10 network virtualization, SDN and data center companies to watch
After years of venture capital and start-up focus on consumer-focused tech companies, enterprise networks and the data center are back in vogue.
But we're not talking sexy Vogue. We're talking virtualization, cloud computing and software-defined networking (SDN) - the sorts of technologies IT managers need to run more efficient data centers and better support increasingly mobile workforces and customers.
"It's an area that hasn't had any innovation in decades," says Jamie Goldstein of North Bridge Venture Partners about the networking industry. "Vendors are still selling the same Ethernet switches they were in the 1990s. Maybe the interfaces are faster, but it's all basically the same technology. Until recently." SDN is a fundamental change in networking, he says, which is causing excitement in the venture capital community.
[MORE STARTUPS: 12 Hot cloud computing companies worth watching
Established vendors such as Cisco, Juniper and Alcatel-Lucent have all aired their SDN plans of late, but it's venture-backed companies such as VMware's $1.2 billion baby Nicira and Stanford University-borne Big Switch Networks that have grabbed more of the attention in this emerging field. And scads of newcomers are right behind them.
In our latest package on companies to watch - vendors that collectively have raised some $250 million in funding - we profile SDN upstarts as well as those in related markets such as virtualization and converged infrastructure.
-Headquarters: Santa Clara
-Focus: Virtualization management
-Product availability: HotLink SuperVISOR for VMware vCenter and HotLink Hybrid Express for VMware vCenter are both generally available.
-Funding: $10M from Foundation Capital and Leapfrog Ventures
-Management: HotLink CEO Lynn LeBlanc and Chief Science Officer Richard Offer co-founded FastScale Technology, a software provisioning platform that VMware acquired. LeBlanc has worked at a variety of eventually-acquired startups, including iReady, ILEX Systems and Atherton Technology
-More information: HotLink
An emerging truth about modern data centers is that they are heterogeneous. They run multiple operating systems, applications and, increasingly, hypervisors. (One survey found that up to 20% of VMware users have deployed the Microsoft Hyper-V hypervisor in their data centers, with Red Hat's KVM and Citrix's Xen Server each gaining market share against VMware's dominate hypervisor market position.)
But the downside to all this: "Heterogeneity is typically synonymous with complexity," says HotLink CEO LeBlanc.
The vendor's SuperVISOR product is designed to simplify management of heterogeneous hypervisors via a software plugin to VMware vCenter. HotLink's recently launched Hybrid Express platform allows vCenter users to control Amazon Web Service's public cloud resources.
In effect, Hotlink's software tricks VMware vCenter into thinking that Hyper-V, KVM or Xen-virtualized machines are running on VMware hypervisors. A variety of management platforms allow users to deploy virtual machines across hypervisors, but HotLink goes beyond just that and allows users to migrate workloads across them, too. "No one else is really doing integration to this degree," says Bernd Harzog, an analyst at The Virtualization Practice.
HotLink gives IT administrators the power to set access controls, track benchmarking and use, make changes to, configure and manage the instances across clouds.
The company has an impressive list of backers. Its advisory board includes CIOs from some of Silicon Valley's biggest names, including from Facebook, Citrix and EA. Partners include VMware, Red Hat, Microsoft and Citrix, makers of the four major hypervisor platforms.
Flush with venture funding and actual customer-generated revenue, and fresh off winning a best in show award at VMworld 2012, HotLink is now looking to broaden its products' integration with additional management platforms and to increase their functionality.
Headquarters: Santa Clara, with offices in Lyon, France.
Focus: Application-defined networking for the cloud
Product availability: Public beta of CloudWeaver launched in January; general availability expected in early 2013.
Funding: $3.3M in series A from Idinvest Partners, a pan-European mid-market private equity firm.
Management: Pascale Vicat-Blanc, founder and CEO, is former research director at INRIA (French National Institute for Research in Computer Science) and former CIO of a Grid 5000 Data Center, as well as project manager at CERN.
