From the blog.

Managing Digital Racket
The more I tune out, the less I miss it. But that has presented me with some complex choices for a nuanced approach to curb
Complexity – My Friend, My Enemy
Over my years of network engineering, I've learned that the fewer features you can implement while still achieving a business goal, the better. Why? Fewer

Friday News Analysis: Arkin Net, Pica8, OpenDaylight, Plexxi, Juniper

1,983 Words. Plan about 13 minute(s) to read this.

Arkin Net Rises from Stealth with $7M in funding, to bring Google Search-like Simplicity to Software-Defined Datacenter Operations

“Arkin Net is coming out of stealth mode with $7 million in funding, led by Nexus Venture Partners, to revolutionize the Software-Defined Networking (SDN) market – projected to be $8B by 2018. Arkin Net is building the industry’s first multi-vendor SDN operations platform that will transform how enterprises operate their software- defined datacenters.

With a consumer approach, the Arkin Net platform will enable enterprises to analyze, visualize and share operational data as easily as a Google search. This fresh approach brings unprecedented visibility across the physical and virtual layers, breaking the traditional IT silos.”

I saw an early demo of this product back in August 2014. My first impressions were positive. The interface is intuitive, and data is presented in a modern, easy-to-consume way. I feel like I’m underselling the interface a bit putting it this way. It’s very clear to me that Arkin Net takes GUI design seriously, and they have a knack for taking a complex SDDC environment and presenting the data in a way that’s easy to understand for operators, yet doesn’t sweep all of the important details under the rug. I think we’re going to hear a lot from Arkin Net – they are filling in a management gap in the industry, and what’s more, I don’t think they’re too early. Showing up right on time.

Read the complete press release here.


Pica8 Introduces CrossFlow Networking

“With this unique capability, customers can run Layer-2 / Layer-protocols and OpenFlow on the same switch at the same time. With CrossFlow, users can leverage OpenFlow for diverse, policy-driven applications to drive business logic into the network, while leveraging their network topology and Layer-2 / Layer-3 protocols like OSPF and BGP for the most efficient packet transport and performance.  Some sample use cases include:

  • In a traditional data center that uses L2/L3 switching and routing, monitoring and tapping can be done on the switches with rules triggered by the OpenFlow protocol.
  • In an OpenFlow data center, the edge devices need to interact with traditional switching/routing devices (using spanning tree, OSPF, BGP, and other Layer-2 / Layer-protocols)

Available in PicOS 2.4, CrossFlow allows every port in the bare metal switch the ability to act as either a legacy or a CrossFlow port.”

Unless I’m missing a key element of this announcement, I’ve heard the technology Pica8 is describing “CrossFlow” as “hybrid switching” in other contexts. The big deal is that Ethernet switches don’t always let you run a switch both as an OpenFlow switch and traditional switch at the same time. Often, you choose one mode or the other. Other switches might let you select on a port-by-port basis what mode that individual port on the switch will be run in. Will the port’s forwarding table be governed by a central controller using OpenFlow? Or will the port’s forwarding table be governed via distributed control-plane protocols like spanning-tree, OSPF, BGP, etc? If you’re exploring SDN and investigating the options OpenFlow offers, this is a big deal to know. Especially in a brownfield deployment.

With CrossFlow, Pica8 is giving consumers flexibility. If you rack a Pica8 switch running PicOS, the use-case doesn’t have to be greenfield to make sense. You can run some ports as traditional, and some as OpenFlow. But AFAIK, this is not a ground-breaking feature. Some other OpenFlow switches already have this capability.

Read the complete press release here.


Compass-EOS, Intracom Telecom and Transmode Join the OpenDaylight Project

“The OpenDaylight Project, a community-led and industry-supported open source platform to advance Software-Defined Networking (SDN) and Network Functions Virtualization (NFV), today announced that Compass-EOS, Intracom Telecom and Transmode have joined the project to advance open SDN and NFV by collaborating to build a common framework for network programmability. Company membership in OpenDaylight now totals 45 with over 250 developers participating in the open source community.”

OpenDaylight continues to grow both in size and diversity. Momentum is sustained for the project. More companies joining means dedication of coding resources to the project, which is good for the project. In my opinion, ODL is the single most important SDN project going on in the industry right now. Yes, it’s early days yet, but ODL is only going to matter more and more going forward. Consider that Extreme, Brocade, and others have announced SDN controllers based on ODL. Also, Cisco’s XNC controller (which is not ACI/APIC related at this point) is tied to ODL as well. If you don’t want to pay attention to any other SDN project, pay attention to ODL.

If you’re more of an enterprise networker, these 3 new companies might not mean much to you. They don’t mean much to me – out of my wheelhouse. But in summary (as best as I can tell), Compass-EOS is in the hardware business  – photonics. Intracom is a telecom service provider. Transmode sells optical networks to telecoms, HQ in Stockholm. That’s interesting because it means ODL is attracting ALL the different industry verticals. Also, ODL is attracting companies outside Silicon Valley, home to the world’s most insular echo chamber. In my mind, that’s a fabulous thing. Everyone is getting their heads together to try to make ODL an even more useful tool in the SDN world. Cisco ACI, HP VAN, and VMware NSX all have their partner ecosystems, but they aren’t putting together the sort of lovefest ODL is. At least, not yet.

Read the complete press release here.


