This is a continuation from Part 1.
SDN and NFV
SDN (intra–data center)
I asked Iwata his views on SDN and NFV. He said that in a narrow scope SDN and Openflow are synonymous, while in a wide view SDN includes automation and management of networking and data centers, including such things as orchestration and overlays. He added that different companies will answer the question differently. A software company like VMware may look at SDN from a purely software point of view, while Google claims that SDN deals with both software and hardware.
SDN (inter–data centers)
Once SDN for intra–data center traffic is implemented, we would like to apply SDN to inter–data center traffic. Can it be applied to general traffic on the Internet? Iwata replied that traffic crossing different Internet providers cannot be easily controlled. He thinks SDN for inter–data center traffic is mostly implemented via leased or owned lines. This is because common SLA and policies across different providers are very hard to implement.
Network function virtualization (NFV) is an attempt to virtualize some networking functions (software defined) so that nonspecialized networking equipment can be used for multiple purposes with different software packages. With this, scalability and swappable modules in software can be applied to add/modify functions without replacing equipment for those new purposes. Iwata mentioned that NFV is currently sought after both by telecom carriers and data centers. At data centers, such things as firewalls and load balancers are implemented as appliances and made available.
Iwata summarized by saying that:
- NFV implements networking functions.
- SDN defines how these implemented functions are connected and made into a system.
But if we look at the combination of the two from an altitude of 30,000 feet, aren’t we looking at a software-defined cloud? Iwata was cautious and pointed out that both terms, software-defined and cloud may be redundant.
How the Internet, data centers, and clouds are changing
We continued to talk about the current status of the Internet, data centers, and clouds. It is well known that the Internet backbone is a combination of a set of networks owned by big companies. They are called tier-1 providers. Until the late 1990s, there were thousands of Internet service providers (ISP) of many sizes. But the major ones were acquired by telephone companies.
Tier-1 providers interact with their tier-1 peers at peering points. In addition to public peering points, there are private peering points between tier-1 players. This is the classic view of the Internet. So those big boys can control how Internet traffic flows. Each packet may stay with one provider or be pushed to other providers at their discretion.
However, this setting seems to be changing. For example, a major colocation provider, Equinix, interconnects directly with Amazon’s AWS cloud, Microsoft’s Azure, Google’s Cloud Platform, and Softlayer’s clouds. When we consider the proliferation of clouds and their services, most traffic is to and from clouds. If you co-locate your computing power at Equinix, you could easily access clouds.
Source: Equinix website. Figure by the author.
When we consider that the ultimate destinations of packets are clouds, Equinix’s architecture makes a lot of sense. I wonder whether carriers are going to sit around and wait for Equinix and cloud providers to eat their lunch. I wonder whether telecom carriers (AT&T, market capitalization $177B) might consider acquiring a company like Equinix ($17B market capitalization).
Iwata is also working in the IoT space. More specifically, he researches how IoT is placed in an environment of SDN/NFV by paying attention to network edges (edge/fog computing), including mobile edges (mobile edge computing or MEC). This means fog computing. In other words, he is interested in comping up with user application use cases along that line.
Being a member of the Industrial Internet Consortium (IIC), NEC as a whole looks at both the industrial and enterprise sides of IIC. Iwata himself looks at the enterprise side.
Iwata seems to have blended well into Silicon Valley culture, and he feels very comfortable with the thick band of experts readily available. He did not want to discuss the details of his team’s mission, but it is concerned with the IoT enterprise space.
Talking with Iwata, I wondered whether Japan could provide an environment like Silicon Valley’s. Only talent matters and anyone with innovative ideas and execution abilities could succeed. Japan is changing but may not change quickly enough. Iwata and his team are definitely an exception to the typical Japanese team.
The following is a set of NEC press releases on relevant subjects.
Biometrics and IoT