More information: Lyatiss
Lyatiss is taking a slightly different approach to software-defined networking and instead is focused on application-defined networking.
"Cloud networks are unpredictable and complex; traditional networks are blind to the applications running on them and rigid," says CEO Vicat-Blanc. "There's a missing link there."
Lyatiss's recently launched CloudWeaver product is an operating system that runs on Amazon Web Services' cloud that makes networks application-aware. It has a series of collectors that monitor application demands in real-time through an API, collects the correlating network requirements and provisions the network to those specifications. If one application needs a low-latency connection for speed, the network will automatically be configured for that. If another app needs high bandwidth for a certain type of traffic, the network can be optimized for that, too.
Lyatiss was born out of the publicly funded French technology research institute INRIA where Vicat-Blanc studied high-demanding applications. Researchers monitored the virtual networking developments in U.S. academia, as well as the market adoption of cloud computing, and saw an opportunity to link the two with network controllers.
"There are a lot of business opportunities in this market," Vicat-Blanc says. "Networking and the cloud used to be separated, but now there's a convergence and there's a lot to be done to make them work together." Users will not embrace SDN technology if it isn't practical for them to use, and making it practical is about ensuring that the network can respond to the needs of the applications running on it, she says.
Headquarters: San Francisco
Focus: Network virtualization for cloud service providers
Product availability: MidoNet software is in beta now, general availability in Q1
Funding: $5.5M in seed funding from angel investors Bit-Isle, and institutional investors NTT Investment Partners, MUFJ Capital and 1st Holdings.
Management: Tatsuya Kato, co-Founder and CEO, was CEO of mobile Internet company Cybird and a former executive at Japanese systems integrator CSK Holdings; Chief Strategy Officer Ben Cherian was GM of emerging technologies at DreamHost; Dan Dumitriu, co-founder and CTO, was an Amazon Web Services architect.
More information: Midokura
The founders of Midokura started the company with one bold goal: To create a public cloud service in Japan that would rival Amazon Web Services. But if anyone could do it, it may be these guys. Members of the team worked with Amazon CTO Werner Vogels to help create Amazon Web Services' distributed Dynamo database system.
But when the Midokura team began planning for the massive data center infrastructure that would be needed, they ran into a big problem. Networking, they found, was not ready for the cloud. To get to their goal, they needed to solve the networking problem first, which is why they created MidoNet software. When testing the technology they believed it could be as big of a business opportunity as creating an AWS in Japan, and have been focused solely on their networking operating platform since.
[MORE MIDOKURA: Japanese virtualization start-up invades the USA]
MidoNet software centrally controls all aspects of an IaaS cloud network, from the switches to the routers to x86 servers to firewalls to any other piece of commodity hardware. It centrally analyzes traffic at the very edge, as soon as it enters the network. Each aspect of a Midokura network is linked, which the company says allows the software to route traffic directly from the entry point to the destination in the most efficient way possible. In doing so, it's "blowing up the router," says Midokura's Cherian.
"Since you're abstracting away the devices from the network controls and pushing the intelligence of the network to the farthest edges, the physical infrastructure is still there, but you're not requiring as much from it," Cherian explains. Switches and routers, he says, simply aren't designed to handle the massive amount of changes that occur within a cloud network. MidoNet is designed for the dynamically scaling network requirements of a cloud. Initial deployments of the system work with OpenStack-powered clouds.
Headquarters: San Jose
Focus: SAN-less storage and compute clusters
Product availability: Generally available
Funding: $71.6M from Lightspeed Venture Partners, Blumberg Capital, Khosla Ventures, Battery Ventures and Goldman Sachs
Management: Founder and CEO Dheeraj Pandey, former vice president of engineering at Aster Data (now Teradata), managed the storage engine group for Oracle Database/Exadata; vice president of operations David Sangste, formerly with EMC, Data Domain, Hewlett Packard; vice president of engineering Bill Culman, formerly of QuorumLabs, Interwoven, BEA Systems and Tandem Computers.
