For the past 15 years enterprises experimented with SaaS applications, took advantage of plentiful cheap compute and storage and migrated desktop workloads to the cloud, yet still have not managed to achieve digital transformation. Panellists at the NetEvents session on multi-cloud (see box) agreed that we are finally on the brink of a generation change in IT driven by business necessity and enabled by new cloud-native solutions to decades-old problems. The revolution will finally see the traditional data centre model retired as monolithic software is replaced with highly distributed, high-performance cloud-native applications. But the major challenge in the battle for the future remains the complexity of building networks fit for the cloud era. According to our expert panel, true digital transformation cannot happen without infrastructure capable of supporting distributed development and IT operations, and able to provide enterprises with full visibility and control of their environments.
Making Multi-cloud and Hybrid Cloud Work on 14 April 2021 brought together an expert panel comprising two of the biggest names in networking, Cisco and Verizon and two of the industry’s hottest start-ups, Alkira and Aviatrix.
Vijoy Pandey, Vice President of Engineering and Chief Technology Officer, Cisco
Vijoy is an engineer and innovator with 80+ patents issued in distributed systems and networking, with over 20 publications in networking, and numerous keynotes, talks and panel appearances.
Oliver Cantor, Associate Director of Verizon Global Products
Oliver has spent more than 25 years with two of the biggest companies in global telecommunications, Ericsson and now Verizon. His background is in engineering and product development, while his professional interests include trends in distributed computing, organisational culture, and the influence of economic and geographic trends on industry.
Amir Khan, President, CEO and co-founder, Alkira
Amir and his brother Atif pioneered the $5bn global market for SD-WAN with their last start-up, Viptela, which they sold to Cisco in 2017. The Khan brothers have numerous technology patents. Their latest venture Alkira was set up to solve the next big problem in IT – networking in the cloud era. The company attracted $76 million in venture funding from, among others, Kleiner Perkins, Google and Koch Disruptive Technologies.
Steve Mullaney, President and CEO, Aviatrix
Steve brings 30 years of marketing, product management, and engineering experience in network infrastructure and security to Aviatrix. Most recently, Steve was the CEO of Nicira Inc., a network virtualization and software-defined networking company, from 2009 until it was acquired in 2012 by VMware for $1.26 billion. He has also held senior roles at Palo Alto Networks, Cisco and other industry leading firms.
Brad Casemore, VP Research, Datacenter and Multicloud Networking, IDC (chair)
Brad is one of the foremost authorities on data center, networking and cloud technologies in the analyst community. Brad provides ongoing research for IDC’s Continuous Information Service (CIS), market forecasts, custom consulting and go-to-market services. He is a regular speaker at major industry event and source of media comment and opinion.
Brad Casemore, VP Research, Datacenter and Multicloud Networking, IDC defines multi-cloud networking as the infrastructure that supports distributed enterprise workloads. “We all know that we’ve entered, as a result of digital transformation, a cloud era where more and more enterprises are harnessing the cloud for agility, for flexibility, for perceived cost savings. And increasingly, as a result of the pandemic, they’re also leveraging cloud for business resilience and continuity,” he said.
But complexity is a major stumbling block for enterprises, whether they’re keen to move to multi-cloud or simply resigned to the inevitability of it.
As Casemore said: “You’re dealing with discrete clouds, you’re dealing with disparate APIs associated with each cloud, you’re dealing with different service insertion models, you’re dealing with different network constructs. Look at some of the infrastructure issues and particularly the network issues that relate to multi-cloud, and enterprises find them daunting, they find them intimidating, and it has become somewhat of an inhibitor to them fully leveraging multi-cloud for business benefits and outcomes.”
With organisations running more and more workloads in the cloud, refactoring existing applications and building new distributed cloud-native apps, the traditional data centre was also becoming distributed, Casemore says, which is why multi-cloud is no longer a choice for many enterprises but a necessity.
Oliver Cantor, Associate Director of Verizon Global Products observed that the shape of networks is being changed not by technology per se but by new patterns of use – a move from what he calls static to dynamic digitisation. “In the past, you could say that networks connected islands of compute together, LANs and telephone exchanges and other things. What are they connecting together now? Well workloads, and workloads are moving, data is moving, users are moving, everything’s on the move.”
