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

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

• Migrating media applications to the private cloud: best practices for businesses
• Infrastructure Q1: IaaS Comes Down to Earth; Big Data Takes Flight
• Cloud computing infrastructure: 2012 and beyond

The party line is likely to be that it was an overhyped technology that was supposed to displace Ethernet in the early 00s that flamed out. VCs lost piles on it, and quickly forgot. A few may recall Cisco’s acquisition of Topspin, or know that Mellanox has gone on the dominate the InfiniBand market from network interconnect cards, to switching silicon and switches.

As time marched on InfiniBand established itself as an interconnect of choice in many high performance computing applications, replacing propriety technologies like Quadrics, Myrinet, etc. InfiniBand even recently passed Ethernet as the most widely used interconnect in the Top 500 supercomputer list (see chart).

Riding the growth in HPC, Mellanox became a thriving public company, essentially a proxy for the entire InfiniBand market. Realizing the limited market of InfiniBand, Mellanox also added Ethernet to its portfolio a few years back.

But over the past two quarters things have changed as sales have accelerated dramatically at Mellanox. The chart below shows the rapid growth in InfiniBand revenue at Mellanox. Mellanox stock has soared accordingly.

The first hint that something interesting was afoot was Oracle’s major investment in Mellanox in October 2010. Oracle uses InfiniBand in its Exadata products among others. InfiniBand is a popular choice for low latency networking in storage back ends. But storage doesn’t seem to be what is driving the growth.

On its most recent earnings call Mellanox indicated that only 15 percent to 20 percent of its revenue was from storage. This means that while HPC and storage are steady and growing, cloud and Web 2.0 applications are driving the large uptick in Mellanox’s sales.

So why after all this time is InfiniBand breaking out of its HPC niche into broader commercial applications? Who is bucking the conventional wisdom of never betting against Ethernet and IP? Are the traditional Layer 2 and Layer 3 data center network architectures so broken that web scale commercial customers are jumping to InfiniBand? It seems so, at least for a few large customers.

InfiniBand does have some very interesting advantages over Ethernet. Its faster — the current generation is at 56Gbps — and offers lower latency than Ethernet. Current InfiniBand switches are also denser than the competing Ethernet options. But these factors have been true for most of the life of this technology, so why is InfiniBand taking off now?

My guess is that this is a mostly a temporal phenomena because of continued delays in the availability of cost-effective 40Gigabit Ethernet and 100GigE and the tremendous cost of building lightly oversubscribed Layer 3 data center networking with the traditional Ethernet/IP vendors. It also highlights the broader recognition that as applications scale out and become more distributed; networking latency has meaningful impact of application performance. Looks like counting microseconds isn’t just for high frequency trading anymore.

But Mellanox isn’t the only company to realize the rising importance of InfiniBand, the massive changes in the networking fabric and lessons that can be learned in HPC. Intel has seen that data center networking is getting in the way of their customers fully utilizing its powerful CPUs. To attempt to address this problem Intel has made three acquisitions in this market, Fulcrum Microsystems, the InfiniBand assets from Qlogic, and the interconnect technology from Cray. Intel buying InfiniBand technology is especially ironic because they were an early proponent of InfiniBand but dropped support in 2002 which was a seen a major setback for the technology at the time.

Whether this is a temporary blip, or InfiniBand is primed to take on a bigger role in the data center, it is fascinating to see how the networking and big data architectures from the HPC world which used to appear so foreign to the commercial market are beginning to make an impact. Even as VCs pour money into the networking sector, there are still opportunities for greater disruptions by taking learning from the world most demanding user in HPC and applying them in the commercial markets.

Alex Benik is a principal at Battery Ventures who invests in enterprise and web infrastructure start-ups. He doesn’t hold stock in either Intel or Mellanox. You can find him on Twitter at @abenik

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

• How tomorrow’s mobile-centric data centers will look
• The importance of putting the U and I in visualization
• Migrating media applications to the private cloud: best practices for businesses

Programmable networks could mean less downtime.