-More information: Nutanix
Efforts by big name vendors such as Cisco, VMware and EMC (via their VCE venture) to combine data center elements such as compute, storage and networking into one plug-and-play system have been mildly successful to date, according to industry analyst Arun Taneja. What Nutanix is doing is different though, what Taneja calls "hyper-convergence."
Nutanix sells a hardware-software combination that combines compute and storage, meant to allow for easy scalability to a massive size with simple management. In 2011 the company won a best in show at VMworld in the virtual desktop category and investors have already poured more than $70 million into the company. Nutanix says in 30 minutes its system can be installed and IT admins have a data center in a box, ready to scale just by adding more nodes.
The key, says CEO Pandey, is the notion that data stored in Nutanix is written locally and stored redundantly across however many nodes are included in the deployment. That allows for easy horizontal scaling - just add more boxes to increase capacity, while not degrading performance. Nutanix is built on Intel Sandy Bridge processors and the VMware ESX hypervisor, but its Version 3.0 software released in December expands support to Red Hat's Kernel Virtual Machine (KVM). Automatic redundancy across the system means virtual machine-level disaster recovery capabilities, while the VMware platform allows for vMotion - the transferring of virtual machine instances - across nodes within the system.
Nutanix is playing in a crowded field with some heavy hitters. Along with VCE, IBM has Pure Systems and NetApp has FlexPod. Taneja, the analyst, says those are "pseudo" convergence efforts - disparate pieces assembled together and optimized to work collectively. "Hyper-convergence can only happen if you start afresh," he says, which is exactly what Nutanix has done.
Headquarters: Palo Alto
Focus: Hardware-independent SDN switching systems
Product availability: OpenFlow-enabled Open Switches are shipping now
Funding: $6.5M Series A from Vantage Point Capital in October 2012
Management: Co-founders James Liao (CEO) and Lin Du (vice president of engineering) were with Woven Systems; other execs from InfloBlox, Dell and Alcatel-Lucent.
More information: Pica8
Even though the SDN market is young, there is already a confusing variety of vendors and products on the market. Where's an interested buyer supposed to start?
Pica8 executives want to simplify the process. The company, which was founded on the basis of being software-focused and hardware-independent, has released a series of reference architectures of how to get going along the SDN pathway - of course using Pica8's network operating system.
An SDN architecture has three main components, explains Pica8 marketing executive Steve Garrison: A physical switch, a controller and a virtual switch. But there are already dozens of options for customers to select from for each of those components, leading to a long onboarding process of typically between 90 to 100 days, he says. Pica8's published guides show how customers can set up and deploy these virtual networks using Pica8 components and commodity hardware supplied by Quanta Computer, ensuring that customers are not locked into buying expensive proprietary equipment.
Pica8's software, which includes the Open vSwitch (OVS) agent from Nicira, also works with open source controllers from Big Switch Networks, although a variety of other OpenFlow-based controllers, such as Ryu, can be used as well. The idea, Garrison says, is that as SDN adoption continues, "no one controller will be best for every application... no one company will own this market."
Headquarters: Cambridge, Mass.
Focus: Application-driven network control
Product availability: Became generally available in Q4 of 2012
Funding: $48.5M from North Bridge Venture Partners, Matrix Partners, Lightspeed Venture Partners
Management: David Husak, founder and CEO, formerly with 3Com and co-founder of Reva Systems; Ephraim Dobbins, co-Founder and vice president of engineering, formerly with Acme Packet.
Fun fact: The ecommerce portion of the company's website includes items such as a sold out bobblehead of company Husak, as well as a $30 million Plexxi jet.
More information: Plexxi
Plexxi executives say much of the talk about SDNs and virtual networks these days is off the mark.
