David Wang - CEO Insights
A Dynamic Network
How does having a fully software driven network improve efficiency, reliability, and real-time dynamic control?
Software is everywhere. It's everything. And you can do so many more things than
what we can do manually. So, in the network today, software is almost everywhere.
But, the physical connectivity aspect, which is fiber, which is cabling, is the fundamental
part of the network. Every network is the connections. Because a connection is done
physically, which is slightly beyond what today's software can control. But, the question
is, why not? If you have the right technology, we see a cost effectiveness and it will
enable the software functions that will bring lots of the benefits. And this will also enable
lots of the newer applications. Now you can have a program to control physical layer
connectivity, to tie together lots of other functions in other layers of the network. And
fiber connectivity can be a full networking function, which is really exciting.
How important is it for key innovators to partner in the creation of the dynamic network of the future?
Collaboration actually creates the innovation. But the mindset being an innovator, in our case, is we're tackling one of the top problems in a very fundamental layer of the network. And the fundamental reason to automate is to enable many other layers and applications in the network. So, collaboration is key. So, what we're doing should be to enable services, new revenue stream efficiencies, and to enable others to offer lots of these capabilities to the network operators, customers and the enterprise. This is really important. You're in the ecosystem and it's also important to collaborate and enable each other, which will simply create even more innovations.
Automation for the Future
In what ways can data center operators plan for integrating automation into their workplace from a technology standpoint?
Automation has happened in many different areas. In data centers, you see many different layers of the network. One critical element is actually missing without automation, which is a physical layer. If you look at the operation in data centers, there are many aspects in terms of energy, power monitoring, redundancies and so on, but a fundamental function data center performs is cross connections between peers, between carriers and customers and service providers. And that's really today's only automated part of the process in the data center. When you have an environment where one link is broken from a speedy and software defined automated process, that really slows down fundamentally the pieces. Today, without automation in the cross-connection space, it would take several days to get a services activated.
That also means that for several days, the revenue and the customer service is not up. If you look at the opportunities and the overall views of how data center operations can be improved, the key aspect is to achieve 100% automation. Then the process, the service delivery, will flow from customer to application from partner to partner, all through software, all in today's modern network environment. When you can turn services on instantaneously, then you can reduce the operational cost. You can increase the customer happiness and the service delivery capability and improve utilization of equipment, resources, facilities, thus generating more revenue. And also, fundamentally it will tremendously close a loophole on security. Now, you have a full software-defined operation, but also a software driven audit trail with data, to close any physical manual access that breaches the security process. This will ensure you have full compliance, with whatever the customer regulations require. So, if you look at the overall picture right now, the physical layer needs to be addressed to bring the entire data center to a fully automated operation.
What are some of the challenges you would bring up to other CEOs when discussing the future of automation if they were interested in the benefits, such as scale, speed and productivity?
To have the innovation such as automation to happen in mission critical environments, it takes decision making from an operational, business, service delivery, finance economic, and a process standpoint. So, how do you make it happen? We touch on many levels from our customer side, both from the network planning and from the technology selection process. We look at which operations, software integration processes, pricing model economics, finance processes, technologies or innovations will make sense and which steps are needed that will cause the least disruption and bring the most economic benefits.
There's always a hurdle for CEOs to look at innovation and something that's new and to look at their decision making in an effort to take the company to a bigger step, accounting for the risks involved. Some companies tend to lead in the marketplace and some tend to be more conservative and try to make a more secondary decision when something new is coming to the market. It is important to be more educated and get more comfortable and use other people's learning experiences as a reference. A big part of my job is to share and to help, to not just educate, but help brainstorm, learn, interact, and understand our customer's constraints and to see what are the things they will need from us.
What makes us unique is we touch on many customers in a lot of countries and different
kinds of players. From a process standpoint, different companies may take different paths
for change. Outside of any sensitive information that within our knowledge and our experience
we always try to help a customer to learn from our experience and to learn from some of the
early mistakes that either someone else or we have made. We help the decision makers to
mitigate risks and learn from our experiences, including those with other customers. We use
it to help the CEOs and the next customer’s decision makers, and help them to make the
most informed decisions.
It’s not whether you're going to do the innovation, do the automation or not. It is about how you do it, where you start and how you do it with the least amount of risk and with the best amount of returns and impact. It’s also important to know what do you do with your operations team, what is the software integration and how are you going to mitigate. Sometimes there is a time-lapse. Automation doesn't happen immediately. How do you shorten it to bring the benefit quicker, with more impact and with less hiccups and hurdles throughout the process. We're working to make our part of the process a lot simpler, both from technology and from deployment, as well as from pricing model standpoint, to make the decision making hurdle, a lot lower. The world can start evolving to more of an advanced network with software capability to the physical layer, with a fully automated operation much sooner, enabling the overall financial benefits much, much sooner. The big part is that we're mitigating, facilitating, informing and sharing, to help the CEOs in their companies to make the best decision, moving forward.
