This excerpt from ZK Research’s white paper Compute Transitions Drive the Need for the New IP Network is sponsored by Brocade. It reflects the views of the sponsor, not ReadWrite’s editors.
The architecture and infrastructure used to build legacy networks have not changed in almost three decades. There have been great leaps in the speed of network connections and the density of ports, and link latency continues to fall.
However, the network is fundamentally the same as it was 30 years ago. Many limitations need to be overcome if organizations are to make the shift to a New IP network.
These limitations include:
Rigid architecture and topology
The architecture used to build traditional networks was designed for an era when IT resources such as compute and application infrastructure were fairly static. Therefore, the network, too, could be rigid in nature. When an event occurred that required the network to change, such as adding or deleting a new application, long lead times were often required to make even simple changes.
In today’s era of increased agility requirements, this is not acceptable because the IT infrastructure can only be as agile as its least agile component—which is often the network. Too often, the rigidity of legacy network infrastructure is a huge impediment to organizations becoming agile businesses.
Siloing of the network inside IT
The network has always been treated as its own silo within the IT department. Network changes can certainly impact the performance of applications and compute infrastructure, but historically the network was managed independent of the technology that sits “up the stack” from it.
Legacy networks built on closed, proprietary protocols
Market-leading vendors have used closed, proprietary technologies as a way to deliver features faster instead of waiting for industry standards to be developed. Often, this leads to “vendor lock-in” and inhibits customers from implementing a best-of-breed technology.
Despite these limitations, closed, proprietary networks were sufficient when the network existed in its own silo. The cloud has driven the need for greater integration among applications as well as compute and network infrastructure. Closed and proprietary technologies can hinder cloud ecosystems and limit innovation.
Inefficient use of network resources
Traditional, multi-tier networks use Spanning Tree Protocol (STP) to protect against routing loops and broadcast storms. When multiple paths exist between two points, STP will enable the fastest path, disable the alternative route, and only make it active when the primary connection fails.
STP was a great leap forward for networking a few decades ago. However, as networks have grown, the use of STP has caused organizations to overbuild networks. Today, as a result of STP, up to half of network ports are inactive because they are passive links.
Consequently, service providers and enterprises have had to overbuild networks significantly, resulting in a less-than-optimal average network utilization of approximately 30%, according to ZK Research.
Hardware-centric
With legacy networks, any kind of scaling for additional capacity or new services must be done through the addition of new hardware. This typically requires “forklift upgrades” of the existing technology over a long period of time, resulting in lower application and data availability.
In addition, the hardware-centric nature of legacy networks makes running a network very expensive, particularly for service providers and global enterprises that often need to add capacity or services quickly.
Manual configuration processes
Traditional network devices, such as switches and routers, are designed with integrated control and data planes. Therefore, the majority of configuration and management must be done on a box-by-box basis. This leads to lengthy change management periods and a high amount of human error.
The ZK Research 2014 Network Purchase Intention Study reveals that human error is the number- one cause of network downtime today (Exhibit 2). To enable rapid, accurate changes, automation has become a top priority, with IT leaders and business executives looking to automate changes to their compute and application infrastructure based on business policies. The device-centric nature of managing network devices makes it difficult, if not impossible, to bring automation to the network.
Lack of programmability
Because converged infrastructure has become more commonplace, it’s important that the network is able to interface with a wide range of applications and infrastructures to optimize performance. For example, when a video call is being initiated, the application needs to direct the network to reserve a certain amount of bandwidth for the voice and images.
This is challenging today because legacy infrastructure lacks programmable interfaces, which prohibits application developers from programming the network to automatically accommodate these types of random traffic spikes.
Designed for “yesterday’s” applications
Traditional networks were optimized for old-school applications such as email, voice, CRM and other static applications. Today’s applications are enabled by the cloud and include mobile applications, big data analytics and social business. These new-age applications have significantly different network requirements from their old-school counterparts, though there is still a need to simultaneously support both. The shift to cloud computing is ushering in a new era in networking. A New IP network that is optimized for the cloud era is now required.
Historically, organizations made network infrastructure decisions based on brand, incumbency and market share versus technical superiority because “good enough” was adequate to support legacy compute models.
But as the cloud era gains momentum and hybrid clouds become the dominant compute model, the network will continue to increase in value. Infrastructure that was good enough for Internet computing will not be sustainable for cloud computing. A New IP network is now a business mandate.
Photo by dominik99
This excerpt from ZK Research’s white paper Compute Transitions Drive the Need for the New IP Network is sponsored by Brocade. It reflects the views of the sponsor, not ReadWrite’s editors.