Vijoy Pandey, VP Engineering and CTO at Cisco agreed: “In the end there are applications and there are users and nothing else really matters,” he said. The impact of distributed applications on the role of the network is profound, however. “These big behemoths are being split apart, their guts are being taken out and they’re being strewn across the Internet.”
Pandey said that software was being increasingly atomised, with collections of APIs and microservices woven together to form enterprise applications spanning on-prem systems and clouds. “The network is becoming the runtime for all these apps” and no longer just an underlying transport but “full-stack connectivity, security and observability from the application layer to the virtualisation bare metal”, Pandey said.
Amir Khan, President, CEO and co-founder of Alkira, said he started his business because he could see that while the cloud was providing enormous resources and enterprises were keen to push more applications and workloads out of the data centre, “networking was lagging, it had not adapted to the cloud”.
“When we started looking deeper into how each of the clouds was providing networking capabilities, they were all different. So we had to build our own virtual infrastructure in the cloud because otherwise all the underlying nuances and details were being exposed to customers and multi-cloud environments were not being adopted at the pace that was needed.”
Steve Mullaney, President and CEO of Aviatrix, agreed that even recent networking constructs were under threat from the mass migration of enterprises to the cloud. “SD-WAN is dead. Why? Because it’s not about going back to the data centre, it’s about going to the centres of data that are in the cloud. I can’t take the old world and jam it into the cloud, I need a cloud-native solution. However, I’m like the government. I need the visibility, the security controls that I used to have on-prem,” Mullaney said.
Mullaney likened the marketing promises of the big cloud providers to the brochures for holiday resorts that appeared to provide luxury and choice to guests until they actually arrive at their destination. Then they’re confronted by a fixed menu with no a la carte options. “They get to the cloud, they’ve been sold a bill of goods and they realise that there is so much complexity, so much that is manual, even within one cloud.”
Enterprises are rushing headlong into even greater complexity by choosing not just one but several clouds, though as Mullaney pointed out, the multi-cloud trend is more ad hoc than strategic. Business teams were picking different clouds to access a particular application or feature set, creating a multi-cloud by default “nightmare” that was left to the infrastructure team to sort out.
Complex or not, enterprises are impatient and IT providers can’t afford to keep them waiting. Verizon’s Oliver Cantor, who likened cloud networking to a “giant utility” the IT industry is coming together to build, said: “We have shifted from a supply-side market to a demand-side market. We’re being dragged forward – and not just by the enterprise customers but by their customers.”
Alkira’s Khan said that as enterprises became accustomed to shorter application development and deployment times in the cloud, they were no longer prepared to put up with plodding network development. “As one of our customers says, we need to move at the speed of the business, not the speed of the network and that’s what we’re changing. Now you can deploy global networks within minutes or hours rather than the months or years it used to take.”
Cisco’s Pandey agreed that customers are not only tired of waiting but bored of hearing about the technical detail. They want declarative, intent-based models, he argues, “systems that take you from A to B not systems that tell you how to get from A to B”.
Cantor also detected a growing frustration among customers: “Enterprises are just shaking their heads at all this complexity and are looking for an as-a-service model. At some point, they fell in love with the utility, the usability and the pay-as-you-go ability of the cloud to enable business. The movement to cloud-native from bare metal and monolithic is going to happen, so how do we make sure our customers’ journeys are simple and straightforward?”
The panellists agreed that it means going back to layer by layer design principles with clean APIs up and down the protocol stack from application to the lowest levels of connectivity. Without such design rigour, programming or operator errors in a complex highly distributed system could have profound consequences.
Cisco’s Pandey said that while it appeared “horribly scary” in terms of connectivity to take monolithic apps and make them cloud-native, the upside is that the resulting discrete components of the application can be swapped out or taken down with fewer consequences to the rest of the system and ultimately to customers. But, he warned, “you need to have the tools and capabilities to monitor it – the full-stack observability piece. You need to have discoverability and you need to have security at the API layer all the way down so that you can manage things properly”.
His comments were echoed by Alkira’s Khan, who pointed out that the problems of a distributed architecture are particularly acute for enterprises trying to apply a security posture in a multi-cloud environment. “There’s a hodgepodge of instances that you need to stitch together on top of which enterprises are creating static policies. And you cannot do even simple routing or segmentation across these environments today.”