The New Era of the Data Center
How are we going to manage it more efficiently, and what type of new technology is needed, from a hardware, software, power, infrastructure standpoint?
The data center, the infrastructure has gone through many evolutions and revolutions. But I think we're about to achieve a major milestone that our operations in these communication facilities are becoming fully automated, which will become a tremendous trigger for a lot of things. For example, if a facility is fully automated, it does not require human access, then the environment requirement can be dramatically changed to save energy. So for example, temperature range, that does not have to be fully compliant to human safety or space requirements or height requirements and the safety levels that are required for day to day human access. So, you potentially will be building something in entirely different form factors that will be more space efficient, energy efficient, and cost efficient. Right now I think many of the facilities in telcos, data centers and any other type of network space still require human handling for the physical layer, which comes down to patching cables, making connections, and physically using a human factor to make a cable connection to a port, a device or to connect to another interface. There's a lot of imagination and possibility that will be unleashed by this tremendous trigger, which is achieving full automation.
It’s becoming more apparent that processing and compute is more effective at the edge for the next generation of services such as 5G, gaming, and streaming. How do we change the mindset of building giant giga-data centers to more smaller edge like data centers?
It's an evolution. It's not a revolution. So the giant data center has been built for more than a decade, that fulfilled tremendous, computing need for our enterprise applications, consumer applications, mobile applications, and is enabling compute and data processing in the cloud. The giant data centers today are our biggest infrastructure for today's computing and digital applications, in many aspects. The fact that we're able to be able to do a video conferencing, with relatively good quality, but on a very intense global basis is because of this infrastructure. Now, I think this allows us to start moving further with how our technology can enable more real time applications and how the application can be more efficient. This means that cloud computing and mobile devices have enabled a lot of the data processing in many different ways to get more personalized, mobile, and much more interactive in real time. These also bring in high bandwidth demand and high quality data, with live streaming information.
Edge computing started becoming much more of a need, to reduce the latency
and the cost, which will facilitate data to not have a much longer round trip but
into many more remote data center locations. We started really getting down to the
physics, the speed of light, the speed of electronics and signals going over a distance.
This will exceed our latency requirement, which today on 5G specs, and data
processing will need to meet a sub one millisecond latency spec. So this is a new
evolution in many ways and it will be a new revolution that's extending our compute to
be able to do much more on a local basis. It's a collaborative infrastructure at least
from the world technology migration standpoint to start becoming more local, more
modular, more on the edge, close to the users, to the enterprise, with much lower
But still using lots of the data that's centralized into these major data centers. I do think this trend eventually will raise a question mark for how efficient we're building data centers. That starts giving us a choice probably as a second step and a comparison, in terms of should I build more modular or efficient data centers, more towards the edge, versus continuing to build the gigantic data centers? I do personally believe the big gigantic data centers have a limitation in terms of energy supply, space supply and cost efficiency. But, it also requires other technologies to really start becoming the enablers. So for example, silicon photonics, and for example, much more host to host communications and data processing such as AI processors. That may not require nearly as much as the electronics and that will help to reduce tremendously the footprint a large data center will require.
The reason you move away from a very large data center is that on a per computing unit basis, you're more cost efficient to build something in a remote, more cost effective area with a low power cost, low labor, low land costs and construction costs, than to build a million square feet building. But the technology will start to bring much more efficiency in terms of computing. Much more optical level, the quantum level computing is also on the horizon, but it's still a little far. But I think the photonics level revolution has already started happening. And, at the AI level, the computing styles in terms of the CPU, GPU now we're seeing a set of new processors that does better pattern recognition with an AI algorithm, to be able to do computing much more efficiently.
And while they're doing that, the connectivity, which today is probably the most convoluted part of any network inside data centers and outside will eventually get condensed because the real connectivity is a piece of wire at the end. And our packetized network requires a lot of electronics, backplanes, transceivers, compensation modules, DSP modules and so on. Those are the components that consume a tremendous amount of energy and I think they will come down and they will give us the opportunity to build a better computing data center in a smaller footprint. So, the world is evolving and I think edge computing is a good direction to start getting data and compute more close to the user, which will bring more efficiency.