We recently completed some research on the current and future state of multicloud networking. I’ve reached the conclusion that the momentum put in place will only continue into 2024.
Let’s explain a few reasons why.
For those of you not yet familiar with the MCN trend, it’s a new requirement among network and cloud architects to be able to securely connect multiple network domains – enterprise, datacenter, cloud, and communications provider – with a more integrated networking platform. The needs of these cloud networking platforms vary widely, but they range from application visibility to network security and policy.
What’s changed in the past year? Interest and activity around AI, AI, and AI has, of course, accelerated. But be careful how you discuss AI with your chief technology officer, or he/she is likely to roll their eyes. The truth in the trenches of IT and cloud infrastructure is that the return on investment remains unknown.
What AI Means for Infrastructure
It’s still early in the AI deployment cycle for enterprises, but it’s clear over time that the need for more AI means more data and more networking. Here are some of the things that engineering leaders are thinking about when they need to connect datacenters with AI:
· Seamless connectivity among multiple clouds and datacenters
· Accuracy and safety of data and content
· Data provenance and privacy issues
· Optimized of supporting AI across clouds (data, infrastructure)
In the case of many professionals charged with infrastructure, this has implications for networking. Networks are required to connect diverse computing stacks together. Namely, how do you assure the safety and privacy of data that is being transmitted across your network to fuel AI models? More importantly, with AI being an outgrowth of existing applications, how do you connect resources together and scale the increased demand for data?
The data that will be used for AI will be everywhere, including multiple clouds and private datacenters. This will pose a challenge for many organizations.
In our annual MCN and network-as-a-service survey of 131 IT pros and cloud engineers, we asked respondents if they thought the arrival of AI services would increase the need for MCN services. In the results, 82% said yes, while only 12% said no. A small minority of 5% said they didn’t know.
Toss AI on top of 5G services, data lakes, and IoT connectivity and what you get is a lot of thirsty apps and devices scarfing down data in massive quantities. But wait, there’s more: These apps and services are more distributed than ever, in a diverse ecosystem. The networks must connect a wide range of cloud providers, content delivery networks, service providers, and enterprises. These connections must increasingly be made on-demand, using automated connectivity technologies.
Networking professionals point out that multicloud networking will have to play a greater role in all of this, serving as a platform for transporting, securing, and managing data across multiple clouds and datacenters. As the chart from our survey below shows, the majority of organizations are using multiple public clouds.
Large Implications for Many Stakeholders
In our research, we determined that this generational shift in cloud architecture, driven by AI, will drive major shifts across the cloud and communications infrastructure.
It’s already clear that these implications are being understood broadly across the ecosystem. Here’s how different segments of the market are responding:
MCN and networking technology providers (startups): There is a significant group of startups being funded in the venture community to help solve broad and complex MCN and hybrid cloud challenges. Key vendors to watch include Alkira, Aviatrix, Arrcus, Graphiant, Itential, Prosimo, and Versa Networks. We covered these companies in detail in our recent report.
One trend we have noticed over the past year is that the incumbents have become more focused on datacenter and hyperscale solutions to help scale AI, and they seem less focused on multicloud solutions. That presents an opportunity for MCN innovation. Startups will remain prominent contributors to this trend.
Networking incumbents: Incumbent networking and cloud software providers will continue to add functionality to their existing platforms to enable MCN. In some cases, these functions will be built organically. Futuriom sees incumbents continuing down the path of M&A to buy innovation. Key vendors to watch for M&A include Arista, Cisco, HPE (and Juniper), and IBM.
Telecommunications service providers: Telcos see the opportunity for MCN and hybrid cloud services, and once again they are trying not to be relegated to “dumb pipe” designation. Their opportunity lies in taking their global networking resources and integrating with cloud services and next-generation MCN provisioning tools to make it easier for customers to leverage cloud networks. As cloud providers increase the connectivity options for their infrastructure, service providers such as AT&T, Deutsche Telekom, Telefónica, Verizon, and others will need to partner with MCN upstarts to add capabilities. Some partners to watch here include Megaport and Packet Fabric.
Cloud PaaS and IaaS providers: The largest cloud operators — including Amazon, Google, Microsoft, IBM, and Oracle – have built large cloud networks that can be used by enterprises to build MCNs. By establishing a global network of points of presence that can be accessed by APIs, PaaS and IaaS operators can provide MCN connection points closer to the enterprise customer. In some cases, the cloud providers aim to be MCN providers themselves – by providing edge services that can be used by enterprises. Some examples include Google Cloud Anthos, AWS Regions and Zones, and Microsoft Azure Virtual WAN. While these MCN functions are useful for customers building inter-cloud networks, many of the public cloud infrastructure services are proprietary in nature. Futuriom believes that most larger enterprises will find these services useful to connect their MCNs to the public cloud providers, but many enterprises will want more control over MCNs that they build themselves.
Content delivery networks, cloud NaaS, and middle-mile services: NaaS and CDN providers, like the cloud providers, have their own global networks that can be used by enterprises to build and connect MCNs. In contrast to some of the startup vendors that supply do-it-yourself tools for building MCNs, many of these providers supply a NaaS service that enables the enterprise to outsource global networking services, often positioned in the “middle mile” of the global network. Some of the providers of cloud-based NaaS, middle-mile, and CDN services include Alkira, Akamai, Aryaka Networks, Cato Networks, Cloudflare, Megaport, and PacketFabric.
Colocation providers: Datacenter and colocation providers such as Equinix and Digital Realty have private cloud infrastructure and networking connectivity services that can be used to build MCNs. Because datacenter PoPs are often located in the same colocation facilities of public cloud providers, these services can be used as a bridge to MCNs – for example, by connecting local private resources directly to cloud services, with cloud onramps. Futuriom sees the colocation providers as playing a key role in MCN development. For example, the Equinix Cloud Exchange Fabric (ECX Fabric) provides software-defined network interconnection across a global infrastructure and public cloud PoPs. Digital Realty’s SX Fabric enables organizations to securely manage cloud networks and applications among PoPs. The large colocation and datacenter providers are likely to continue to make moves to enable more extensive cloud networking integration for customers.
In 2024, we’ve seen the MCN ecosystem focus on feature upgrades, partnerships, and pursuit of the ever-present AI craze. We expect this activity to increase in 2025. In the next couple of weeks, I will post another article detailing some of this activity and what we expect to happen.