Plexxi Names New CEO to Lead Company into New IT Era

“Plexxi, a pioneer of next-generation networking products, today named former top EMC executive and serial entrepreneur Richard Napolitano as CEO to continue the company’s expansion and help drive the networking industry’s transition to an application- and data-focused era powered by the explosive growth in mobile, social media, cloud and Big Data.”

Plexxi’s focus has been on hardware Ethernet switches connected in a variety of ring topologies. The frontside of the switches are Ethernet that you’re familiar with, but the ring interconnect between Plexxi switches is the unusual bit. The ring is a lightwave multiplexed backbone that allows multiple data paths in the same piece of fiber. This allows a logical “any to any” connection between switches in the ring, for any-to-any up to 11 switches, last I knew (although you could make much bigger rings than that, including stacked rings). It’s been a while since I heard a briefing – those numbers could be different, but that’s not really my point at the moment. The hardware is interesting, but is more a platform for their software, which is where I believe Plexxi’s strongest value proposition lies. Plexxi can do continual deep traffic analysis and optimize traffic patterns across their fabric based on a number of criteria.

Plexxi’s technology is fabulous – genuinely ground-breaking stuff that is application-centric and highly automates the network, making the best use of what’s there. The over-provisioning most of us as architects do isn’t the only answer, in other words. There’s much more to Plexxi’s tech than I can discuss here, but the rub is that Plexxi isn’t exactly taking over gobs of market share. Although the fact is that the Plexxi platform is far more than simple Ethernet switching, it’s brutally true that the most visible part of the product is an Ethernet switch. And that’s a tough market to get a foothold in. In the last few years, Ethernet switching has gotten ever more competitive, with a number of whitebox entrants adding cost pressures to this busy space, as well as Dell shouting loudly about “open networking” and Cisco adding yet more switch models to the crowded segment.

In short, it’s hard to be heard, and this is the challenge I see Napolitano needing to address. Let’s assume that Plexxi’s real strength is in their software. Recognizing flows traversing the network and providing a programmatically defined path is one of SDN’s most immediately useful value propositions, and Plexxi has some of the smartest people in the business (literally Ph.D’s from MIT) writing the algorithms that address it. When this is more clearly understood across the networking consumer space – and maybe when the algorithms stretch their legs beyond DC and campus, as could be happening with their ODL participation – Plexxi’s name will be one certain conversations can’t leave out.

One more thought – with new CEOs often comes staff churn, for a variety of reasons. Something to watch out for. 

Read the complete press release here.


Juniper Networks Redefines Networking Industry With First Carrier-Grade Virtual Router

“The Juniper Networks® vMX 3D Universal Edge Router, which operates as software on x86 servers, gives service providers and enterprises the ability to seamlessly leverage the benefits of both virtual and physical networking so they can rapidly deliver services and cost-effectively keep ahead of customer demand.”

When you read this announcement in its entirety, the hyperbole is nauseating. You’ve been warned. New products rarely redefine anything, and this is no exception. Virtual appliances have been around for a while, and in essence, that’s all we’re talking about here. Short version – the MX router is available as a vMX now. Caper about, hug your cat, wave arms in delight, fist pump joyously, have another slice of pizza, etc. Whatever response you think is appropriate.

Now, I’m genuinely keen on this idea. I use the MX platform today for Ethernet routing services. Assuming I can use a generic Ethernet switch to terminate the physical connections coming in from carriers and pump the data up to the vMX via 802.1q, I’ve just eliminated a certain amount of wiring, power draw, and RU requirements from my rack. That comes at a cost of eating some RAM and CPU in the virtualization cluster, of course. But assuming I also get the benefits of fault-tolerance, vMotion portability, snapshots, etc. that seems like a worthy tradeoff. Instead of running 2 MX routers for redundancy to minimize exposure during an outage, I could run one vMX (in theory, I’m making several assumptions) while still having all the benefits of a fully redundant physical router design.

Thinking through this vMX announcement a little further, this is a fascinating move for Juniper to make in that they’ll brag all day about the Trio chipset in the MX hardware platforms. But now they’re running the MX code on x86? There will be x86 scale issues for some applications, no question. But for someone like me who is running at gigabit or lower throughput rates? I bet the vMX is a perfect fit. What if I wanted to spin up several small vMX instances, rather than doing the multiple routing instance model I follow today? Oooh – nice choice to have if the price is right. Can’t wait to get one in the lab. Which, incidentally, is a scenario Juniper cites in this whitepaper.

FWIW, I believe I heard Juniper was getting 160Gbps of throughput max on a vMX. But to achieve that number, it owned the entirety of a big multi-core system. I don’t recall specifics of the hypervisor or bare metal. I’m sure LOTS more detail will be forthcoming. But the shift to x86 is something that customers are asking for. I know I am, and I don’t even live in uber-big cloud land where the unicorns run freely.

Read the complete press release here, but put on your anti-hyperbole propellor beanie first.


For the PR Folks…

If you’d like me to analyze your news, be sure ethan.banks@packetpushers.net is on your distribution list. I focus on the following areas:

  • SDN & NFV for data center and enterprise
  • Open networking
  • Enterprise & data center networking
  • Ethernet fabric, switching, standards, and optics
  • Application delivery controllers
  • Visibility fabric
  • Network virtualization
  • Network management
  • Network automation
  • Public/private/hybrid cloud
  • Wireless LANs

Other topics (like security and storage) are secondary for me; I am less likely to pay attention to them.