Blog Post

In a distributed world cache is king. Why routers are becoming the new server.

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Much like Amazon(a amzn) sparked a revolution in not just corporate IT but also in how companies build out their infrastructure (you could probably credit VMware(s vmw)’s virtualization software with a role in this as well), the rise in connected sensors and devices will rock the world of information technology again.

Cisco(s csco) calls this architecture shift “fog”, while others (including me) are desperately trying to some up with something else to describe what’s happening and why. The short rationale for this shift is that, as we connect more devices at the edge, we need a highly distributed architecture that encompasses not just data centers but the edge nodes, with applications that can handle the distributed compute and the databases that such an architecture requires.

In the last few months, two companies have both hit upon a way to handle this shift — and they are doing it with the realization that in this brave new connected world the network is a potential bottleneck. Their solution involves turning the humble router into a way station for data, for content and for real-time processing that can’t afford to make the trip up to the cloud.

To paraphrase the old Sun ads, the router is becoming the computer.

Yesterday Cisco showed off its new architecture for a world with connected devices trying to communicate real-time data to the cloud. Designed for the “fog,” Cisco’s IOx architecture will be a Linux-based operating system that will be embedded in forthcoming industrial routers.

And unlike its previous box software, Cisco says it plans to open the IOx architecture up for others to run their own applications on, but that will remain to be seen. Cisco sums it up in its release like this:

To better monitor, manage and respond to the massive amount of data generated by IoT, IOx allows applications to run as close as possible to the data source and create automated responses that drive value.

Cisco isn’t alone. In the residential world, Qualcomm(s qcom) has a similar vision for the humble router, turning it into a content cache, and a computer set to manage a house full of connected devices, including running algorithms that govern which devices things can talk to and when they might have access to the internet. Late last year, Qualcomm introduced its internet processor chip to be the brains behind this router/computer combo, and at CES it showed off a content caching concept with Akamai. The idea here is similar to Cisco’s. That many connected things will need a local compute and cache of data and applications to function optimally.

That two companies that have strong ties in the networking world have settled on the router as the source for both computing and caching of content or even databases closer to the edge makes sense. Other ideas might involve hubs that can handle many radio protocols or plain old servers, but I think given the types of mesh networks and myriad sensors that our homes and manufacturing floors will soon have, conscripting the router to handle both networking and computing might make a lot of sense.

This is a major theme of our Structure conference in June, as we deal with the implications that sensors and connected devices will have on our IT architectures. We’ll have to rethink databases and possibly even our concept of the network stack to deal with a world where there are both more protocols at the physical and transport layer, as well as a variety of ways for devices to communicate instructions and state to each other.

It’s going to be awesome.

18 Responses to “In a distributed world cache is king. Why routers are becoming the new server.”

  1. “The world isn’t just flat, it’s exponentially flat”-John Chambers, CEO of Cisco. Why does anyone listen to this stuff anymore? So routers are going to do more? Just put some network cards in a linux box. Oh, Cisco is going to do that for you, and it will be secure, and easy for Everyman to use! What lies ahead?

  2. foggycloud

    “the network is a potential bottleneck”

    it´s 2014. this has been known since 2001 u know… but the hipsters in sv are still incapable to code. what was needed was some kind of smp, but they only can create buttons in ruby. they used the net to scale, here is you bill.

    now even cisco makes jokes of them all, to call it the fog.

  3. Richard Bennett

    Incidentally, security is a bigger challenge for IoT in the home than network architectures and bandwidth. IPv6 makes this stuff somewhat easier, but it took the easy way and refused to address any of the Internet’s long-standing issues with security, identity, mobility, and resiliency. That’s what happens when you leave Next Gen architectures to the JV.

  4. Richard Bennett

    This is all pretty cool stuff. When I made home routers in 2006-8, carriers were interested in developing plug-in architectures to make a market for home control & security applications, and they over-spec’ed CPUs to ensure that enough processing power was available to do the job. The shift from desktops to laptops means you need a dedicated device in the home to run all that stuff, and the router and set top box are the best candidates.

    Relying on the Cloud doesn’t go very far because there are too many technical and regulatory issues, such as net neutrality, that cripple the development of advanced services, but you can do all the prioritizing you need on your own home network.

    I doubt that IoT needs massive bandwidth; a lot of it is things like the fridge announcing its current temp and door status. But forecast of exafloods and that sort of nonsense always get people excited.

  5. Andrew Lochart

    The cloud folks think it will all be in the cloud and the networking folks want to push some smarts to the network, naturally. But more of the smarts, e.g. caching, should really be on the endpoint device to truly optimize performance and reduce network traffic. It will interesting to see if the device makers, the OS companies, or the ISVs take the lead.

  6. Still missing the real need.

    Connected Homes are increasingly beginning to look like small datacenters from a decade ago with dozens of connected devices with needs for local segmentation (read VLAN) and QoS at the switch (My wife streaming the DVR to the bedroom should not kill a motion sensor reporting an issue to the security controller), along with things like multiple Wireless Access Points instead of single Points of Access or stupid repeaters that kill throughput and deeper security mechanisms and even potentially IDS capabilities. Other than improved throughput and some USB ports on the back, these systems have barely moved an inch from even 5 years ago.

    Cisco has no play as they have been rapidly selling off most of their last mile assets like Linksys and no interest in bringing their considerable expertise and IP to bear on the space because the margins suck.

    Also, I have no desire to have most of these new Internet enabled things directly connected to the Web and have any dependency on that connectivity to provide their functionality, so Cisco suggestion of Edge caches that are still outside the local network simply do not cut it from my PoV.

  7. Rick Bullotta

    Hardware is one thing, but the real value will be in the software that runs at each point in the network. The requirements go way, way beyond plumbing (security, routing, etc.) and there is a definite need for richer application level functionality at all layers in the IoT. It’s an area we’re doing a lot of work in here at ThingWorx, enabling intelligence to reside on devices, gateways/routers, and distributed in cloud and on-premise servers as needed.

    • Rick, that’s true, but I think many assumed that the smarts/software would primarily live in the cloud. At least a few years ago. But that’s changed. I am super excited about the concepts of database caches near the edge for things like identity or policy.

  8. Art Fewell

    Would be a good play for cable providers to offer enhanced services on one hand for an upsell opportunity, but perhaps more importantly as the younger generation consumes less and less traditional media and internet service is generally less than cable or a package, They need new ways of getting the younger generation to pay more to keep revenue streams up. I think an industrial play here is good, trying to sell direct to consumer could only go so far as, the ISP could raise strong barriers to competition for this device category at any point if the business were lucrative/strategic enough to them.