HP is following other big systems makers into the world of software defined networking with a line of 16 OpenFlow-enabled switches. That’s a pretty serious commitment to OpenFlow, a protocol that helps take the intelligence associated with routing packets off of the high-priced switching gear and puts it on commodity servers.

HP not only introduced OpenFlow enabled switches, but said that customers with existing HP switches can download software that will add OpenFlow capabilities to their current gear. This looks like a far bigger committment than IBM’s and NEC’s effort to build out a hardware and services package around OpenFlow and software defined networking from earlier this month, and is a continuation of the trend toward OpenFlow making it into production environments this year.

OpenFlow and software defined networking has been a topic for academics, webscale vendors and carriers, as they seek to do to routers and switches what virtualization did for servers — make them more agile and scalable. OpenFlow is just one tool to build SDNs while Juniper, Cisco and other vendors also offer tools for network virtualization. Of course, most vendors say they will support the OpenFlow protocol as well, including Cisco, the vendor that stands to be hurt the most if OpenFlow ushers in an age of folks buying cheap switches and shifting the networking intelligence to commodity servers.

As we add more devices to the network they have to scale out better, and as IT relies more on on-demand compute and storage, the networking has to become as flexible as the virtualized servers that spin up and down. The network becomes a bottleneck if every time you want to add capacity to your cloud or associate new networking policies with a series of virtual machines, someone has to manually unplug boxes or install new load balancing or firewall gear. Virtualization and software defined networks are seen as the solution.

HP said that so far it has more than 10 million OpenFlow-capable switch ports deployed, which is tiny number compared to the overall switch market. However, it’s not alone in pushing OpenFlow, and it has made quite a commitment with a full upgrade of its existing switches and 16 new ones on offer.

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

• The promise of SDNs in the enterprise
• The promise of software-defined networking
• IT spending update, third quarter 2012

Setting up your Roku

One of the biggest issues that people run into with streaming media devices is a slow or inconsistent Wifi connection. Big, solid walls, microwave ovens or cordless phones can ruin your Netflix fun, especially when you are watching in HD. If you have a Roku 2 XS, I’d strongly recommend that you hard-wire the device with an Ethernet cable. And if you don’t want to string a long cord through your living room, adding a HomePlug AV device to the mix might do the trick. Check the end of this article for more details on this kind of accessory.

Unfortunately, the Roku 2 LT, HD and XD devices don’t have an Ethernet port, so you’ll have to make do with Wifi. Try to move the router closer to your TV and avoid unnecessary obstacles if you run into issues, or upgrade to a 802.11n router if nothing else helps. Also, check out the video embedded below for more advice on how to improve your Netflix streaming:

Watching great content on the Roku

Roku’s media streamers offer access to a few hundred content channels, and hundreds more are available with a little bit of tinkering. Make sure to check out the following choices:

• Netflix is a no-brainer: The video subscription service now offers access to more than 57,000 movies and TV show episodes for as little as $8 per month. Check out our Netflix tipps and tricks to get the most out of your Netflix subscrtiption.
• Hulu Plus is a great choice to catch up on currently running TV shows.
• Amazon now offers free access to thousands of TV shows and movies, including Showtime shows like The Tudors and Fox classics like Arrested Development, to subscribers of its Amazon Prime shipping service.
• Roku also offers lots of live sports, religious services, video podcasts and other specialized content through its channel store. Most of these channels are free, but some come with subscription fees. But wait, there’s more: A few hundred additional private channels, including adult content and foreign TV networks, can be accessed through special access codes. Check out this post to learn how you add private channels, and then browse Roku-Channels.com to find all those private channels.
• Roku is optimized for online content, but you can also use the device to play your locally-stored photos, music files and home videos. Only the Roku 2 XS comes with a USB port to play content from a thumb drive or hard disk, but you can play videos stored on any PC or Mac in your home network with any Roku model. Check out the video below to learn how:

Note: The YouTube channel shown in this video is unfortunately not available anymore.

