Expecting too much from Class of Service?

CoS, what is it good for? Maybe not “absolutely nothing” but not far short.

With the exception of the EF class for real-time apps, CoS actually delivers almost none of the expected benefits promised by the marketing gloss. Let’s pull it apart and see why.

CoS only operates on outbound links (no point in queuing traffic after it’s already traversed the WAN). OK, but managing traffic outbound from each end does give bi-directional control, right? Wrong. CoS bandwidth allocations are proportional to the overall outbound bandwidth which is usually much bigger at the DC end than any of the branches. Traffic in any individual CoS leaving the DC could easily flood the smaller inbound branch links.

OK how about outbound from the branch? Nope, not much better. Normally the outbound branch traffic runs at a lower volume than inbound. CoS only kicks in when the link reaches capacity and this rarely happens. Worse still, the sum of many branch networks can oversubscribe the inbound DC bandwidth and as there’s no inbound control at the DC end the traffic profile can be chaotic.

In summary, CoS has no meaningful control of traffic outbound from the DC or outbound from the branch. CoS cannot help with inbound control. Include the possibility of branch to branch traffic wrecking best laid plans and any notion that CoS is useful flutters out the window.

If you want to monitor CoS utilization and other end-to-end Path characteristics like packet loss, RTT and jitter try PathView . If you need to manage traffic it’s far better to use traffic engineering devices like PacketShapers, even if you can only put one in the DC. We provide PathView and PacketShapers separately or bundled in low cost service options.

Riverbed wins at WAN optimization but there’s more to the mix than just that

Although Riverbed won the competition for WAN Optimization, it’s heartening to see that NW recognize that WAN Optimization alone is a bit of a one-trick pony these days. Visibility, bandwidth management and prioritization are critical components to match applications with the appropriate style of control.

“…However, WAN optimization has evolved to encompass other features, such as traffic management and visibility. Here, we find that Riverbed’s Steelhead could use some work. (Its) Traffic management is good, but not great. Visibility is limited in a way that pushes network managers to use Riverbed’s own tools, rather than opening up to the growing world of standards-based flow analysis products. And new features, such as WAN path selection, don’t live up to Steelhead’s traditional technology leadership.”…”

When you see the indicative pricing, you’ll need to add some serious contingency to include visibility for the players that don’t quite make the grade.

Read the article here



Five Areas That Will Shape the Application Performance Management Market in 2014 from TRAC Research


TRAC’s research shows that 61% of organizations are listing “demonstrated effectiveness in relevant usage scenarios” as the key criteria for selecting APM solutions.

Customers are more likely to select solutions that are best aligned with use cases that they care about, as opposed to looking for the most complete solution in the market overall. They want APM solutions that are easier to buy, deploy, and manage with time-to-value being by far the most important selection criteria.

See the report here

Poor web performance? Hit your own Too-Slow button

We were discussing how to know when a user was unhappy about the performance of a browser based application and how they might most productively vent their spleen about it. My suggestion was to have a button on the browser named ‘Too Slow’ and encourage users to click on the button if they felt a page was loading too slow. Here’s one way which is free and easy. There are others.

The process had to be quick, simple, timely, and completely intuitive for the user with zero training. And, it had to provide maximum detailed information for technical resources capable of analysing the information.

I loaded the free performance bookmarklet from GTMetrix into my browser (Chrome) and changed it’s name to ‘Too Slow’. I viewed a page that was too slow (the Linkedin group page of the discussion, actually) clicked on my ‘Too Slow’ button and shared the resulting report to Google+ limited to within my own organization. The whole process took less than 30 seconds and cost me the effort of two whole clicks. It produced this report.

If I, as a user, cared enough to complain about a page load time, I could certainly handle 2 clicks to register my complaint and automatically provide a detailed breakdown of the incident to the right people. I didn’t need to understand what I was sending, just be confident that it captured every technical detail for someone else to analyse.

That said, in addition to the detailed reports about performance it also spells out the things you need to do to improve performance with reference to standards published by Page Speed and Yslow

This will give you a one-time snapshot each time a user is willing to do it and it is a good way to focus on the pages that users decide are problematic. Then you can automate that navigation trail with continuous monitoring using Appneta’s PathView / TraceView suite. It includes integrated browser waterfall, detailed network performance stats and code-level application internals analysis. Quick, easy and amazingly comprehensive.

I still have the ‘Too Slow’ Button on my browser tool bar and often use it to delve into poor web page performance, but soon move on to the next problem after Appneta starts taking care of the continuous duties.