Vote for Sessions! OpenStack Summit in Paris

Big Data, High Performance Networks,

We have submitted several speaking sessions for the upcoming OpenStack Summit in Paris, France. Please review the descriptions for each session below, click to vote and share these presentations with your colleagues.  Remember:  Voting closes on Wednesday, August 6, 2014 at 11:59 pm CDT.



OpenStack, My Love & Challenge story: A story of experience with distributions   

Share this session!  [Click to tweet]

Level:  Beginner

Presented by:

  • Eyal Gutkind, Director of Big Data & Cloud Market Development, Mellanox
  • Arik Kol, Director of Business Development, Mellanox

In this session we review our experience in building OpenStack deployments over high performance networks. We are looking on 10Gb Ethernet, 40Gb Ethernet and FDR InfiniBand as the management, private and storage networks and the benefits for the users.  We will walk through the consideration of each choice and the work we done to integrate hardware offloads to the different distributions. We will talk about our experience with the difference open source distribution and our recommended hardware stack. We will show usage of OpenStack with different block and object storage solution and the points we believe can help you make a more efficient OpenStack Deployment.



Deploying Red Hat Block and Object Storage with Mellanox and Red Hat OpenStack 

Share this session!  [Click to tweet]

Level: Intermediate

Presented by: 

  • Eyal Gutkind, Director of Big Data and Cloud Market Development, Mellanox
  • Thomas Cameron, Chief Architect, Central US, RedHat
  • Erez Cohen, Sr. Director CloudX Program, Mellanox

As data continues to grow exponentially storing today’s data volumes in an efficient way is a challenge. Many traditional storage solutions neither scale-out nor make it feasible from Capex and Opex perspective, to deploy Petabyte or Exabyte data stores. A novel approach is required to manage present-day data volumes and provide users with reasonable access time at a manageable cost. This presentation summarizes the installation and performance benchmarks of Red Hat Ceph storage solution for Red Hat OpenStack orchestration. Ceph is a massively scalable, open source, software-defined storage solution, which uniquely provides object, block and file system services with a single, unified Ceph Storage Cluster.

The testing emphasizes the careful network architecture design necessary to handle users’ data throughput and transaction requirements. Benchmarks show that a single user can generate read throughput requirements able to saturate a 10Gbps Ethernet network, while the write performance is largely determined by the cluster’s media (Hard Drives and Solid State Drives) capabilities. For even a modestly sized Ceph deployment, the usage of a 40Gbps Ethernet network as the cluster network (“backend”) is imperative to maintain a performing cluster. In the session we will review a reference architecture and cost estimation of a solution.


VXLAN Offloads – Overcoming Performance Challenges with Hardware Offloads

Share this session!  [Click to tweet]

Level: Intermediate

Presented by:

  • Eyal Gutkind, Director of Big Data & Cloud Market Development, Mellanox
  • Arik Kol, Director of Business Development, Mellanox

Virtual Extensible LAN (VXLAN) is a network virtualization technology that enables a VLAN-like encapsulation technique to incorporate layer 2 Ethernet frames within layer 3 packets. This approach helps scale to millions of virtual layer 2 networks over IP networks. In addition, VXLAN enables layer 2 expansion over wide area reach.

However, the encapsulation creates an overhead on the CPU of the virtual machines (VM), since common offload techniques cannot operate on the VM TCP/IP packets (inner payload).

In this session we will review the following approach for hardware VXLAN offloading:

  • Stateless Hardware offload which move the most critical VXLAN processing to hardware, and maximize CPU core parallelism, by offloading Rx/Tx Checksum of the inner packets,
  • Large send offloads for inner packets, GRO and Multiple receive and send queues, packet steering based on inner payload to improve parallelism

We will review the architecture and performance results of the hardware acceleration.


Pushing Packets through OpenStack: How Good is OpenStack for the NFV Use Case?

Share this session! [Click to Tweet]

Level: Intermediate

Presented by:

  • Gregory Elkinbard, Senior Technical Director, Mirantis
  • Arik Kol, Director of Business Development, Mellanox
  • Aleksandr Shaposhnikov, Network Architect, Mirantis


OpenStack Neutron has evolved into a very flexible network platform and has risen up to the challenges of generic cloud applications. However, virtualized network functions need more. High performance, low latency, reliability at the software and hardware levels and fast convergence after a fault. This paper will focus on two topics: Network performance and scalability and growth strategy.

In the network performance section, we will examine performance of a VM as a package engine using standard software switching with openvswitch and contrast it with SRIOV approach provided by Mellanox. This will be done using good old VLANs.

As the NFV clouds scale they will need to go beyond VLANs, and this paper will examine the two possible approaches to driving the scalability of the networks, OpenFlow and software overlay using VXLAN


Additional Proposed Sessions:

  • Light up performance of your LAMP apps and DevOps with a Power optimized stack  
    • Presenter:  Alice Spencer –
  • A Case for Fabric-Aware Datacenters to support a flexible OpenStack Storage Infrastructure  
  • VXLAN Gateways – Bridging software overlays to physical workloads using white box top-of-rack switching
    • Presenter:  Michael Ford – 
  • Accelerate Network Access to OpenStack Block Storage

Comments are closed.