Buying accessories

Funny how good gifts make you want to buy more stuff, isn’t it? The Roku is no exception, and there are a few things worth considering:

• As mentioned above, nothing beats a hard-wired connection to stream HD content. HomePlug AV devices like the WD Livewire simply use your home’s electric circuitry to stream video from your router to your Roku, which makes them a great choice for any user of the Roku 2 XS. Check out our detailed review here.
• Roku’s media players aren’t just good for streaming, you can also use them to play games like Angry Birds on the big screen TV. The Roku 2 XS already comes with a motion-aware remote control, but new owners of the Roku 2 XD or HD don’t need to miss out either: Roku sells its motion remote online, albeit for a somewhat steep price of $30. So if you really want to play games on your Roku, but you don’t have the option to return the one you got as a gift and exchange it for a Roku 2 XS, this could be one way to go.
• The last suggestion may be a little more for tinkerers than your average Roku user, but owners of an EyeTV tuner can turn their Roku into a DVR that works in concert with a Mac in your home network. Check out the video below to learn how:

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

• When video gets democratized, who wins and who loses?
• Connected consumer third-quarter 2012
• Who and what to watch in the new era of the living room

That means rapid data transfers up of to 500 Gigabits per second (current Ethernet cables in data centers are generally 1, 10 or maybe 40 gigabits per second) and less mess with physical cables. Of course, every switch at the top of a rack would have to get a radio card slotted into it, and there’s also the matter of putting reflective panels on the ceiling for the wireless signals to bounce off of. The top of the servers would also need some kind of signal-absorbing surface so the signals don’t continually bounce around the data center. From the article:

To maximize the bandwidth and reduce interference between signals, it needs to be focused into narrow beams that require a direct line of sight between endpoints. “Any obstacle larger than 2.5 millimeters can block the signal,” she says.

One way to prevent the antennas from blocking each other would be to allow them to communicate only with their immediate neighbors, creating a type of mesh network. But that would further complicate efforts to route the data to the appropriate destinations, says Zheng. Bouncing the beams off the ceiling directly to their targets not only ensures direct point-to-point communication between antennas but also reduces the chances that any two beams will cross and cause interference. “That’s very important when you have a high density of signals,” she says.

While it sounds kind of out there, the researchers hope to build a prototype data center to try the idea out. Mark Thiele, the EVP of data center technology at Switch Communications’ SuperNAP data center, says the research is worth following as low-latency networking inside the data center can be a bottleneck today for applications that range from financial trading to trying to move gigantic data sets around.

The choice of 60 GHz for the data center is possibly inspired. Intel is one of several chip firms pushing 60GHz for consumer use, under the WiGig brand. This means the chips would be cheap. Additionally at 60 Ghz, signals deteriorate rapidly, which sucks if you want to transmit data over long distances, but is a boon if you are worried about someone standing outside the data center trying to eavesdrop on the data you are transmitting. However, using wireless, especially wireless with somewhat persnickety propagation limits (can’t travel far, requires line-of-sight between endpoints), means that data center technologists will suddenly have to learn a different type of network engineering, skills more familiar to their brethren setting up base stations in the cellular world.

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

• Will cloud computing push the BRIC market to the front?
• The Internet of Things: What It Is, Why It Matters
• The role of converged infrastructure in the data center

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

• Cloud computing infrastructure: 2012 and beyond
• Infrastructure Q1: IaaS Comes Down to Earth; Big Data Takes Flight
• Infrastructure Q2: Big data and PaaS gain more momentum

In the last year, Qwest had tiptoed into the on-demand services market by providing application recovery and interactive voice recognition on demand, but it wasn’t as large a player as Verizon with its Compute-as-a-Service products. And in January, Verizon said it would purchase Terremark, a provider of private clouds and data centers in a deal valued at $1.4 billion, saying cloud services were a big growth area for the company. Ancell agrees, although when I asked about the likelihood of a CenturyLink acquisition in this space, he replied, “Glen [Post] is an acquirer and has been the CEO for 19 years and has a track record of acquiring companies, and I would expect that to continue.”

Ancell says the merger gives CenturyLink scale and larger free cash flow, which could boost its deal-making capabilities. The company is trying to figure out what customers want from Infrastructure-as-a-Service providers and what they want as software. Qwest had 17 data centers that it brings to CenturyLink, which as of the end of 2010, reported $172.94 million in cash and $2 billion in operating free cash flow. It does have more than $7 billion in debt, however, and none of that includes Qwest’s contribution.

