Simple, scalable, predictable and efficient: leaf-spine architecture is the modern standard for medium-to-large data centers. This streamlined topology is replacing the previously standard three-tier design, improving upon the limitations of Spanning Tree Protocol (STP) and the Access-Aggregation-Core model.
The media industry faces unique challenges when it comes to building data centers; today’s audience expects content at their fingertips, without interruption, and on a wide range of devices. Virtualization has forever changed the industry and requires increasing amounts of bandwidth.
Most businesses expect to see the need for storage capacity to increase by at least 25% and some even predict they will need to double or triple storage capacity. In this application-driven media marketplace, embracing new technology is among the clearest and easiest ways an organization can be more competitive.
What Is Leaf-Spine Architecture
Essentially, leaf-spine is a full mesh network topology that is constructed of two components: leaf switches that connect to servers and storage, and spine switches that connect the leaf switches. Leaf switches provide connectivity to endpoints such as routers, switches, firewalls, and other physical or virtual end-users. The role of the high port-density spine is to provide traffic-forwarding operations for the leaf switches.
The spine switches do not connect to each other and in most cases the leaf switches do not interconnect either. The access layer of a leaf-spine topology is formed by each leaf switch directly connecting with each spine switch. Instead of several interconnection points, the traffic payload must only travel from the ingress leaf switch to a spine switch and back out to the egress leaf switch. This direct connection cuts down on traffic bottlenecks and latency, which is the measurement of time between an instruction being made and commencement of the data transfer.
Why Transition to Leaf-Spine Architecture
The rise of virtualization and convergence in cloud environments means network topologies must be efficient, adaptable, and agile. Today many media organizations face rapidly increasing traffic and storage needs, but budgets remain static. In the past, bulk-purchasing more capacity was a practical solution to these increasing demands but is no longer a sustainable option. Hyper-converged solutions combine computation, networking, and storage into a single system, offering data centers increased speed and predictability. In order to maximize these data center systems, transitioning from the three-tier Access-Aggregation-Core model to modern leaf-spine topology is necessary.
With the previous network architecture model, most data center services and functions ran on physical hardware that moved “north-south” traffic. In this model, hosts communicate with other network segments via access switches; traffic flows down and back up the network. STP blocks redundant links to prevent catastrophic loops, but also results in access switches only able to use one link at a time. When a host on one access switch needs to communicate with a host on another access level switch (“east-west” traffic) probable congestion points are created. This presents a dilemma for data center technicians, as east-west data center traffic has increased dramatically over the past decade.
Straightforward Scalability
One major benefit of leaf-spine architecture is the ease of scalability. If a link becomes oversubscribed, leading to bottlenecks or long and unpredictable latency, expanding capacity is straightforward. The leaf-spine topology scales horizontally – meaning in order to increase capacity of the system, IT managers simply add more spine switches and connect them to every leaf switch. Another option for increasing capacity in a leaf-spine system is new interfaces between the existing leaf and spine switches. This flexibility means more reliable access to your content, even if it begins to go viral.
More uplink ports can also be added to the fabric of the leaf-spine model, thereby increasing port density. Higher port density at the spine tier creates more bandwidth and has the effect of reducing link oversubscription. Further, adding leaf switches to the system does not adversely affect the provisioned capacity of the existing infrastructure to handle east-west traffic. When new leaf switches are added they need to connect with each spine switch, which may require additional ports.
Hosts Are Equidistantly Spaced and Utilize Redundant Paths
With the rise of multi-tier applications and open source, data-intensive computing system platforms, traffic predictability is more important than ever. Because of STP, traditional three-tier topology does not always direct traffic in a uniform manner, making for unpredictable behavior and potential bottlenecks. Telemetry in the leaf-spine model is highly efficient. There are multiple redundant paths for every leaf switch; if one of the pathways fails several routes are readily available.
With the simplicity of the leaf-spine network hosts know that traffic will only make one stop before reaching the egress host. Every payload only needs to travel from the incoming leaf switch to the spine switch and back to the outgoing leaf switch. This leads to predictably low levels of latency. For media companies streaming to thousands of viewers across a variety of platforms, latency must be predictable and low.
Cost Effective and Energy Efficient
While convergence and virtualization present many opportunities for media companies, they have also resulted in inflating IT costs. With many organizations operating on a tight budget having little additional capital for the data center, any cost-saving measure is a worthy pursuit. The simplified setup that utilizes interconnectivity to reduce the amount of equipment needed for a data center makes the leaf-spine network topology a budget-friendly network topology.
In the three-tier design, switches are forced to communicate with two switches at the adjacent layer, requiring the larger chassis switches. Leaf-spine architecture solves this inefficiency with full-mesh topology that provides multiple paths through which traffic can be directed. Secondarily, with less physical equipment and an increased capacity for the east-west traffic of virtualization, your data center will likely see reductions in energy costs.
Conclusion
Virtualization and infrastructure convergence are trends that present unique challenges to media organizations but also great opportunity. The need to move traffic and storage both within and to-and-from a data center requires increasing amounts of bandwidth. The leaf-spine network topology presents a simplified approach to handling this growth. Evolving from a three-tier Access-Aggregation-Core network architecture to the modern leaf-switch topology is not a complicated, expensive or labor-intensive process.
With a need for fewer switches that cost less to obtain and maintain, working with a managed services company to move to leaf-spine topology is a smart choice for media companies. Once the change has been made, your organization will reap the benefits. Leaf-spine architecture provides an agile, reliable system that utilizes multiple redundant pathways to solve the problem of heavy intra-network traffic from virtualization and convergence.
Having a hard time convincing staff to embrace new technology? Read our step-by-step guide with tips for how to overcome that problem.
Comments