Why would enterprises peer?
Louis Winthorpe III
Enterprises have traditionally kept focused on their core competencies and outsourced non-strategic services to the brand leaders. Need telephone services? Contact AT&T. Internet? Call Verizon. Rental car? Hertz. They choose the market leader, the most recognized names in the Fortune 500 and stick with them. Let the best of breed handle things is their philosophy.
However, today, with the emergence of increasingly high-profile and high-value cloud services, these market leaders may not be sufficient for what cloud service customers require.
Peering for Enterprise and Cloud
As Andreas Sturm so eloquently puts it,
“Important Traffic Is Peered”
Peering is an essential building block for enterprises for at least two reasons.
With the commodity Internet, your packets are aggregated with every one else’s packets. All packets along a particular path are metaphorically on the same fate bus and share the same fate along their path to their destinations.
Most of the time the commodity Internet works. Until it doesn’t. There may be a denial-of-service attack that congests one of the many intermediary networks (as a rule-of-thumb, there are on average 4.5 networks involved in any end-to-end flows). There may be congestion along a link or within an intermediary network due to a spot event (major software update, webcast of a popular TV show, etc.) that adversely affects all traffic along a particular path. There may be a network or link compromised by hackers. As a result of this shared fate bus architecture of the commodity Internet, traffic along some paths will suffer from packet loss, queueing delay latency and jitter, or various forms of compromise due to hackers infiltrating one of the networks along the path. Those of us that have been around for a while know that, given enough time, these unpredictable and unreported anomalies will rear their ugly head and break things.
To depend on the commodity Internet for mission-critical 100% uptime requirement applications is simply negligent.
The Good, The Marginal, and the Ugly
Because of these inherent vulnerabilities, I tell my clients to perform their Peering vs. Transit performance comparisons over several months - that is enough time for them to witness the good, the marginal, and the ugly of the commodity Internet.
After this comparison, the performance of both Internet Peering and Internet Transit can be compared, and the disparity between what is needed and what is provided is revealed.
The enterprises that depend on cloud storage services for example expect that their cloud storage services will always be available to their employees, will always work, and provide consistent and reliable performance for their staff. What they witness instead is a service that is good enough most of the the time, but fails occasionally due to issues such as those mentioned above. Once they acknowledge this disparity, enterprises seek reparations via direct interconnection with their cloud solutions provider. For mission-critical applications like storage, peering this important traffic directly, bypassing the commodity Internet and these vulnerabilities, is a way to maximize reliability.
As discussed in an earlier blog, Internet Peering also improves security, another hot button topic for the enterprise.
Traffic between peers is architecturally and automatically segregated from the commodity Internet, and therefore is immune from the side effects of denial of service attacks across the shared-fate commodity Internet. Fewer network hops and links are involved, minimizing the amount of infrastructure that can be compromised. And if there is a problem, troubleshooting involves fewer parties, and there peering usually involves the exchange of contact information and escalation details. Your upstream may or may not see your trouble ticket as critically important to them as it is to you, but when you directly connect via peering with your cloud provider, the two interested parties can tackle the problem together from both sides.
Peering will lead the enterprise to high-reliability access for mission critical applications with a bonus of enhanced security beyond what they can get from the commodity Internet.
The first implication of this trend is that cloud services companies will embrace direct connect methods for their premium customers or perhaps for those that complain enough. During times of troubleshooting, the choice will be to troubleshoot intermittent issues across many autonomous ISPs, or to simply bypass the Internet. Most network savvy cloud firms and almost all enterprises will demand the latter.
The ubiquitous nature of the Internet will continue to enable access to cloud services from anywhere, but for important customers, for premier trading partners, and for important traffic, mission-critical cloud services companies will provide direct peering options.
Secondly, there will be a need for traffic management services to identify these dominant enterprise applications, their performance profiles (peering vs. transit), their dominant traffic flows and their destinations of interest. When network glitches are identified, software will provide or advise on automatic remediation such as requesting and implementing direct connect peering solutions from key networks and solutions providers.
Finally, these enterprises will need to be introduced and trained on peering. They may buy a book, attend a class or a peering workshop, they may hire people or outsource to a network SWAT team. In any case, they will need to be network savvy enough to go beyond multi-homing and enter into the peering ecosystem.
Peering has never been more important.