Q&A With Umesh Mahajan, CEO of Avi Networks, Regarding Its Cloud Application Delivery Platform

Cloud Computing Today recently had the opportunity to speak with Umesh Mahajan, CEO of Avi Networks, about the company’s cloud application delivery platform (CADP). Avi Networks uses analytics to help ensure consistent end-user experiences of applications by way of a platform that delivers the application delivery and load balancing technologies used by internet giants such as Facebook and Yahoo to the enterprise. The Avi Networks HYDRA-powered Cloud Application Delivery platform empowers enterprises to optimize load balancing and application delivery within private cloud deployments as well as hybrid cloud infrastructures. Cloud Computing Today engaged Avi Network’s CEO Umesh Mahajan about the cloud application delivery space, the need for hyperscale application delivery platforms within the enterprise and HYDRA’s analytics capabilities.

Cloud Computing Today: How do you envision the cloud application delivery space? What is the key differentiation of Avi Networks within the cloud application delivery space?

Umesh Mahajan (CEO, Avi Networks): Over the past decade, requirements for application delivery have gone through a complete transformation driven by changing modes of application consumption and an evolution of application architectures. In an attempt to keep up, traditional ADC vendors have brought forth incremental improvements in performance, scale, security and availability. I call this device-centric innovation, that is, making existing appliances better, bigger, faster. However, today we are faced with mega trends such as hybrid cloud adoption and a mobile-first access policy in most large enterprises. This has resulted in the emergence of the software-defined data center, which legacy networking technologies are hard-pressed to support.

In this new world, Avi Networks bring three key pieces of innovation. First, we use real-time analytics that tracks and uses a number of telemetric data to dynamically adapt the application delivery services being provided. This enables IT to guarantee application SLAs and accounts for sudden spikes in user traffic. Second, we have developed the industry’s first distributed load-balancing architecture that’s based on SDN principles with a clean separation of control from data plane. This provides a dramatic simplification of network operations via a single centralized controller while the data plane services can span, serve and scale multiple data center and cloud locations. And finally, as opposed to a “pay-upfront” model that exists today with appliance based application delivery solutions, we price our software-only solution based on the actual network services our customers consume.

Cloud Computing Today: Avi Networks attempts to bring the application delivery and load balancing technologies enjoyed by the likes of Facebook and Google to the enterprise. Describe the business need to bring the application delivery capabilities of companies such as Google and Facebook to enterprises that have vastly different workloads and business needs than internet companies that ingest petabytes of data daily.

Umesh Mahajan (CEO, Avi Networks): Companies such as Google, NetFlix and Facebook who deal in hyperscale environments are important because they have redefined the online end-user experience. It amazes me how a Google app can almost predict my next move and suggest ways to auto-complete my text. The other important benchmark hypersale vendors have established is the expectation that their applications will never, ever go down. These are realms that most enterprise apps can dare dream about today.

We build Avi Networks with the goal of enabling the same level of application experience for enterprise apps enjoyed by Web 2.0 and hyperscale users. We’ve also taken inspiration from and have integrated ways to make data centers more efficient through software that can run on any hardware processor and appliance. Of course, the use of real-time analytics to drive smarter even predictive load balancing is another hyperscale innovation that we’re using. A relevant example is the Facebook Autoscale project that the company designed to enable energy- and CPU-efficient load balancing. Finally, these companies have shown the power of automation that drives a “self-service” and agile operating model for the network admins, which also distinguishes Avi from legacy ADC vendors.

Cloud Computing Today: What is most notable about HYDRA’s built-in analytics capabilities?

Umesh Mahajan (CEO, Avi Networks): At the highest level, we’re committed to making the job of the network administrators dead simple so that they can focus on strategic initiatives instead of simply keeping the lights on and troubleshooting application performance issues. It’s very common that the networking teams get the brunt of the blame for poor application performance, irrespective of the real reason for the issue. That’s where Avi comes in by arming network administrators with real-time data and troubleshooting capabilities to shorten what some people refer to as the “mean time to innocence.”

But what’s truly notable about the HYDRA architecture is that it is a true software-defined network model, composed of a centralized controller and a distributed, scalable data plane (called service engines) that can be co-located with the applications within and across the cloud. This tight alignment not only enables service engines to serve as micro load-balancers but also as integrated data collectors that pick up the ambient statistics about every user-to-application transaction. This way they become hundreds of eyes in the data center, ever vigilant. This data is streamlined via reduction filters and compression techniques and then sent to a continuous data store within Avi Controller. The output is very granular and gives real-time insights about the end-to-end timing of application transactions, application health score, client logs, and client insights that can predict and proactively prevent any degradation in application performance – at any scale, any location for every user.