Aside from a possible rush to the cloud, CenturyLink closes this deal with 190,000 miles of fiber, after adding 75,000 miles of Qwest local fiber and 30,000 miles of Qwest backbone fiber. As it seeks to serve businesses and data centers, the biggest constant is bandwidth demand continuing to grow. Ancell echoed what I had heard recently about larger corporations wanting to upgrade to 100 Gigabit access from 10 Gigabit pipes — completely skipping over the 40 Gigabit access. He cautioned that not everyone was skipping 40 Gigabit access, but that the awesome demand for video and other applications was causing companies to look hard at their upgrades. With 40 percent of network traffic being video, the demand is continuing to rise, he said. “People just need more and more bandwidth at rates that almost have us saying, ‘What is going on?’”Ancell said.

Image courtesy of Flickr user Torkildr.

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

• In Q3, Big Data Meant Big Dollars
• Cloud computing infrastructure: 2012 and beyond
• Infrastructure Q2: Big data and PaaS gain more momentum

According to the Berkeley Lab, the amount of data emerging from climate research and being accessed by scientists is growing “exponentially.” The “next-generation archive” of climate modeling data is expected to reach 650 terabytes or more, up from the 35-terabyte archive now maintained by a program at the Lawrence Livermore National Lab and used by some 2,500 scientists. Long term, the ESnet team aims to build a network that can handle as much as 1 terabit (1,000 gigabits) of data per second.

Where’s this data coming from? Computer modeling can be used to study things like storm patterns over hundreds of years, helping to put today’s weather patterns and changes in a much broader context. That was the intent of researchers out of Pennsylvania State University who published a study this week finding that the North Atlantic Ocean has had more big storms over the last decade than any other time in the last 1,000 years.

In addition to analyzing physical evidence (layers of sediment, for example), the researchers fed into their computer model more than 150 years’ worth of information about “factors known to influence hurricane activity,” as ClimateWire explains, and then simulated 1,500 years of Atlantic storms. With projects like the one being funded at Berkeley, this kind of data crunching could move much faster, and in the meantime, provide broadband gear companies with a valuable customer and test bed for high-speed networking.

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

• The fourth quarter of 2012 in cleantech
• The role of converged infrastructure in the data center
• Cleantech and investment in 2013

Cole Brodman, chief technology office of T-Mobile USA, in a recent GigaOM interview said that the company is currently providing 6 Mbps per site. “Tomorrow I think the first steps are going to be something more like 20-25Mbps, quickly followed by 50Mbps, and eventually getting to 100Mbps-plus,” he said. T-Mobile isn’t alone in its scramble to bulk up the backhaul as according to some forecasts, there will be more than a billion mobile broadband phone subscribers by end of 2010. While data has been a good way for the wireless operators to scoop up easy cash, they’re all facing a future in which they’ll be spending a lot of money on infrastructure. According to market research firm Infonetics Research:

• By 2010, data traffic will surpass voice traffic on mobile networks.
• Mobile backhaul equipment investments jumped a healthy 19 percent in 2008 to $4.6 billion worldwide, and revenue is set to explode over the next five years and beyond.
• Demand for mobile backhaul equipment is seen topping $10 billion by 2011.

Why? Because as In-Stat, another market research firm, notes, carriers worldwide will need some 90,000 Gbps of capacity in the last mile of the backhaul network by the end of 2013 to support the world’s cellular and WiMAX networks. The big boost will come for the backhaul Ethernet-based equipment, whether in the form of microwave, fiber or copper. Microwave will be a big winner, Infonetics predicts. Roughly 60 percent of backhaul cell site connections in most regions of the world are microwave, the firms estimates, and demand is growing fast, including here in the U.S.

Related Posts:
* 4G Wireless & the Ensuing Bandwidth Boom
* Verizon Rents Out Its Fiber for LTE

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

• CES 2012: a recap and analysis
• 12 tech leaders’ resolutions for 2012
• Connected world: the consumer technology revolution