Multiclouds are flattening from end to end. As we move into the coming decade, centralized and hub-and-spoke intercloud networking will decline in importance in enterprise architectures.
As data gravity pushes more cloud topologies toward the edge, Wikibon sees multiclouds evolving into radically distributed mesh architectures at the infrastructure, platform, and application levels. In fact, the more complex the enterprise multicloud, the more likely it is that it will be implemented as a service mesh that sprawls across disparate on-premises and public cloud environments.
Trends accelerating the flattening of multiclouds
As this trend picks up speed, multiclouds will flatten along several planes of cross-mesh distributed management:
- Resource management: In future mesh multiclouds, there will be more even distribution of compute, storage and memory resources across all tiers, clusters and nodes, with more workloads parallelized to execute across increasingly powerful edge devices. This will require sophisticated resource management controls such as load balancing and fine-grained routing, rate limiting, flow control, protocol translation, authentication and authorization and monitoring and logging.
- Workload management: As mesh architectures take hold, there will be more flexible movement, routing and control of workloads, with streaming, publish-and-subscribe and stateful continuous processing becoming the dominant approaches for handling real-time, low-latency, distributed workloads across the multicloud. This will require fine-grained control of microservices traffic behavior with rich routing rules, fault tolerance and fault injection, as well as automatic zone-aware load balancing and failover for diverse traffic types.
- Interface management: In the mesh multicloud, development abstractions will deliver programmatic access to all routing, policy, security and other control-plane functionality. This will require distributed catalogs for managing APIs, service definitions, machine-learning models and metadata to facilitate discovery, delivery and management of application interfaces.
- Orchestration management: In the decentralized multicloud, there will be more peer-to-peer orchestration of nodes within and across all tiers, all the way out to mobile, embedded, “internet of things” and other edge devices. This will require proxy servers that intermediate the network path between service nodes.
- State management: In the edge-oriented multicloud mesh, there will be management of shared application state as a shared context. This will require a distributed persistence plane — distinct from hypervisor, container, serverless and streaming application backplanes — that manages state, context and other metadata as a shared resource.
- Performance management: As command-and-control gives way to dynamic cross-mesh operations, there will be more software-defined, artificial intelligence-driven monitoring, orchestration, optimization and assurance of end-to-end application performance across the multicloud. This will require continuous monitoring of traffic and workloads, using this data to enforce policy decisions such as fine-grained access control and rate limits.
- Identity management: As the edges begin to dominate the multicloud, the need for distributed strong authentication — built on multifactor identity assertions — will grow. This will require end-to-end trust relationships, role-based access controls and confidentiality across all nodes, applications and microservices, perhaps leveraging blockchains for secure credentials management.
Recent industry landmarks in the multicloud’s flattening
Wikibon has seen ample validation of these trends in recent announcements and strategic directions from in Cisco Systems Inc., VMware Inc., Dell Technologies Inc., IBM Corp., Google LLC, Amazon Web Services Inc. and other cloud solution providers. Clear recent signs of the flattening of the multicloud include:
- Hyperconverged nodes for simplifying hardware deployments from cloud to edge: Wikibon expects that more of these multicloud network backplanes will run on hyperconverged infrastructure platforms. These will support flexible scale-out of the compute, memory, storage and bandwidth needed to supporting growing network traffic and more complex machine learning-driven, closed-loop optimization patterns. Hyperconvergence simplifies information technology resource provisioning by combining storage, computing and networking into unified scalable nodes. Hyperconverged platforms incorporate virtualized computing, networking and storage resources. Deployable anywhere in a multicloud, these systems enable multiple nodes to be clustered together to create pools of shared compute and storage resources, designed for convenient consumption. Paving the way for use of commodity hardware anywhere in the flattened multicloud, Cisco recently enhanced its HyperFlex family on hyperconverged infrastructure hardware solutions to enable enterprise deployment of a consistent hyperconverged architecture across on-premises, hybrid cloud and edge environments. The new devices come equipped with embedded connectors to the cloud-based Cisco Intersight, which enables AI-driven IT operations management, data resiliency and centralized lifecycle management capabilities. This capability enables HyperFlex Edge devices to be installed, configured, monitored and optimized in fully automated, zero-touch fashion on HyperFlex Edge clusters.
- Kubernetes for orchestrating containerized microservices all the way to the edge: Enterprises are evolving their multicloud network backplanes to incorporate more routing, policy, security and traffic management capabilities in containers that are orchestrated through Kubernetes. Software-defined wide-area networking on a Kubernetes backplane is a powerful way to do application-level routing to edge devices. It supplements the network-layer routing that has long been the mainstay of most networks. It also adds the ability for infrastructure components to introspect containerized content payloads to drive more congestion management, intrusion detection and security functions. Standard approaches for extending these network-layer functions to edge devices are essential. Wikibon applauds the Cloud Native Computing Foundation’s KubeEdge initiative and its collaboration with the Eclipse Foundation in the Kubernetes IoT Edge Working Group as the right steps in this direction. Wikibon expects that Kubernetes-orchestrated network routing and traffic policies will increasingly be deployed in infrastructure controllers. This will allow enterprises to deploy rapidly only those networking features required everywhere in their flattened multiclouds. This will have the benefit of reducing the complexity of networking routing and policy updates. It will also reduce risks by enabling more rapid and consistent updates to routing, policy and security rules throughout the multicloud. These Kubernetes-based controllers, to be provided by all leading network virtualization vendors, will support 24-by-7, lights-out policy-based programmable automation of multicloud network management operations. The controllers will translate enterprise private-cloud network and security policies into equivalent instructions governing the behavior of network components, including virtual switches and firewalls, that run in target public clouds. As this trend intensifies, vendors will embed this containerized network routing capability into a growing range of edge gateways, on-premises computing/storage racks and device-level container runtimes to manage distributed Kubernetes backplanes.
- Istio for managing traffic and routing across mesh interclouds: Increasingly, enterprises are bringing service mesh topologies into the core of their multicloud initiatives. This will enable them to build flexible bridges between containerized on-premises assets and a growing range of public and private cloud fabrics in their distributed computing environments. To address these growing requirements, the principal public cloud providers have built out impressive service mesh capabilities. Cloud providers are ramping up their support for managed services that simplify interconnection and management of thousands of virtual private over mesh, hub-and-spoke and other complex multicloud architectures. Adoption of cloud-native industry microservice-mesh management initiatives — most notably Istio — will enable enterprises to proactively monitor, control and optimize meshes through the proverbial single pane of glass. In this way, enterprises will be able to extend software-defined wide area networks over service meshes. AI-enhanced software-defined networking capabilities will include intent-based networking, application-aware firewalling, intrusion prevention, health monitoring, anti-malware and URL filtering across the meshes. In addition, blockchain and other hyperledger backbones will evolve to provide an immutable audit log for the network-, system- and application-level data that is used to train all of these AI-driven policy controllers throughout the multicloud.
Here’s a further discussion of recent industry announcements that push flattened meshes to the heart of the multicloud revolution.
