In part one of this series on prioritizing WAN application delivery, find out why simple WAN QoS settings won't ensure that critical business applications will always get the network priority they need. In parrat two, we discuss which technologies can help with WAN application delivery.
Prioritizing wide area network (WAN) application delivery is not as simple as it used to be.
Network engineers have historically relied on carrier-class WAN QoS tagging to prioritize WAN application delivery; however, Quality of Service (QoS) is not a panacea for protecting critical applications. The number of applications that require protection from latency and bandwidth constraints is growing within enterprises. VoIP over WANs is now mainstream, and it has no tolerance for latency, jitter or packet loss. Virtual desktop infrastructure (VDI) traffic is also hitting WANs, and users will scream if they notice any latency with their mouse clicks. Video communications loom as well.
"These sorts of latency-sensitive traffic streams are becoming more common," said Jim Frey, research director with Enterprise Management Associates. "Anyone who is planning their WAN architecture needs to understand [that] deploying these kinds of technologies … [requires a] plan for how they're going to manage their WAN and optimize it."
Why WAN QoS alone can't guarantee WAN application delivery
With no universal QoS standard among carriers and vendors, there is no guarantee that the QoS tags an engineer applies to his applications will work in all settings. "You don't necessarily know what is going to happen once you get out into the provider network," Frey said. "Although QoS tagging is done at the edge and is generally honored as you go through the rest of the delivery chain, that oftentimes isn't enough."
Carrier-class QoS tagging is also very inflexible and enterprises find that the nuances of a diversified set of applications don't fit.
"The QoS you can get from carriers right now is kind of a blunt instrument," said Mark Blomquist, chief technical architect for Atlanta-based NCDR LLC, a provider of nonclinical services to Kool Smiles, a national provider of dental care. "It kind of works, but it doesn't perform well when things go haywire on the network."
NCDR provides IT services for more than 100 Kool Smiles sites, including an MPLS network that CenturyLink delivers. While NCDR's service-level agreement (SLA) with the carrier promises three nines of reliability, it rarely delivers on that SLA in some of Kool Smiles' more rural sites. That level of service can be highly disruptive to critical applications like VoIP and Citrix-based thin clients. Blomquist wanted the ability to fail over to another provider when CenturyLink's MPLS went down. Unfortunately, WAN QoS tags don't readily translate from one carrier's MPLS cloud to another carrier's.
"CenturyLink will give me three buckets for QoS: real time, interactive and bulk. I can tell my Cisco router, 'Here is how you're going to prioritize your traffic when you push it out on an MPLS link.' And you can load a T1 line to 100% capacity and still hold a voice call. But you can only do that on a single link, so if you need any kind of carrier diversity, you can't do it. If you need to do anything with the traffic other than just put it into a bucket, you can't do that. Let's say you want to [duplicate] the traffic [over multiple links], or if you are missing one packet out of 40 when you try to say something, the Cisco router doesn't have the ability to reassemble things."
WAN application delivery: QoS can struggle with Web-based applications
More on WAN application delivery
The cloud shakes up WAN application delivery
The new age of application delivery: more than just install
Top five WAN application delivery mistakes
Tips for application delivery optimization
With WAN QoS, an engineer can usually tag voice for high priority, and one or two other applications for middle priority, but most applications get shuffled into a bulk queue. Many enterprises are dealing with an application landscape that is far more complex.
"Just a couple of years ago people would say, 'We have five or six -- maybe eight -- applications that are business critical.' These days I've seen tremendous growth in that number," said Jim Metzler, vice president with the consultancy Ashton, Metzler & Associates. "I think it's just a natural evolution where more and more business processes are run over IT. If the applications and infrastructure aren't doing well, then neither are those business processes."
Finding the applications that need prioritization is easy, Metzler says. Just keep track of what users complain about the most. Figuring out how to prioritize WAN application delivery is far more difficult. Five or six years ago, an engineer could assign QoS settings by network port number, but applying QoS at the network layer doesn't work like it used to.
Originally, enterprises deployed QoS to prevent users from misbehaving, according to Sidney Rabsatt, group product manager at Riverbed Technology. It would throttle down traffic on the network ports that corresponded to the FTP and HTTP protocols, and business applications would be fine.
"[QoS] made sure [users] weren't doing a lot of peer-to-peer and recreational Web browsing,” Rabsatt said. "The presumption there was that application performance could be good enough just as long as you can remove all the bad influences."
"Everything works now over HTML and HTTP, so even internal applications are more Web-based. Even document management systems and ERP systems are running over Port 80 and Port 443," said Mark Urban, senior director of product marketing for WAN optimization solutions for Blue Coat Systems. "The old way of doing QoS was, 'Hey, I have my Microsoft Exchange working over Port 137, FTP over Port 121.' So you were able to do a lot of QoS rules at the network port level because you had rough equivalencies between port numbers and applications. Now SharePoint is the main way people collaborate or intranet pages are the way people post documents within a company, and that goes over Port 80 or secure Port 443. The webification of applications has really broken the linkage between network port numbers and what applications are."
The application landscape gets even more complex as enterprises start deploying cloud applications. For instance, enterprises typically access software-as-a-service (SaaS) applications via HTTP or HTTPS (Port 80 or Port 443).
"How do I find 10 cloud-based applications, and how do I identify those out of the millions of websites out there? And what happens if those applications are delivered over SSL, which is secure and encrypted?" Urban asked.
A number of vendors are moving beyond old-fashioned WAN QoS. They are developing WAN application delivery technologies that focus on the application layer rather than just the network layer. These technologies have familiar names: WAN optimization, content delivery networks and application delivery controllers.
In part two of this feature, we will explore how enterprises can move beyond QoS and guarantee WAN application delivery for critical business applications.
For more information:
- Learn how to achieve application visibility through network visibility in this tip.
- See which WAN application visibility solutions are available to enterprises.
- Get a better understanding of application performance monitoring and management (APM).
Let us know what you think about the story; email: Shamus McGillicuddy, News Director
Dig deeper on Application performance on the WAN