This represents a huge opportunity for those who can help enterprises consolidate their facilities. In their Strategic Roadmap for Data Center Infrastructure, Gartner notes that “by 2019, 80% of enterprises will have an IT strategy that includes multiple Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) providers.” This is up from only 10% in 2015, while “by the end of 2018, 10% of enterprises will close their on premise data centers entirely.”
How Are Enterprises Dealing with their DC Needs?
End customers--both enterprises and cloud service providers (CSP)--for IXP/CXP expect agility, in terms of the ability to deploy services rapidly. They also expect stringent service level agreements (SLA) and operational efficiency.
7 in 10 respondents believe they will always have a blend of traditional IT and cloud.
9 in 10 “front runners” (those who have implemented a hybrid cloud) say hybrid cloud gives them greater ROI than either an all-traditional or all-cloud environment
The report also notes that “improving productivity is the number one goal as companies hope to offload some of their IT resources and management complexity to the cloud.”
But gaining in top-line revenue is also a benefit, as enterprises aim to capture new markets by using hybrid cloud to pivot to new initiatives such as Internet of Things (IoT). All of this will deepen the importance of visibility, analytics, and reporting on IXP/CXP networks going forward. The network infrastructure decisions that IXP/CXP make today, and the resulting architectures, must ensure that the business can meet emerging enterprise customer needs.
And What About Cloud Service Providers?
Accordingly, cloud service providers are adopting a carrier-neutral cloud exchange interconnect strategy to allow enterprises and other service providers to be able to reach their cloud through high performance services delivered by IXPs and CXPs.
For example, one of the largest cloud providers connects with almost all major IXPs and CXPs around the world and has stringent pricing agreements with them. To be able to meet CSP requirements without sacrificing margins, CXPs/IXPs need to have high bandwidth at the lowest TCO possible.
This means that they need low price/port, visibility and automation to reduce CapEx and OpEx. These services need to be able to meet SLAs at all times.
How Enterprises Evaluate Colocation Facilities
Currently, the largest need of an enterprises for high performance connections is to peers or to the cloud. This should also include the ability to dynamically change the number of ports and to optionally receive bandwidth on demand.
Once basic connectivity and cloud exchange is achieved, growing enterprises soon find the need for more services and application workloads (e.g., big data analytics, content, storage, Citrix, email, IP telephony, virtual desktop, and disaster recovery). Some will need higher bandwidth, perhaps multiple connections at 100 Gbps.
Here is a summary of applications that are moving to the cloud:
Development & Testing
Project Mgmt, Expense Reporting, Expense Mgmt
Cloud based Anti-spam and Anti-virus
IXP/CXPs can partially fulfill the needs of these applications with a very low cost/port. To further reduce the cost, and manage a high concentration of users and application types, VLAN and VXLAN virtualization at a granular level (per application or service type) is also essential.
Furthermore, each of these application workloads require different service levels. For instance, some may require strict priority, and others may need individualized markings and weighted priorities. Some may need “bursts” of bandwidth or other resources, and here it’s important to remember that not every application workload is going to peak at the same time.
These Class of Service (CoS) levels need to be set for each type of workload. For instance, video, unified communications, and back office applications each require different classifications and attributes.
With many tenants and applications per tenant, the support of these categories needs to be automated. IXP/CXPs need to be able to create, delete, and upgrade services per site and per tenant. This avoids the need for huge upfront CapEx, as seen in the step functions in the following graph.
For instance, they may need twice as many servers in busy times as opposed to off-seasons, and will want corresponding business-aligned spend on power, as well as servers and storage.
The actual usage is represented by the white curve, and the ability to plan accordingly (especially when taking on the tasks in house) are shown in the orange lines. Grab too much capacity and it’s a waste--too little and customers are dissatisfied.
It is very difficult to match needs exactly, and that’s where outsourcing to a business that specializes in adjusting services to business needs can help. As much as possible, enterprises need to let someone else worry about routing, latency, security, load balancing and all the rest of their IT functions that, while essential, are peripheral to their core business.
To fulfill these requirements, Enterprises are increasingly adopting CXPs and IXPs to meet their hybrid cloud and peering needs. CXP’s and IXP's need to be stable, reliable, and accessible by any network that meets participation requirements. This means always-on stability and the ability to coexist with standard, open and unrestricted interconnection policies.
The goal, after all, is to interconnect as many enterprise businesses and CSPs as possible, with continuous addition of virtualization, visibility, analytics, automation and agility, with required quality per tenant, service or application workload type.
It’s clear that the hybrid IT cloud strategy is here to stay. Gartner underscores this strategy in their Infrastructure Agility Primer for 2017, noting that enterprises who include public cloud service providers as well as (where higher security and control is needed) private clouds for their own mission critical workloads, will gain the highest benefits.