The issues of visibility and management control are among the hardest to solve in multi-cloud environments. Aviatrix’s Mullaney argued that these requirements would put enterprises off relying on SaaS offerings, which he characterised as “black boxes”. It was up to the industry to provide the best of both worlds – maximum agility and total control, he said.
“I have to be able to double, triple, quadruple click down if I need to, to understand what’s going on. This is a problem with the as-a-service and the cloud providers. You say to them ‘What will happen if something goes wrong?’ and they say ‘Nothing will go wrong’ and you have this stand-off. If I’m an enterprise, I have to have the visibility and control I used to have in my on-prem infrastructure, but I also want the velocity and agility that go the cloud,” Mullaney added.
Alkira’s Khan agreed that enterprises wouldn’t settle for agility without also getting assurances of control over day-to-day operations, security, regulatory compliance and governance.
The other important thing they need to be able to control is cost. This is an increasing problem as the cloud made it easier for anyone in the business to turn on IT resources.
Or as Khan put it: “An employee goes and spins up an extra-large instance in the cloud when they could have got away with a medium sized instance. That’s where the cost comes from. Also, at least from a networking perspective, the automation you apply needs to be elastic, so capacity can be scaled up and down according to your needs. That’s what Azure and AWS allow you do for compute, storage and database applications. And that’s also what we’re doing for the network.”
Aviatrix’s Mullaney agreed that to reduce complexity and impose control, enterprises would need a common set of services across all cloud platforms that provided a “single pane of glass for visibility across all these clouds as well as on-prem”.
Although the panel agreed on the fundamental importance of cloud networking infrastructure, there was less agreement about exactly how it should be delivered.
Cisco’s Pandey argued that too much emphasis on control to lead enterprises away from a strategic view of IT and back towards an operationally heavy approach. “Control is the harbinger of complexity. The design layer, the policy layer is the right place to provide intent. Anything beyond that drives complexity,” he said.
Verizon’s Cantor cautioned enterprises against taking the DIY freedoms of the cloud too far. Acknowledging the appeal of SaaS and service-based infrastructure, he argued that there was still a place for service providers like Verizon with deep engineering expertise. Whatever magic was performed at the top of the protocol stack, the performance, security and resilience of enterprise networks in the cloud era would still ultimately depend on being able to manage everything from the application at the top to the fibre and bare metal at the bottom of the stack, Cantor said.
Aviatrix’s Mullaney predicted that SaaS would become a “four-letter word” and could even be on the way out, claiming that his company’s customers did not want to buy their network infrastructure as a service. His view is that enterprises want to optimise their own infrastructure and bring in the orchestration skills of vendors such as Aviatrix to provide the management overlay.
Alkira’s Khan disowned the “black box” label for his company’s cloud network as-as-service (CNaaS) offering, which he said allows for deep visibility of the network, granular control of services and complete end to end management.
Challenging the approach of vendors like Aviatrix, Khan said that orchestration solutions would always be one step behind the hyperscale cloud providers in much the same way that the third-party network management solutions of the past were always playing catch-up with the big network vendors. “That’s why we took the approach of building our own infrastructure in the public cloud so that we didn’t have to rely on ten different cloud providers to bring out a feature before we could offer it,” he said.
All the panellists were agreed that in 2021 the IT world is entering a new phase driven not by technology but business imperatives.
Aviatrix’s Mullaney said: “For every organism, the number one thing is survival. That’s what is driving this. This is the business people saying we have an existential threat to the survival of our company. And it’s not from the other legacy vendors: banks aren’t worried about other banks, they’re worried about the neo banks, they’re worried about people that have digitally transformed and have the attribute that everybody’s talked about here, which is agility.”
IDC’s Casemore said that cloud was changing not just the nature of technology or the skills required by IT departments, but the way businesses work. “At IDC we see enterprises going through a transition where not only the infrastructure is cloud-like but the operating model is too, and it includes all the agility, flexibility, speed, and intelligence that you would expect from fully leveraging cloud technologies.”
By Guy Matthews, Editor, NetEvents
See forthcoming events from NetEvents
Find out more about the companies taking part: