New features

First off, the default installation is for a virtualized environment. As Mark Baker, Ubuntu Server product manager at sponsor company Canonical, told me, this is because users are increasingly deploying the OS on hypervisors and Canonical wants to show off the OS’s capabilities there.

“While KVM has been big on Ubuntu since 2008, it’s not the only game in town,” Baker said. “We’re seeing customers wanting to understand integration or compatibility between ESX and Ubuntu, or even Hyper-V and Ubuntu, and we’re ensuring testing on these – and of course KVM and Xen — so when we are engaged with customers or users we can say we know Ubuntu provides a robust experience on the prevalent hypervisors.”

The other major aspect of this release is its integration with the new Grizzly release of OpenStack. Canonical has been involved with OpenStack since the start, and the release cycles for the two products are aligned (Grizzly came out a few weeks ago).

Ubuntu 13.04′s Juju orchestration “charms” have been updated to deploy OpenStack for high availability – for example, when the user deploys MySQL, the charm will set up 3 nodes in a failover configuration, and a similar approach applies to the deployment of the Rabbit messaging server. Of course, those deploying in a test environment won’t be too keen on running 2 or 3 of everything, so it will still be possible to install in a “less highly available way”, as Baker put it. The Juju GUI has also seen a lot of work this cycle “to improve usability”, he added.

Meanwhile, the Ceph storage subsystem is now fully integrated with Ubuntu and OpenStack, in order to please Canonical’s telco and service provider clients, and Ubuntu’s Floodlight OpenFlow controller has also been updated. Although Canonical and VMware are working closely on Nicira, “having an open-source alternative to Nicira is also important,” Baker pointed out.

Carrier adoption

Speaking of carriers and service providers, this is the market segment where Canonical appears to be thriving.

“OpenStack certainly has been the biggest growth areas for us in the last 12 months,” Baker said. “We have got engaged with the types of customers that we could only have dreamed of, looking back a few years. OpenStack is gaining adoption with carriers, and most people doing that to scale are doing that with OpenStack on Ubuntu. Most of the major telcos, the global names that you’ll see, are deploying their OpenStack on Ubuntu.”

Baker also claimed that OpenStack is seeing traction in the big data space, with users deploying Hadoop and Cassandra on Ubuntu – he suggested this may be out of “developer affinity” with the Linux distro.

“It’s fair to say the bread and butter of our user base is running web infrastructure,” Baker said. “A lot of that user base is moving that web infrastructure into the cloud. We’ve gained significant popularity on Azure – there is a fair proportion of that running Linux. While you wouldn’t think it a natural fit to provide Ubuntu on a Microsoft cloud, we actually think it’s quite exciting.”

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Cloud computing infrastructure: 2012 and beyond
• The promise of SDNs in the enterprise
• Cloud and data third-quarter 2012

Intel is showcasing its networking silicon, but it’s also offering two reference designs — one for new switches and one for new servers that would use Intel’s new chips. It’s also showing of a software layer called the Intel Data Plane Development Kit for OpenVSwitch that will accelerate packet processing on Intel’s CPUs instead of on dedicated network processors. One of the reference designs is aimed at the data center and the top of rack switches made by Cisco, Juniper, Arista and Force10 (owned by Dell), and the other is more of a punch for Cisco and Juniper in that it’s aimed at service providers. In fact, at the event Intel said Verizon was testing a prototype version of its reference design.

With these offerings, Intel is putting Broadcom, a popular maker of merchant silicon, on notice that it’s going directly after its business. That’s not surprising. The only question is how low Intel will go in pricing to put the hurt on Broadcom. But it’s also providing designs and capabilities that could obviate the need for special-purpose silicon that Cisco and Juniper currently rely on in their high-end boxes. Intel has gone after special purpose hardware before when it took on Sun and IBM in the server world with its x86 chips for personal computers.

Intel’s moves into the networking world are a symptom of the broader shift in computing. On the consumer side, mobility is changing the devices we use. In the enterprise, considerations of power consumption can now trump performance. In fact, super chips like the ones Intel traditionally sells can cause their own challenges in a virtualized world because using all of that capacity requires data center operators to virtualize the hardware and complicate their lives.

On the enterprise side the architecture to support our computing needs is changing as well as the workloads. The business considerations are changing too. This is a trend that’s shifting the ground underfoot all of the large IT vendors. So to see Intel going after its fellow chipmakers is perhaps unsurprising, but to see it going after Cisco’s and Juniper’s markets is like watching a rat resort to cannibalism during a time of starvation.

IT companies aren’t starving yet, but they are under stress, as Intel’s lackluster earnings indicate. The very real disruptions caused by a new generation of computing and web infrastructure are going to eat away at the margins and business models supporting today’s giants. Seen in this context, Intel’s moves aren’t surprising, but they are a symptom of the overall disruption in the IT world.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Cloud computing infrastructure: 2012 and beyond
• The promise of SDNs in the enterprise
• The promise of software-defined networking

He began with a rueful acknowledgment that back in the early 70s, when creating the addressing scheme for the internet, 32 bits were enough. The point of the story — we ran out of 32-bit addresses two years ago — was to illustrate how the common knowledge at the time influenced the architectural decisions the creators of the internet made.

Yet, 40 years later, the internet is still the valuable foundation of our communications infrastructure, and Cerf hopes that in building out this next generation networks we learn a bit from the creation of the internet. For example, he calls for the creation of open standards where differentiation doesn’t come from companies patenting protocols, but rather from branding their services or branding their unique implementations of a standard protocol. That’s because interoperability is important for building networks that are stable and resilient. As Cerf said: “Stability is your friend in networking environments.”

“If you want things to interoperate, standards are important,” Cerf said. “That’s not to say you can’t explore new ideas, but when you want something big to happen then you need to think about standards.”

In that same vein, Cerf also explained how as companies build out software defined networks they should consider the things that made the internet a success: the loose coupling of the gear that underlies the internet as opposed to a heavily integrated and brittle solution; a modular approach allowing new companies to develop solutions that might work between layers in the stack; and open source solutions, which are recommended but not required.

SDN can build a web for the future.

Cerf then went into some of the opportunities that SDN can offer to improve some of the shortcomings of the internet. For example, the current way we route traffic relies on the network having a physical port to send a packet to, but the OpenFlow protocol changes the destination address from a physical port to a table entry, which enables a new type of networking. One that might be more suited to the collaborative web we’re building today.

Content based routing also could be an option — something we’ve covered at our Structure conference in 2011. In content based routing you take the content of a packet and use that to determine what to do with it. It turns routing into something that’s closer to the way Twitter works as opposed to how the U.S. Postal System does. For example you would look at the content of a packet and route it to people who said they want to receive that information. It becomes multi-cast instead of a one-to-one connection.

As for the core tenet of software defined networking, separating the control plane from the data plane, Cerf said. “I wish we had done that in the internet design, but we didn’t.”

But that also means people can build new networks that resemble older networks while sneaking in revolutionary new features. Cerf is excited about the ability of those building SDN products and networks to mimic the core functions of today’s networks in order to drive adoption but then introduce something new like content-centric routing. Or perhaps they can implement better security to protect people from identity theft, from inadvertently becoming zombies in a botnet attack or from any number of security threats that exists online.

Cerf is confident that SDN can help address those issues and more. He envisions using SDN to perhaps define areas where people can access intellectual property in a controlled manner that may prevent people from making illegal copies. SDN might also be a way to bridge the divides between different networks today.

He pointed out that when the internet was developed researchers built different networks depending on the medium, so a mobile network and a wireline network today don’t look the same to software running over those networks. You can’t run traffic seamlessly across both at the same time. With SDN you could.

He closed with a few examples of how SDN is helping Google, from its implementation of an intra-data center WAN to using software defined networks to boost the utilization of spectrum through tools like Google’s white space broadband database. This example, as well as the idea of creating a unified network using different medium, has me really excited to see what Google might do with its own fiber network and a corresponding Wi-Fi network.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The promise of SDNs in the enterprise
• Cloud and data third-quarter 2012
• Takeaways from the second quarter in cloud and data

Last night I attended the opening cocktail reception for a radically different ONS and had the chance to reflect on how rapidly the once-staid field of networking is changing. There were about 1,500 people registered, which was the limit of the venue. The event had grown to the Santa Clara Convention Center and attendees were a fairly even mix of suits and engineers.

The biggest change was the exhibitor section. Where in 2011 the exhibitor hall was a narrow corridor at the Stanford conference center where a little more than a dozen students, companies and non profits had set up “booths” to showcase their ideas for Open Flow, there was now a few rows of booths — most of which were quite professional.

In October 2011, I attended the show for one day and moderated a panel where I recall asking Dave Ward, who was then CTO and Chief Architect of the Platform Systems Division at Juniper, what he would do if Stuart Elby, the VP of digital architecture at Verizon, a Juniper customer, got so excited about the promise of OpenFlow and SDN that it stopped buying expensive Juniper gear.

Ward danced a bit but essentially said that Juniper had the features and expertise to pull networking gear together that Verizon would pay for. The subtext (and knowing Ward, it may have been directly stated) was that he wasn’t an idiot and he was well aware that the networking industry was shifting. But his company would figure it out.

Six months later, the same conference had grown to 700 people and had Google showing off its own networking coup — it had built a software defined network using OpenFlow that connected its data centers. And Ward was still on a panel I moderated, only now he was at Cisco: preaching the same ideas but now at a company with the resources to carry it through.

Fast forward to the opening of the summit this year on Tuesday, and I’m eager to see what awaits. All I can tell is that so far the industry has moved from the excitement of translating a new technology into a commercial endeavor — one that scored a $1.26 billion transaction when VMware purchased Nicira — to one where use cases are more common and vendor fighting has started capturing a bit of the event conversation.

Indeed, mixed among the many case studies I’ve heard so far is speculation about the vendor-led Open Daylight Foundation that includes IBM, Cisco and VMware as strange bedfellows trying to build an open source controller for the software defined data center.

Just eighteen months removed from its inaugural event, software-defined networking has clearly learned to walk — if not run.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Cloud computing infrastructure: 2012 and beyond
• The promise of SDNs in the enterprise
• Takeaways from the second quarter in cloud and data

Last week Cisco offered an update on the beta customers using its new OnePK program to implement software-defined networking on its gear, and it’s an exercise in realpolitik that Kissinger would have admired, if Kissinger cared about networking hardware. I spoke with Dave Ward, Cisco’s CTO of Engineering (pictured above) about the network giant’s strategy, and he told me there are more than 50 beta customers using the Cisco software (and gear) to implement everything from load balancers to a cloud-based DVRs.

Cisco’s big plans tie the network to the data center

The how is more interesting than the what. Last year I wrote a post about Cisco’s strategy, saying it would focus on big vendors such as Cisco’s core service-provider customer base and that it would be open in as much as an exclusive nightclub is open. This is still true, although Cisco is supporting protocols such as OpenFlow in its gear and is also supporting all of the big hypervisors with its controller software. And remember, Cisco has a lot at stake here.

So let’s look at what Cisco has pulled together. There are two prongs in an enterprise SDN strategy — one inside the data center and one associated with traffic between data centers (or between the data center and an enterprise office). Cisco has both and has tied them together. For SDN on the wide-area network, it had purchased Carriden as the orchestration layer. Here it competes with vendors such as BTI or Alcatel-Lucent but also is solving problems similar to what Google solved with its homegrown OpenFlow network between data centers in 2011.

Inside the data center, where companies like Nicira and Big Switch are focused, Cisco has built something it calls the elastic service control. The software works on all of the popular hypervisors and can handle networking gear that doesn’t belong to Cisco. But because this is Cisco, and CEO John Chambers had warned that his plan to deal with the threat of SDN would rest in the ASIC, Cisco also had introduced new hardware containing specialty silicon.

Custom ASICs to save the day!

In many of its routers, the Cisco-design ASIC tracks data about the application running on the network and transmits that information to other routers containing that same chip. This hardware gives Cisco the ability to deliver service-level agreements and enforce rules for apps at the hardware level without IT intervention. Of course, other firms are trying to build this capability into software, but Cisco is trying to maintain its edge on the box market, so it makes sense that it wants to retain this capability in the hardware.

Its overall pitch is that is can basically wed the WAN virtualization with the data center virtualization to deliver exactly the resources an app needs when it needs it. This is a hot area with startups teaming up to offer similar capabilities or software to make this happen. The hardware containing its ASIC allows buyers with the right boxes at each end a level of control that Cisco hopes will justify the need to purchase its boxes — at least at some points in the network. And because Cisco knows that blindly shutting out all hopes of OpenFlow and cheaper gear would be crazy, it is supporting other protocols and even controllers through its orchestration software.

