring Service Injection into an OpenStack-based cloud, enabling a range of new applications in NFV and IT Cloud.

Speakers
avatar for Eshed Gal-Or

Eshed Gal-Or

Chief Architect Open Source, Huawei
I'm an avid technologist, innovation leader and problem solver, with 20 years of R&D experience in diverse domains, such as networking, virtualization, telecom, BSS, IT infrastructure, cloud and security. | I currently manage a team of open source developers and PTLs in Huawei Tel... Read More →


Tuesday March 15, 2016 4:10pm - 4:30pm
Grand Ballroom F Santa Clara Convention Center

4:35pm

Performance Considerations When Deploying Open Source NFV Infrastructure - Eric Carmès, 6WIND
Performance bottlenecks persist in VMs within NFV use cases and new Web 2.0 applications slow down the adoption of open source-based virtual applications in OpenStack deployments, questioning the wisdom of adding under-utilized multi- 10GE and 40GE interfaces on multi-core platforms. By leveraging data plane acceleration within the OpenStack hypervisor, these bottlenecks are resolved and provide line rate throughput across multiple 10G, 25G, 40G, 50G and 100G interfaces to meet all virtual application bandwidth requirements, which can help accelerate the adoption of open source networking.

This session addresses the means to integrate and automate OpenStack hypervisor acceleration to eliminate performance bottlenecks that challenge the adoption of NFV, with the end result goal to scale data centers beyond 10G leveraging open source NFV infrastructure.

Speakers
avatar for Eric Carmès

Eric Carmès

Founder & CEO, 6WIND
Eric Carmès is a Founder and the Chief Executive Officer of 6WIND. In this capacity, he drives the corporate strategy and growth. Under his leadership, 6WIND was the very first company to understand the impact of multicore architectures on the networking software industry and became... Read More →


Tuesday March 15, 2016 4:35pm - 4:55pm
Grand Ballroom F Santa Clara Convention Center
 
Wednesday, March 16
 

3:05pm

MIND: A Machine Learning-based SDN Control Paradigm - Yanhui Geng, Huawei
SDN has emerged as the preferred networking approach for large datacenters. Despite its benefits, the rapidly changing traffic patterns and the dynamic nature of the network pose significant challenges to the decision-making process. To address this, we will present in this talk the MIND (Machine learning based Intelligent Network Dynamics) framework proposed by Huawei Noah’s Ark Lab. The MIND architecture consists of components that employ online machine learning techniques to predict/infer the spatial-temporal traffic information or network state. These results are then used by a policy generation module, which devises an optimal routing policy by learning from the historical data based on a reinforcement learning framework. Extensive experiments with a real SDN testbed demonstrate the feasibility and effectiveness of the self-learning SDN paradigm.

Speakers
avatar for Yanhui Geng

Yanhui Geng

Project Manager, Huawei Tech. Investment Co. Ltd
Dr. Yanhui Geng is a researcher and project manager at Huawei Noah's Ark Lab (Hong Kong). He manages one project as the Principal Investigator on optimising SDN performance by applying big data analytics on real-time network flows and adopting more intelligent network control strategies... Read More →


Wednesday March 16, 2016 3:05pm - 3:25pm
Grand Ballroom F Santa Clara Convention Center

4:35pm

Enhancing the SDN Ecosystem - Evolution of OpenFlow - Ben Mack-Crane, Corsa Technology
SDN run-time control solutions need to be more capable, modular, interoperable and robust. The ONF is evolving OpenFlow to accomplish this by defining common core protocol that is independent of datapath protocols (i.e., eliminates assumptions about switch functionality). This enables OpenFlow to be applied equally to any datapath technology. Protocol elements addressing specific datapath technologies are moved to separate specification modules. Network architecture considerations are captured in Table Type Patterns (TTPs) that define the capabilities of SDN elements playing specific roles in a network. Another aim of this work is to support emerging technologies for programmable dataplanes (e.g., the P4 datapath programming language). This presentation will describe how OpenFlow and related SDN software components are being (re-)structured to enhance the open source SDN ecosystem.

Speakers
BM

Ben Mack-Crane

Corsa Technology


Wednesday March 16, 2016 4:35pm - 4:55pm
Grand Ballroom F Santa Clara Convention Center

5:00pm

Adapting Open-Source NMS to an SDN Reality - Jeff Gehlbach, The OpenNMS Group
Since the late 1980s, the discipline of network management has kept up well enough with virtual notions in networks while continuing to rely heavily on the Simple Network Management Protocol (SNMP). The rise of SDN and NFV poses similar challenges to the ones brought to the systems management space by server virtualization and cloud computing, leaving practitioners to wonder how much retooling will be needed to cope in this new world.

As the oldest enterprise-grade, open-source network management platform, OpenNMS has proved fairly adaptable so far. In this presentation, Jeff Gehlbach will highlight inventory integrations against OpenDaylight and other controllers; serviceability, topology, and metric data collection both via controller APIs and directly from switches; and what opportunities for growth exist today and in the near future.

Speakers

Wednesday March 16, 2016 5:00pm - 5:20pm
Grand Ballroom F Santa Clara Convention Center

5:25pm

Operating OpenStack on a Budget - Dan Mihai Dumitriu & Susan Wu, Midokura
Interest in deploying private clouds based on OpenStack is on the rise, as more and more web scale companies and enterprises consuming public cloud infrastructures alike want to replicate a similar set-up for their in-house software development.

There is a growing trend of modern enterprises leveraging both Amazon Web Services (AWS) and on-premise OpenStack-based private clouds. However, the default networking option in OpenStack remains broken and the plethora of confusing plug-ins makes networking in OpenStack mysterious and difficult to manage.

Enter MidoNet, the open source network virtualization solution favored by DevOps cultures in web scale enterprises and service providers alike. This session will show how Blue Jeans Networks uses MidoNet to build, run and manage large scale virtual networks in OpenStack clouds.

Speakers
avatar for Dan Mihai Dumitriu

Dan Mihai Dumitriu

CEO and Co-Founder, Midokura
Dan is responsible for the technical innovation and development of designing, building and operating Midokura technology. Prior to founding Midokura, Dan served as Chief Architect at Ballista Securities, a New York City ATS offering an electronic block trading system for options... Read More →
SW

Susan Wu

Director of Technical Marketing, Midokura
Susan is the Director of Technical Marketing at Midokura. Susan previously led product positions for Oracle/Sun, Citrix, AMD and Docker. She is a frequent speaker for industry conferences like OSCON, OpenStack Summit, Interop ITX, Container World, All Things Open, Linuxcon/CloudOpen/Containercon... Read More →


Wednesday March 16, 2016 5:25pm - 5:45pm
Grand Ballroom F Santa Clara Convention Center