Bio: Umesh Mahajan, Chief Executive Officer & Co-founder of Avi Networks

A seasoned executive and entrepreneur with 25+ years of experience in tech industry, Umesh has helped develop the vision, strategy, and execution plan for several innovative technology products. Before co-founding Avi Networks, Umesh was the Vice President / General Manager for a $2B Data Center switching business at Cisco where he led engineering, product management, marketing and operations for Nexus 7000, MDS switching products, and the NX-OS operating system.

Prior to his work at Cisco, Umesh led the software team at Andiamo in architecting and delivering SAN-OS. Umesh has a Master of Science degree in computer science from Duke, a Bachelor of Science degree from IIT Delhi, and has 29 patents to date.


Midokura Partners With Cumulus Networks To Bridge Virtualized and Physical Infrastructures

Network virtualization vendor Midokura recently announced details of a partnership with Cumulus Networks that expands its ability to connect virtual, software defined networks to physical infrastructures and workloads. Midokura’s Midonet platform is a distributed, software-defined, networking platform that specializes in virtualized networking in a space that features the likes of Juniper Contrail, PLUMgrid, and VMware NSX. Cumulus Networks, meanwhile, offers Cumulus Linux, a Linux-based operating system for networking hardware. Whereas Midokura’s Midonet platform previously was able to connect virtualized networking infrastructures to non-virtualized infrastructures by way of its support of Layer 2 Gateway technology, its partnership with Cumulus Networks expands its ability to support non-virtualized infrastructures by adding enhanced speed, automation, scalability and performance to the connection between virtualized and physical workloads. Midonet connects to physical network switches that reside within a Cumulus Linux-based OS by means of a VxLAN Tunnel Endpoint (VTEP) gateway.

Because VxLAN represents an industry standard networking technology, Midokura’s partnership with Cumulus Networks affirms its commitment to supporting industry protocols that mitigate against proprietary networking protocols and operating systems that encourage vendor lock-in. Overall, the Midonet collaboration with Cumulus Linux promises dramatic improvements in the speed, scalability, performance and simplicity of networking infrastructures. Midokura and Cumulus Networks plan to offer a preview of their partnership in May 2014 and a general availability release in Q3 2014. The partnership between the two companies promises to push the envelope with respect to open standards in software defined networking in addition to improving the economics of building large-scale networks that encompass a combination of virtual and physicalized infrastructures.

Midokura Releases Upgrade To Midonet With Support For Layer 2 Gateway And OpenStack Havana

Midokura today announced the latest release of Midonet, its software defined networking platform for IaaS infrastructures. MidoNet now integrates with OpenStack Grizzly and OpenStack Havana and supports Layer 2 Gateway networking. MidoNet’s support of Layer 2 Gateway means that customers can now “connect bare-metal servers to the overlay network” and thereby add physical networks to its network virtualization platform. The latest release also features an enhanced management console marked by the ability to understand how a virtualized network would operate if it had experienced a physical installation. In addition, today’s release boasts an enhanced command line in its web-based user interface. Midonet’s distributed, decentralized, software defined network virtualization platform represents an alternative to the Open vSwitch (OVS) Plugin provided by OpenStack. In an interview with Cloud Computing Today, Adam Johnson, General Manager of Midokura, notes that Midonet claims competitive advantages over OVS such as improved security, efficiency, scalability and performance as it strives to become the networking solution of choice for the OpenStack community.

Midokura Announces Integration With CloudStack

Building upon its finalization of $17.3 million in Series A funding, Midokura recently announced the integration of its MidoNet technology platform with Apache CloudStack, the open source IaaS platform that Citrix submitted to the Apache Foundation in 2012. MidoNet is a software-defined virtual network solution for IaaS that leverages a distributed, decentralized virtual networking architecture. MidoNet’s value proposition consists in its ability to streamline the management of networking infrastructures for IaaS while reducing costs by means of the simplicity of the platform’s networking topology. Midokura CEO Dan Mihai Dumitriu remarked on the significance of MidoNet’s integration with CloudStack as follows:

We view Apache CloudStack as a leading open source platform, so we are pleased to offer a tightly integrated solution which gives CloudStack customers the ability to take advantage of the MidoNet network virtualization solution. The major benefits of the joint solution will be increased scalability, functionality and management of the network layer, along with a simpler approach to network design.

Dumitriu notes how MidoNet brings enhanced simplicity and scalability in addition to improved “functionality and management” to CloudStack and other IaaS platforms. MidoNet already claims integration with OpenStack, and hence its recent CloudStack announcement stands to widen its adoption and consolidate Midokura’s positioning as one of the leading software defined networking vendors for the IaaS space. As MidoNet’s adoption accelerates, the industry should expect to hear more in the way of customer success stories that underscore, elaborate and explain the uniqueness of MidoNet’s distributed, decentralized virtual networking platform in contradistinction to VMware’s acquisition Nicira.

7 Key Features Of OpenStack Grizzly

OpenStack Grizzly was launched on April 4, 2013 and contains over 230 new features that enhance the OpenStack platform’s computing power and ability to integrate with other technologies. The release was enabled by the collective work of 517 contributors who merged 7,620 software patches. More than 45 companies employed developers that contributed to the release including Red Hat, Rackspace, IBM, HP, Nebula, Intel, eNovance, Canonical, VMware, Cloudscaling, DreamHost and SINA. Given that OpenStack is used in production environments by the likes of Best Buy, Comcast, CERN, HP, NSA and Samsung, the Grizzly release focused on supporting the day to day operational work required to manage IaaS platforms based on OpenStack distributions.

Key features of Grizzly include the following:

•Support for VMware ESX and Microsoft Hyper-V hypervisors, whereas previously OpenStack had focused on the KVM and Xen hypervisors.
•Support for bare metal provisioning
•Enhanced OpenStack compute scalability by means of “cells” that manage distributed clusters alongside a design that reduces the centrality of a core database via a “NoDB” architecture
•Enhanced ability to automate the management of storage platforms by means of quotas
•Ability to manage diverse storage platforms from a central point of access
•Support for additional third party storage solutions such as Ceph/RBD, Coraid, EMC, Hewlett-Packard, Huawei, IBM, NetApp, Red Hat/Gluster, SolidFire and Zadara.
•Advanced support for software defined networking (SDN) that allows users to write rules to configure networking infrastructures from a broad array of virtualized networking platforms that now includes Big Switch, Hyper-V, PlumGrid, Brocade and Midonet. Previously, OpenStack had supported the Open vSwitch, Cisco UCS/Nexus, Linux Bridge, Nicira, Ryu OpenFlow, and NEC OpenFlow SDN platforms.

Jonathon Bryce, Executive Director of the OpenStack Foundation, remarked on the maturity of the OpenStack software development process as follows:

The Grizzly release is a clear indication of the maturity of the OpenStack software development process, as contributors continue to produce a stable, scalable and feature-rich platform for building public, private and hybrid clouds. The community delivered another packed release on schedule, attracting contributions from some of the brightest technologists across virtualisation, storage, networking, security, and systems engineering. They are not only solving the complex problems of cloud, but driving the entire technology industry forward.

Here, Bryce notes how OpenStack Grizzly features important enhancements across a wide range of attributes such as virtualization, storage and networking. Central to these enhancements is a significant increase in OpenStack’s ability to integrate with other virtualization, storage and networking vendors in ways that dramatically enhance the attractiveness of the OpenStack product given the inherent heterogeneity of enterprise IT platforms. Expect subsequent releases to add more orchestration and automation to the OpenStack platform, most likely beginning with the Havana release in October 2013.

Midokura Raises $17.3M In Series A Funding For Network Virtualization For IaaS

Today, Midokura announced the finalization of $17.3 million in Series A funding in a round led by Innovation Network Corporation of Japan with additional participation from NTT Group’s Venture Fund, DOCOMO Innovations Inc. and NEC Group’s Venture Fund, Innovative Ventures Fund Investment L.P. The funding is intended to expand Midokura’s product development teams and contribute to the refinement of its go-to-market strategy. Coincident with the news of Midokura’s Series A funding is the announcement that Dan Mihai Dumitriu, formerly CTO, replaces Tatsuya Kato as CEO. Co-founder Tatsuya Kato will assume the responsibility of Chairman of the Board. CEO Dan Mihai Dumitriu remarked on the significance of the capital raise and the company’s software defined networking technology as follows:

As enterprises and carriers embrace and build out IaaS clouds, an overlay-based network virtualization platform will soon be a must-have technology. The financial support of Innovation Network Corporation of Japan, and other key backers, validates our strategy as well as the work we’ve done over the past three years developing our industry leading product MidoNet. This funding will enable us to accelerate our product engineering, the establishment of partnerships, and the growth of our customer base. We look forward to delivering the most performant, scalable and fault tolerant network virtualization solution to the IaaS infrastructure market.

Here, Dumitriu neatly summarizes the concept behind Midokura’s Midonet platform, a distributed, software-defined virtual network solution targeted toward IaaS cloud-service providers as well as enterprises. Midokura plays in the space made famous by VMware acquisition Nicira, but is designed differently because of its decentralized, fully distributed architecture as elaborated by Brad Hedlund. The company is already a veteran contributor to the OpenStack community and is in the process of gearing up to announce further partnerships with other cloud platforms in the upcoming months according to Chief Strategy Officer Ben Cherian. Now that the company is flush with a significant infusion of cash, the industry should expect to hear more news about Midokura’s penetration into the U.S. market as its unique software defined networking platform increases its footprint in IaaS infrastructures and delivers simplified network infrastructures and protocols for data transmission in addition to high performance and high availability virtual networking solutions.