In Cisco’s ideal scenario, its customers will continue to buy Cisco hook, line and sinker and use it to deploy SDN. Undoubtedly some will. But what’s more likely is that customers, even those afraid to get their hands dirty messing around with complicated OpenFlow or open-source SDN controllers will buy Cisco gear like I might buy a Hermes scarf or a Gucci bag to accessorize my outfit from The Gap. Thus, they might pick up some boxes running merchant silicon and tie that into some Cisco gear.

“There isn’t going to be some flag day for a conversion to SDN,” Ward noted. And he’s aware that there will also be new and open protocols that Cisco will have to support. “It does appear that there are going to be more [protocols], and we don’t want to be trapped, waiting for them to mature,” he said. That’s why Cisco is going to support those protocols through agents on its OnePK platform. Customers will leave the engineering to Cisco, giving Cisco the opportunity to add value and keep its margins.

Will it work? That’s the $64,000 question.

However, much of this still seems like a stopgap measure that relies on service-provider customers deciding that they’d rather stick with Cisco than let their own network engineers build out products that might help lower costs. Cisco has already done a restructuring to help put its business in line with the new infrastructure realities, but my hunch is that it will have to continue adapting to a new sales environment (and style) as well as see margins erode.

Guido Appenzeller (left) and Kyle Forster of Big Switch

“They are copying the innovation that they see in the outside world, and are bringing it in,” Forster said. “But they are competing against much more open architectures that can contribute faster and offer more choice.” In Forster’s opinion, Cisco will find itself paddling like hell to keep up with the more open, software-oriented market and will fall farther and farther behind. Eventually, even the customers who want to do it themselves will see the benefits of switching over to more innovative hardware and software. Obviously, Forster is hoping that those customers will also pick up Big Switch’s software and support when they make their move.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The promise of SDNs in the enterprise
• Takeaways from the second quarter in cloud and data
• Cloud computing infrastructure: 2012 and beyond

SDN involves the abstraction of the network’s brains, as it were, from its hardware. This is analogous in some ways to server virtualization, in that it makes it much easier to build smarter systems on top of commodity hardware. Juniper’s take on this sees the network as four layers, namely forwarding, control, services and management — in Juniper’s vision, everything but the forwarding layer should be centralized.

Lin, who joined Juniper through the company’s acquisition of SDN specialist Contrail Systems, said at GigaOM’s Structure:Data conference in New York on Wednesday that federating the control function and eliminating “manual error-prone processes” would help the big data because:

“We’re seeing a huge opportunity here to reposition the role of the network as a curator of big data and make sure that role is easily exposed through abstractions of the network. The role of the network is interesting because the network is the only thing that’s globally pervasive and … uniquely knows a lot of the contextual information that is required to drive insights back into the system.”

Lin argued that this “rich context” would enable new types of business models such as collaborative data exchanges, without anyone needing to worry about the technology architectures involved. “The role of the network is changing quite heavily and the pace of innovation for hyperconnected data is really astonishing,” she added.

Check out the rest of our Structure: Data 2013 coverage here, and a video embed of the session follows below:

A transcription of the video follows on the next page

Go to page 2 (of 2) on GigaOM .

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Cloud and data third-quarter 2012
• Takeaways from the second quarter in cloud and data
• Cloud computing infrastructure: 2012 and beyond

It also marks VMware’s commitment to working with other players in the industry regardless of the underlying hypervisor and even networking protocols they might be using, as it seeks to move up the software stack in the data center.

VMware’s NSX software, which draws from the company’s existing vCloud Networking and Security software in addition to the Nicira offering, will let customers spin up virtual switches, firewalls, routers and other network elements in conjunction with hypervisors, regardless of whether they come from VMware. The NSX controller cluster will expand or contract customers’ virtual networks in harmony with changes in computing power that are reported by way of northbound API requests, according to a VMware blog post describing the many abilities of the new software.

NSX will ship in the second half of 2013, about a year after VMware said it would buy Nicira for $1.26 billion. Nicira has not named more than a handful of customers, large though they may be, and now its software-defined networking innovations could enter many more data centers.

