We stand on the brink of two huge advances that will define the next-generation data center. One is a new architecture for Internet applications, where microservices running in containers on cloud infrastructure vastly improve hardware utilization and accelerate app dev and deployment. The other is the proliferation of big-memory analytics and machine learning, which will enable self-optimizing infrastructure, enhanced security, and better business decision-making.
Hyperscale Internet giants like Google, Twitter, and Facebook have already put those technologies into production by building their own solutions and reaping the benefits. But for containerized microservices and ubiquitous analytics to trickle down and transform mainstream enterprises, a new array of solutions must be developed.
At first blush, a networking and server giant like Cisco seems an unlikely company to take up that cause. But in an exclusive interview with InfoWorld last week, Zorawar Biri Singh, who joined Cisco as CTO in July 2015, made it clear that his company plans to play a key role in driving these two important trends.
Formerly a senior cloud executive for both HP and IBM, and more recently a partner at Khosla Ventures, Singh’s vision represents a new direction for the company, with product details to emerge later this year in a series of announcements.
Orchestrating the future
A key tenet of Singh’s vision is orchestration. Microservices need to be orchestrated for application coordination, portability, and management, with partial solutions such as the Mesos and Kubernetes open source projects already available. Singh says Cisco will build on those existing projects but that ACI (Application Centric Infrastructure), Cisco’s sophisticated twist on SDN, will play an essential orchestration role. Cisco UCS (Unified Computing System), the popular converged server first introduced in 2009, will also be part of the solution:
Our future vision, and it’s coming pretty soon, is a very tight coupling of UCS and ACI with an orchestration layer, largely leveraging the way ACI has been architected. I would argue it’s almost the same thing. That is our core orchestration knitting. We think of it as a stack. You have compute, networking, and storage bundled together.
The way we frame it is: How do you straddle all of the requirements for everything from cluster management and what these Web-scale technologies like Mesos and Kubernetes and Docker Swarm are focused on -- which is very specific, well-defined cluster infrastructure? How do you manage that all the way down to very pointed orchestration, lifecycle-level orchestration in terms of config management tools like Chef, Puppet, Salt, all of that? We have a lifecycle embedded in this and we’re playing with it. We’re getting some pretty meaningful traction.
Although breaking monolithic applications into microservices running in containers has significant advantages, networking swarms of containers at scale using today’s tools gives operations a migraine. Singh cites an open source project launched last year as an example of Cisco’s work in this area: The Contiv container network plug-in, which enables admins to implement infrastructure and security policies for microservices deployments. In addition, says Singh, Cisco has actually built its own lightweight, high-level container PaaS called Mantl. Singh outlines how all this will come together:
The broad vision would be a very tightly coupled next-generation stack; compute, storage, networking, ACI-centric networking and orchestration with some of the things we talked about around Contiv or Mantl totally baked into that, certified by Cisco, pushed through our direct team, pushed through our partners as the next-generation data center buildout. We think we’re going to be very well positioned for that.
Clearly, Cisco is setting its sights higher than merely growing its server business or consolidating its dominant position in networking. To my knowledge, this is the first time Cisco has publicly discussed such a forward-looking, all-encompassing view of the data center.
That vision extends to analytics. For years, companies have used solutions such as Splunk or the ELK stack to pool log files and analyze them to get a deeper view of infrastructure, mainly for the purposes of optimization and to identify patterns that indicate security threats. More recently, machine learning has been applied to parse all that log data and surface significant trends and events.
Cisco has already taken a step in this direction with its Connected Analytics software portfolio, which targets not only data center infrastructure but also streaming data from the Internet of things. Singh wants to take this to the next level:
We’re doing a tremendous amount in deep learning, machine learning, and analytics. We’re going to flood all of Cisco’s portfolio with analytics capabilities, L2, L3, and above. I think there will be a lot of meaningful things we can do but I’m very excited about the data center piece of it. That’s what I’ve been working on the last few months.
Singh notes that MapReduce/Hadoop has not taken off like the industry thought it would, but that with the advent of streaming analytics it’s a whole new ball game, and Cisco intends to play:
We see a world where you have real-time streaming analytics at petabyte scale, sub-second analytics. This is a world that has something like, from a technology viewpoint, a Kafka-like pipe streaming in hundreds of billions of events, raw JSON into a streaming engine like Spark or Samza into a time-series database like Druid which you could then hold in Cassandra and persist it. Then you could throw analytics tools on top of it and actually interact with that data flow in real time and start building next-gen applications running on a container stack ... and redefine the way apps are done -- much more real-time and interactive.
If you think about that and you think about our role in networking and in data center and some of the things we’re doing as it relates to security, for us we’re going to take our analytics platform ... and basically streamline it across all of the business at Cisco.
As for machine learning, Singh says that Cisco plans to invest heavily, including hiring “a bunch of PhDs” to start adding machine learning and analytics capabilities to almost all of Cisco’s products. “My thesis has been that machine learning-based analytics is the secret sauce with which every future, next-generation software offering is going to be built,” he says. “We think that will play to our core networking data center offerings,” as well as to Cisco’s security and the Internet-of-things efforts.
According to Singh, this vision extends to cognitive computing in a collaborative workspace environment:
You walk in with your phone, your location, your avatar identity biometric and your payment is all there. It should know. I think the future of work gets completely redefined with contextual computing. The amount of processing power from a data center standpoint, the amount of ability to automate pattern-matching and anomaly detection, how to solve for false positives, when you tie all that together having analytics capability in everything we do is going to be super important.
So when will such solutions arrive from Cisco? Singh says to “stay tuned,” because the company will showcase some of these capabilities in a matter of months.
Building the next generation
Altogether, this seems like a lot of technology development for one company to bite off, even one the size of Cisco. At a strategic level, Singh and Cisco have clearly identified where enterprise computing is headed. Execution may be difficult, in part because the prevailing view of the next-generation data center assumes a commodity hardware infrastructure -- not exactly what Cisco is known for.
On the other hand, what other industry giant can lead enterprises to the promised land? No one thought Cisco could dominate enterprise networking as long as it has. Its highly scalable data center network fabric, ACI, which Singh says will be crucial to the company’s orchestration efforts, recently won an InfoWorld Technology of the Year Award. Plus, thanks to UCS, the company is tied for fourth place in the server market (according to the latest IDC numbers) and seems likely to climb higher.
That strong position could be a foundation for Cisco to lead the way in enabling a future of gazillions of orchestrated microservice containers in production and machine learning-infused analytics that pervade every corner of the data center. At the least, Singh makes a persuasive argument that Cisco is looking forward, not backward, and has a bead on the most important enterprise trends. We look forward to seeing the actual solutions that emerge in the coming months.