Having knocked down a number of the myths of why enterprises "need" MPLS over our last few columns, let's now address the most difficult – and in many ways "legitimate" – myths around MPLS. They can be worded different ways, but boil down to the claim that MPLS delivers reliable, predictable application performance for the enterprise, while using Internet connectivity does not. These myths can be worded in any number of ways, including:
- "There is no QoS on the Internet, so I need MPLS."
- "Packet loss and jitter on the Internet are too high, so I need MPLS."
- "To get reliable, predictable performance, I need MPLS."
- "I need MPLS for VoIP."
Per one of the themes of this column, this is one of those places where I believe WAN managers have been correct to be conservative up until now, but with the introduction of the Next Generation Enterprise WAN (NEW) architecture and the set of technologies incorporated in it, there is now a better way. Let's go through the reasoning.
First, that reliability and performance predictability for all applications is critically important to any enterprise WAN manager worth his or her salt. On this point, I wholeheartedly concur. And, in fact, public Internet connectivity, which has taken on a larger and larger role in almost all aspects of computing over the last 20 years, has barely made a dent as the primary means of enterprise site-to-site connectivity precisely because it cannot deliver performance predictability on its own.
I covered many of the reasons for this in an earlier column, the gist of which is that even though IPSec VPNs long ago addressed the issue of network-level data security, the network-of-networks which is the Internet won't on its own deliver the 4 nines reliability that enterprises rightly want and expect, partly from design, and in large part because there is no economic incentive for any Internet Service Provider to make the system as a whole work that way. So it is true that any Internet connection will face a lot more congestion, and thus deliver less performance predictability, than almost all MPLS connections.
Given this, enterprise IT buyers have been correctly risk-averse until now, sticking with at first Frame Relay and now MPLS for most of their intranet connectivity needs. With limited IT staffs at most remote locations, it makes sense to be risk averse when it comes to the WAN, as the yearly savings of going with a cheaper but less reliable WAN service can sometimes be eaten up in handling even a single incident. Further, given limited staffs, paying lower WAN costs and handling connectivity problems on their own when they arise – and arise fairly frequently – is not even an option for some businesses with a lot of locations.
Of course, for TCP-based applications, WAN Optimization and/or Network-as-a-Service today deliver superior, and usually more predictable, application performance than the typical enterprise MPLS deployment. WAN Optimization appliances do end-to-end application optimization, and Network-as-a-Service leverages colocation facilities to deliver much better, and more predictable, performance in the face of network packet loss. Neither of these technologies, however, can do much for real-time applications like VoIP, other than prioritizing the real-time traffic and minimizing the bandwidth consumed by the TCP traffic on the first and last mile.
With the introduction of the most recent NEW architecture technology, WAN Virtualization, and by leveraging multiple connections at each location (a preferred approach for providing higher availability even when using MPLS), the ability to deliver performance predictability and superior support for all IP traffic, including real-time traffic, even when using Internet connections is now quite, er... real!
WAN Virtualization provides multipath multiplexing for aggregating bandwidth and delivering reliability, and in some implementations offers sub-second reaction, dynamically engineering around network trouble – not just outright link failures, but also high packet loss or excess latency – as it occurs.
With RAID-like techniques and benefits, it delivers end-to-end reliability and predictable application performance for TCP-based applications, and ultra-reliable support for real-time traffic, even when using those "imperfect" public Internet links. By reacting to connectivity or congestion-based packet loss or jitter sub-second, it delivers responsiveness fast enough to handle any application that can run on a cloud IP WAN like MPLS.
For real-time applications like VoIP and videoconferencing, WAN Virtualization can deliver reliability and predictability by choosing network paths with the least packet loss and lowest jitter, and switching sub-second to a better path in the face of high loss or jitter. Where sufficient bandwidth is available, some WAN Virtualization implementations can provide still greater application reliability by replicating real-time traffic flows along a second path, suppressing duplicates at the receiving end, and so delivering essentially "perfect" sound and voice quality even in the face of failures or congestion on one of the connections.
Combined with WAN Optimization, colocation and the inexpensive Internet connectivity these technologies together enable, WAN Virtualization is the linchpin to knocking down the myth that only MPLS can deliver application performance predictability and enterprise quality support for VoIP.
Next time, we'll wrap up this series on knocking down the myths of why enterprises need MPLS.
A twenty-five year data networking veteran, Andy founded Talari Networks, a pioneer in WAN Virtualization technology, and served as its first CEO, and is now vice president of product management at Aryaka Networks. Andy is the author of an upcoming book on Next-generation Enterprise WANs.