Thinking of Multi-Cloud Peering Between Cloud Service Providers?
NOTE: This blog exists as part of a six part series.
Organizations are moving towards hybrid multi-cloud environments to take advantage of best-in-breed cloud services. However, this transitional path is not devoid of its challenges, be it design or operational.?If you are planning your multi-cloud journey, you better be ready to roll up your sleeves, as developing a peering relationship between service providers, and securely connecting those environments together is equivalent to a major home DIY remodel project, where you are the designer, the contracting agent, and the project manager coordinating various sub-contractors and utility providers (while at the same time keeping your customer happy).?
Before we discuss these challenges, let us take a minute and revisit the business model behind the Internet services that we so take for granted. Given its simplicity, it is easy to forget that the Internet is not one Internet Service Provider (ISP); it is hundreds/thousands of ISPs across the globe, all working together.
ISP Peering Simplified Internet Connectivity for Customers
When a customer purchases Internet service, they are buying into a global consortium. With that global consortium, customers have access to any URL, any IP address, anywhere in the world. The foundation of this global partnership is the connectivity agreements that each ISP has with other ISPs that define their interworking relationship. These working agreements are known as peering agreements. Peering agreements are two-way agreements established between two connecting ISPs and is extremely important in improving the efficiency of operations.??A single ISP can have dozens of peering agreements, where each agreement defines the connectivity and working relationship between each ISP.
To enforce the traffic rules established in the agreement, at the operational level, ISPs utilize the Border Gateway Protocol (BGP). BGP is a very powerful and flexible protocol used to discover service adjacencies and manage Internet traffic passing between one provider and another. The strength of the protocol is its inclusion of various network discovery metrics, called attributes that providers exchange with each other. Each ISP passes these attributes on to the next ISP, which helps everyone establish Internet service adjacencies.?These adjacencies are incredibly important in providing efficient routing connectivity, not just between the peering ISPs, but those that are two, four, or seven hops away. These attributes (LocalPref, AS Path Length, Origin type, MED, etc.) enforce business, operational, and networking rulesets established in the contractual agreement, across the boundary.
In the early days of the Internet, most ISPs physically connected at commonly known facilities known as Network Access Points (NAP). These meet points (Who remembers MAE-East and MAE-West?) made it easy for ISPs to come together and cost-effectively connect with each other. The primary advantage of these common independent locations was cost, latency, redundancy, and bandwidth scalability.
The reason why it is important to review this history is that it highlights the brilliant simplicity of the Internet?as seen by customers?purchasing access; a simplicity that is often taken for granted. All of the complexity of hundreds of companies, thousands of peering agreements, their ongoing operations, and the governance necessary to ensure all parties comply, is hidden from customers. Customers simply purchase Internet bandwidth and connect.
In contrast to the ISP model, today’s cloud providers are operating on their own, leaving the responsibility of creating a seamless business and operational environment to each enterprise customer.
领英推荐
Hybrid Multi-Cloud Peering Challenges
Enterprise customers venturing into the hybrid multi-cloud environment soon realize it comes with challenges. Some of these operational challenges are:
As enterprises, continue to explore ways to optimize their cloud-based operations by choosing to work with best-in-breed cloud service providers, the operational challenges continue to baffle them.
The Emergence of Hybrid Multi-Cloud Networking
The first step in utilizing any cloud provider is connecting to them. It is important to realize that when an enterprise shifts a workload to the cloud, it often requires significant bandwidth at significant costs to connect multiple Virtual Private Clouds (VPCs) and the computing resources of a private data center together. It is also important to realize that the current cloud providers do not provide connectivity from their environment to remote cloud environments, thus making it the enterprise’s responsibility.
Multi cloud networking (MCN), a relatively new and rapidly developing market has emerged. This market was developed because traditional carriers did not necessarily have a cost-effective means (usage-based pricing, any-to-any connectivity) to deliver high-speed connectivity to the cloud. In its most?simple?form, multi cloud networking is the implementation and management of the basic connectivity required to interconnect and secure an enterprise customer’s private data center with multiple public cloud providers to distribute IT workloads across these environments. This network connectivity can be complex and challenging considering the inclusion of regional/edge cloud providers, public colocation space, local access points of presence, and planning for disaster recovery scenarios.
The Criticality of a Service Adjacency Matrix
Circling back to the history lesson regarding ISP connectivity, just like ISP peering agreements are more than just a physical connection, networking across on-premises data centers and multi clouds is a lot more complicated than basic connectivity, and the fact that each cloud provider operates on their own makes this even more complex. As business IT workloads are distributed and more containerized applications are deployed in on-premises data centers and multi cloud environments, it will be extremely important to understand the service adjacencies across the hybrid multi-cloud environment. Which application instances are running where, what is the ability of the network to support those in terms of various network metrics (BW, latency, jitter, delay, etc), and what the various QoS metrics and cost metrics are that need to be considered on individual flows to enable the network services to reroute and converge rapidly in the event of workload re-distribution events.
Needless to say, working in a multi-cloud environment, begins with the design process.??That process needs to accommodate several complex issues that each enterprise moving to a multi-cloud needs to address on its own.??But if history has taught us anything, it is that the pace of change is not going to slow, and the design process needs to yield a network that is as agile as the applications that it supports.