Instead of just re-architecting the network to make it more elastic, scalable and controlled through software, they say the network should be orchestrated from the top down, by applications telling the network what resources they need and the network automatically provisioning itself. That's what Plexxi says its switches and software combination enable customers to do. "We view ourselves as building a solution based on an SDN foundation," says Mat Matthews, a co-founder and vice president of product marketing.
The software layer collects information about applications, analyzes what network resources it's using and creates algorithms to predict what will be needed. It takes that information and relays it to the underlying integrated network hardware devices made by Plexxi - mostly switches - that configure the network to the specifications of the applications.
Plexxi has already seen some success in the financial sector, where organizations are looking for ultra low-latency private cloud infrastructures. The company's system is ideal for networks that include a high number of server connectivity points, and for content delivery and big data applications, according to Plexxi. One early customer has created a premium service cloud computing offering that guarantees no two servers are ever more than one hop away from one another, for example.
The company is made of proud Bostonians. On its website, CEO Husak shows his East Coast pride, noting: "The last time there was a shift of this magnitude in enterprise networking, Boston-based companies led the way. And, it's going to happen that way again."
Headquarters: Sunnyvale, Calif.
Focus: Network virtualization-plus
Funding: $10.7 million in Series A funding from US Venture Partners and Hummer Winblad Partners
Product availability: First half of 2013
Management: Lot of former Cisco employees and engineers, including Co-founder and CEO Awais Nemat, Co-founder and CTO Pere Monclus, Head of Product Management Valentina Alaria and Vice President of Engineering Lele Nardin
More information: PLUMgrid
PLUMgrid has issued an open challenge about why the company is called what it's called. When executives reveal details about the company's plans during a launch event in the spring, CEO Nemat plans to hold a drawing for anyone who guesses right and the winner will take home an iPad.
Nemat is also secretive about exactly what the 50-employee company is up to, but has shared some clues. He promises it will be a virtual networking advancement. The problem with many existing solutions, Nemat says, is that they're not able to work on top of existing network gear, requiring a "rip and replace" model.
PLUMgrid's technology, described as an I/O control system for connecting networks, storage, compute and applications, "lives on top of servers as a software layer that protects existing investments, while also providing a path for greenfield investments," Nemat says. PLUMgrid has decided not to be OpenFlow-based, and is instead developing its own network controlling software.
Headquarters: Palo Alto
Focus: Hardware-accelerated network virtualization for public/private clouds
Product availability: Limited availability of NetVisor software and Server-Switch hardware, with general availability in Q2
Funding: $42 million from NEA, Menlo Ventures, Mohr Davidow Ventures.
Management: CEO Robert Drost and CTO Sunay Tripathi, both formerly with Sun; Vice President of Engineering Ken Yang is on sabbatical from UCLA
More information: Pluribus
While working at Sun as chief architect for storage and network virtualization of Solaris, Pluribus CTO Tripathi kept running into the same problem, as did Sun customers. "Coordinating between multiple networking components is really hard to do," he says. Between the compute virtualization layer, the persistent environment and the different operating systems, hypervisors and applications on each server, it is really difficult to centrally control them.
Tripathi (who holds 80 patents, many related to Solaris, but a variety in virtual switches, too) left Sun to solve this problem, and for the past three years has been building a team to create its Netvisor network operating system and Server-Switch hardware. Similar to how the x86 server ushered in an era of compute virtualization, Tripathi says a new era of networking chips from manufacturers like Fulcrum, Intel and Trident will allow increased programmability of switches. But they still need an operating system to control them, and that's what Pluribus provides.
Tripathi says he started Pluribus before the hype around SDNs heated up. Even though there are a bunch of startups in this emerging field - just look at the rest of this list of companies to watch - Tripathi says Pluribus's Netvisor system can work with next-generation networking switches and controllers from the likes of Nicira and Big Switch.