Gains for VMware with NSX could negatively impact Cisco, F5, Riverbed and other vendors of network appliances that take time to deploy and can’t scale as quickly as modern storage and compute resources. It also puts VMware in competition with the ecosystem that Big Switch is trying to build around OpenFlow and its own controller software called Floodlight.

Big Switch open sourced Floodlight in January, and several startups and a few large companies are building products using that software. So while VMware is opening up to other hypervisors and promising customers that it will support a variety of protocols including OpenFlow, it is building its software-defined data centers play around a closed system.

That’s not a bad thing, and VMware’s ability to offer tested integration and the possibility of a more controlled experience will undoubtedly win over some customers, especially those already using VMware’s software. Meanwhile the more open approach from Big Switch will win its own adherents. With a $35 billion market cap, VMware has to not only keep innovating beyond its core business, but it also needs to protect its IP and margins while selling more software. The NSX product and strategy offers a way to do this. Now we just need to wait and see if customers bite.

Stacey Higginbotham contributed to this report.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The promise of SDNs in the enterprise
• Takeaways from the second quarter in cloud and data
• Cloud and data third-quarter 2012

Mellanox, primarily an Infiniband vendor, recently started gaining wins on the Ethernet side as well. Its history of providing networking gear and silicon for the highly-distributed and IO-intensive high-performance computing market has given it something of an edge in certain scale-out deployments. So the news that it’s trying to drive the creation of an Open Ethernet initiative makes sense. It’s listening to its customers and attempting to position itself as the replacement to traditional networking gear for the scale out and software-defined networking era.

According to an EETimes article, Mellanox won’t open source its firmware drivers for its chips and so will retain its proprietary edge on the silicon. Once again, it’s clear that server and now networking gear is getting stripped down to its most basic construction, where the primary hardware value is in the silicon and whatever software runs on top of it. On the server side, the value of that software has slowly been driven down by open-source alternatives, but on the networking side that process is just beginning.

Mellanox, like others in the space who have offered OpenFlow switches or APIs to their networking gear, has now announced its strategy for dealing with the threat open-source software poses to networking. Will it find partners and customers willing to buy into that vision?

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The promise of SDNs in the enterprise
• Takeaways from the second quarter in cloud and data
• Cloud computing infrastructure: 2012 and beyond

Research from Synergy shows that Cisco has the lion’s share of the market in switches and routing, reaching roughly 65 percent and 70 percent respectively. In 2012 the six main segments within the enterprise networking market generated $45 billion in revenues for technology vendors, with Ethernet switches now accounting for almost half of all spending.

So even as Cisco comes off a successful reorganization and faces existential threats to its networking business from the commodification of the router, it’s daunting to see what its fighting to keep. It will not go gently.

Check out Cisco at our Structure Data conference in New York City March 20 and 21.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The promise of SDNs in the enterprise
• Takeaways from the second quarter in cloud and data
• Cloud computing infrastructure: 2012 and beyond

The company has been providing wide area networking optimization products for telecommunications companies, optical gear and variety of other products in its 13-year history, but in conjunction with its funding it has launched a software-defined networking product designed to connect multiple data centers. In much the same vein of Google connecting its data centers using OpenFlow, or firms like NTT or Calligo connecting their data centers using Nicira’s software, BTI hopes to also help network providers make multiple data centers look and behave more like one.

BTI is offering a chassis-based product (it’s a big box) that customers put in their data centers network and connect via fiber to other BTI boxes in other data centers. BTI expects to announce customers using the product in the second quarter of 2013. The idea behind layering a software defined network between data centers is that it gives operators granular control on how they can route traffic between data centers based on customers and their service level agreements, but it can also lower costs associated with networking.

The promise BTI offers is that operators might not have to over provision to the extent they do today, because they can better manage their traffic and charge for bandwidth based on need. If packets don’t need to travel during peaks times, then the operator has the ability now to use pricing or service level agreements to move a customer’s traffic to less congested periods. To be clear, these customers are not consumers, but corporations that are buying bandwidth.

The vision here is for a telco-grade SDN offering for service providers and big content companies that own their own networks, but that don’t have the engineering talent or the interest in building their own boxes and code to do this. Despite the excitement around SDN inside the data center, using some type of software defined networking between data centers is actually gaining adherents just as quickly — if not more quickly.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The promise of SDNs in the enterprise
• Cloud computing infrastructure: 2012 and beyond
• The promise of software-defined networking