SDN architectures

Japan Has an SDN Startup, Too

Midokura is a startup that says it's doing what Nicira Networks is doing, only for Layers 2 through 7. With a description like that, you wonder why VMware Inc. (NYSE: VMW) didn't buy these guys instead.

But then you find out Midokura has only three U.S. employees, and their office -- when they bother to use one -- is in an unused conference room inside a space leased by DreamHost, a Web hosting company. The connection? Ben Cherian, Midokura's chief strategy officer, used to work there.

Not quite the same as the clout of Nicira, which has the cachet of SDN luminaro Martin Casado and a founding team that did the original work on OpenFlow. (See VMware to Buy SDN Startup for More Than $1B and VMware Insists It's Not Warring With Cisco.)

Aside from some contributing work to OpenStack, Midokura has sailed under the radar. Most sources contacted by Light Reading (all of them in the United States, admittedly) had barely heard of the company, if at all.

But the company was founded in 2009, so it's not as if it's trying to cash in on the sudden SDN craze. Rather, Midokura reflects the depth of the data-center scaling problems that are helping SDN become so popular.

The Amazon of Japan
Cherian is not the face of a Japanese company. He's a bright-eyed, clear-spoken emissary of the Web 2.0 world, fitting comfortably in the hip San Francisco startup world.

And Midokura's U.S. arm is very much like a startup. The San Francisco team started out in a spare corner of a Rackspace facility, but as Rackspace grew, Midokura got squeezed out. Now they're working from home and borrowing space from Dreamhost, Cherian's former employer. "We're cloud couch-hoppers," Cherian says.

Midokura started out because founders Tatsuya Kato and Dan Mihai Dumitriu wanted to create the equivalent of Amazon Web Services Inc. for Japan. But in planning out the company, they realized that the biggest roadblock was going to be the network.

Amazon isn't telling how it got around those problems. Without a simple guidebook to copy, Midokura's team decided they would tackle the problem of network scaling and make that the company, selling a platform to cloud providers.

Dumitriu provided some of the necessary background. He went to Cornell and worked at Reliable Network Solutions -- and at Amazon. Working from that background in distributed systems, he decided to assemble a team to tackle the problems of network scaling.

To cite one specific but overused example -- Layer 2 VLANs are great for connecting one virtual machine to another, but you're limited to using 4,096 of them. Technologies such as QinQ purport to fix that, but then other problems come up. "Where do you end up trunking those QinQ things? Do you end up trunking everything on every machine? You end up with a very complex networking arrangement," Cherian says.

The company has raised $5.8 million from Japanese venture capitalists -- Sunbridge Partners is the only one listed on Midokura's website -- and other investors. It's raising another round that could help double the head count, to 40, Cherian says.

Reaching Layer 7
Midokura's product, still not in 1.0 version yet, is called the Midonet. It's software, distributed among a data center's servers, can create a virtual switch or firewall or any other network element without changing the physical network.

Those network elements target Layers 2 through 7, and that scope is unusual, Cherian claims. Nicira is working at Layer 2 and isn't working on Layer 3 yet, although Cherian predicts that will change with the help of VMware. Embrane Inc. is creating virtual network elements for Layers 4 through 7 -- SDN-based firewalls and load balancers, for instance -- but doesn't include the switching and routing of Layers 2 and 3.

One catch: Midokura hasn't only gotten to Layers 2 through 4 so far. "Midonet 1.0 will be a polished version of what we do today. It might include Layer 7. Our team is working hard to figure that out," Cherian says.

For now, though, Midokura is "very much like Nicira, pushing the ability to go up in the stack," says Dante Malagrino, CEO of Embrane. He adds that Embrane's customers in the U.S. or Europe haven't ever mentioned Midokura.

Midokura is making some progress in Japan, though. It's gotten revenue from some proof-of-concept projects, including one with NTT Communications Corp. (NYSE: NTT), Cherian says.

— Craig Matsumoto, Managing Editor, Light Reading

Vishnu Goel 12/5/2012 | 5:24:49 PM
re: Japan Has an SDN Startup, Too

In the network virtualisation space NICIRA really created flutter by isolating the hardware and software platform elements and now getting acquired by VMWare really early in the game! However the like companies need to create unique value by changing strategy somewhat.Perhaps this Japanese startup need to outdistance others by incorporating the Test & Management (T&M) virtualisation as part of their offering.It can then ensure Telecom and many other markets more deeply and sharply.Vishnu Goel T&M +919810101238

Pete Baldwin 12/5/2012 | 5:24:49 PM
re: Japan Has an SDN Startup, Too

Plenty more SDN startups are going to pop up, I'm sure. What interested me about Midokura was the way they got started, trying to be a cloud provider -- then realizing that someone needed to fix the networking scale problems they were anticipating as a would-be cloud provider.

In other words, it shows there's a legitimate reason to be working on this stuff.

SDN still might not take over the world, but it's still interesting to see the ideas develop.

patentchoi 12/5/2012 | 5:24:47 PM
re: Japan Has an SDN Startup, Too

Going from a cloud provider to a data-center networking virtualization solution provider is a long way and I not sure what circumstances caused them to realign. Anyways good luck to them.

There is very little info on their website about what load-balancers, firewalls etc they use. Is this all open sources, How easy it is to manage them, etc.

Given that they are unheard of, it seems like a long haul for them.

Pete Baldwin 12/5/2012 | 5:24:46 PM
re: Japan Has an SDN Startup, Too

Nooser -- One thing I didn't make clear: Midokura never actually became a cloud provider. They wanted to, but they discovered this technology gap they'd have to overcome... and they decided to make the gap their business.

They seem to have started with the right skill sets for it.  So, I'll give them the benefit of the doubt there.

I agree they've got a long ways to go, although if they can at least get Layers 3 and 4 nailed down before Nicira does, that'll give them some differentiation. They'll also have to grapple with Embrane, which is stocked with some of the former Cisco "spin-in" crew (who, come to think of it, are now going to be competing against those people again, sort of).

There is indeed very little info about Midokura's specific technology; they're still prepping for the product stage.  I found their story intriguing, anyway. It'll be fun to watch all these companies develop.

patentchoi 12/5/2012 | 5:24:46 PM
re: Japan Has an SDN Startup, Too

It will be useful to hear from the well established cloud service providers, those L4-7 virtual network services for which it is important to have elastic capacity allocation. Is it just the load-balancer and firewall or is there anything else beyond that?

t.bogataj 12/5/2012 | 5:24:20 PM
re: Japan Has an SDN Startup, Too

Hope they succeed... But how can anyone trust a startup with a mindset from ten years ago? Cherian's "Where do you end up trunking those QinQ things?" was a valid question before 2005, i.e. before 802.1ad (Provider Bridging).

Asking this question today means that either they are not familiar with the Ethernet progress of the last decade, or that they willfully ignore this progress. Neither qualifies them as a trustworthy SDN vendor.


Sign In