VC Jim Smith of Mohr Davidow, which has backed Pluribus, says a couple of things stood out to him about this start-up, including the team and its devotion to offering a complete package. "We're already seeing the pain-points for this next-generation networking technology," Smith says. "It's clear there will be a new control layer to manage these super-complex compute and networking environments we're building for the future. Pluribus Networks is one of the few companies we've seen that offers a complete fabric for handling that complexity."
Headquarters: Westborough, Mass.
Focus: Data center in a box
Product availability: Limited availability of OmniCubes today; general availability in March.
Funding: $43 million from Accel Partners, Charles River Ventures and Kleiner Perkins Caufield and Byers.
Management: Chairman and CEO Doron Kempel, founder and CEO of Diligent Technologies, and with IBM and EMC before that; Vice President Engineering Brian Nadeau is a former Dell executive director; Vice President Marketing Tom Grave was director of product management and marketing for IBM's storage division.
-More information: Simplivity
CEO Kempel doesn't have the traditional background of a start-up tech exec: He holds degrees from Tel Aviv University in law and philosophy and served as a major in the Israeli Defense Forces before pursuing a technical career.
That amalgamation of experience is emblematic of what SimpliVity brings to the market: A unified data center system that combines best-of-breed technologies. It offers an alternative to a common data center rack filled with possibly hundreds of thousands of dollars worth of servers, shared storage units, backup and deduplication appliances, a WAN optimization box, a cloud gateway and more.
Simplivity's OmniCube 2U appliances, which work with VMware software, each boast up to 40TB of usable capacity, 12 cores and 800GB of memory. They start at around $55,000 and if users need more scale, they can add OmniCubes, creating a highly-available OmniStack that can be dispersed geographically and managed centrally.
A key driver behind the OmniCube is data management. Kempel and his team came from Dilligent, a de-duplication startup that IBM bought in 2008. Using that experience, the company has integrated deduplication and compression into the OmniCubes for all data at inception and throughout all layers of the OmniCube. This eliminates the need for separate WAN optimization tools and speeds the connections for data requests, the company says.
Headquarters: Santa Clara
Focus: Geographically distributed relational database management
Product availability: TransLattice Elastic Database (TED) launched in 2010, Version 3.0 was released in February
Funding: $9.5 million, led by DCM Ventures
Management: CEO and Co-founder Frank Huerta and CTO Michael Lyle, both formerly with Symantec, Recourse Technologies and Exodus
More information: TransLattice
Let's say an organization has multiple sites across the country or world, and has a database accessed by workers in all those locations. The database needs to be synchronized so that everyone has all the same information, no matter where they are.
The cloud might seem like one option: Spin up a relational database instance via Amazon Web Services, right?
Not quite says Louise Frank, marketing vice president for TransLattice. AWS database instances do not automatically replicate across sites, so changes made in one availability zone of Amazon's cloud will not appear immediately at another site. TransLattice databases do.
A software and hardware appliance from TransLattice sits at each site where an organization has a relational database and automatically synchronizes the nodes across locales. It uses sharding techniques that split up the data into bite-size chunks as well as caching technology to get the job done.
The company was born in part out of the vision of CTO Lyle, who before co-founding the company served as an entrepreneur-in-residence at venture firm DCM. He kept hearing pitches from companies doing WAN optimization, load balancers and database optimization tools. "All those individual components just make the system more complicated and they don't actually solve the problem," Funke says. So Lyle - who had previously served as CTO of distributed security/threat management firm Resource Technologies, which was later acquired by Symantec - left the VC world and began pursuing the idea of a globally-dispersed database system.
Now, TransLattice is humming along, having recently announced the third generation of its TransLattice Elastic Database (TED), which now spans across Amazon Web Services and Dell public clouds to ensure high availability and resiliency. The company has also rolled out the TransLattice Application Platform, which applies the same technology used to synchronize the databases for customers to host applications at sites across the globe.
Read more about data center in Network World's Data Center section.