Josh Stephens

QoS Tweaks

Posted by Josh Stephens Nov 28, 2007

Well, the Thanksgiving break is over and I'm back in the office after being sick for a few days so I reckon it's time to get back to work... I'll probably post some details about the crazy goose hunting trip I took last weekend and how I ended up sick as it's a good story and worth sharing.

Anyhow, I've been talking to some customers lately about their implementations of traffic prioritization and/or traffic filtering and I ran across some interesting concepts/trends. First off, it seems that most people fall into one of two camps - they either try to filter out all non-essential traffic or they try to significantly prioritize essential traffic. The engineers that are fans of the "filter out non-essential traffic" method are blocking all user access to well known non-essential sites (YouTube) and or traffic types (streaming video from untrusted sites). In this camp you are probably still going to implement some traffic prioritization so that latency sensitive traffic such as voice and video are weighted correctly, but the overall need for traffic priorization is less.

In the second camp engineers aren't limiting access to specific sites or filtering certain traffic types but they are spending a significant amount of time implementing and tuning their traffic optimization strategies. This isn't to say that in these cases people aren't limiting access to some inappropriate content - but for the most part it's a live and let live type attitude. In this case not only is there a need to prioritize latency sensitive traffic but there's also a need to prioritize essential traffic over non-essential traffic. In rare cases I'm seeing where companies deploy completely separate connections for essential and non-essential traffic. For instance, one company deploys DS-3 connections to all of its larger sites for internet connectivity and leverages these connections to carry the VPN connections to its other sites and for business oriented web traffic. The DS-3 connections are very locked down - essential traffic only - but each site also has a T-1 that is wide open for generic internet browsing.

I tend to fall into the second camp as I'm not a big fan of "The Man" limiting where I can surf to and everybody has had those days where a funny YouTube video is all that has kept them from going postal. One of the more creative strategies I've seen is where non-essential traffic is either blocked or de-prioritized to the point that it's not usable during most business hours but during well known "break times" such as lunch and maybe a well established breather in the afternoon the gates are opened and the YouTube videos flow like Niagra Falls in the Spring Time.

Regardless of which camp you fall into, implementing and using a solid configuration management solution such as our Cirrus product will be a huge time saver in managing your bandwidth optimization strategies. In the case above where traffic prioritization is changed based upon time of day, this would be darn near impossible to do by hand but with a simple script Cirrus can do it automatically every day at the prescheduled times.

Anyways, would love to hear your thoughts on this subject...


Flame on...


Well, I'm back in the office from vacation just in time for a short work week (couldn't have planned that better). In case anybody's curious, I've spent the last week chasing pheasants across Western Kansas and had a blast. Greg (a.k.a Sparky) one of our senior architects and Dan (a.k.a LT) our most senior tech support lead accompanied me and we had a blast. Here' a quick pic...

We're doing our 2008 budget planning right now, so if there are any really cool or especially important things that you're looking for from SolarWinds that you haven't already told us about this is the time to mention it...
Anyhow, just trying to get re-engaged here so this week I'll keep it light as I realize lots of people are out of the office. In case I forget to mention it - Happy Thanksgiving to all of you.


Hi, I’m Joel Dolisy, chief architect for SolarWinds; over the next few months I’ll try to provide insight of what’s happening in the engineering side of the house.


As Josh mentioned a few weeks ago I’m researching the types of failover scenarios we need to support in Orion, and how those relate to the existing hot-standby polling engine strategy. As usual, the complexity of the solution depends on the level of failover support that the product needs to operate under. Claiming that a solution is fail safe is actually a complicated problem to solve and involves the interaction of different players such as the OS, the hardware (computer and network infrastructure), the storage subsystem and the application.

As part of the research, I’m considering if leveraging the built-in clustering services offered by the OS (such as Windows Server 2003) would be something that would be acceptable for our customers. Leveraging those services has lots of advantages from an engineering standpoint as we can concentrate on building features that are actually useful for you instead of having to write more plumbing, but on the other side such a solution comes with its own set of constraints. Every time I encounter additional constraints I always have to weigh the benefits of the proposed solution against its additional configuration burden.
One of the big assets of our products is their simplicity of installation/configuration and that they don’t require days of planning before a rollout like a lot of our larger competitors. Every time something has the potential of disturbing this ease of use/configuration I need to make sure that I understand under which circumstances those additional configuration steps will impact our customers.
I feel that customers asking for failover capabilities are ready to take additional configuration steps required by such configurations, in fact, most of those users will already have SQL Server clustered and requiring the OS clustering service in this case should not be an issue.

At this point I would really like to hear from you what your thoughts/comments/requirements/experience on failover scenario are and how do you expect us to support them.



I'll be out on vacation bird hunting in Kansas starting tomorrow for about 10 days. Wish me luck and I'll see you when I return...


p.s. Don't be surprised if some dude with a Belgian accent adds some content while I'm gone. That's Joel our chief architect :)

I'm curious to see how many of you are utilizing diffserv vs. RSVP in your QoS implementations. Personally, I've found diffserv to be more flexible and generally easier to implement but I wouldn't necessarily count myself as an expert on RSVP. Are there any RSVP users out there?

Not sure how many of you are doing it, but you can use the Orion NetFlow application to monitor traffic levels via QoS. This feature is new as of the last couple of months, but has been helpful in some of the testing I'm doing.

Also, since I've been doing some research on QoS lately I thought I'd share a couple of interesting sites I found:



Howdy y'all... Before I get started talking about QoS, I'd like to thank everyone that helped me prepare for the webinar we hosted today on Effective WAN Management. The webinar went very well and I really appreciate those of you that responded here on the Blog and those of you that send me info directly via e-mail. The t-shirts are on the way...
Over the next few weeks we're going to be sponsoring some activities around QoS including some free web-based training, blog articles, and conducting/posting some interviews with industry experts that'll be posted as podcasts on our website. To that end, I'm looking for people that are actively leveraging QoS type technologies and/or any interesting WAN optimization techniques. Drop me a note if you fall in this category and I'll get you some additional information.

More to come tomorrow night but I'm gonna run for now as I've got some shows on the Tivo that are calling my name. Not sure if any of you out there are watching the new sitcom "The Big Bang Theory" but it's quickly become a favorite for me. It's nice to finally see some dudes on TV that are even geakier than I am

Flame on...


I'm hosting a webinar next week on Effective WAN Management and I'm trying to come up with the content. Problem is - I can't seem to put anything down on paper today. You know how it is - sometimes either that part of your brain just doesn't want to engage or the other parts don't want to give up control (probably a little of both in my case).

Anyhow, I'm looking for some help. Send me your ideas and I'll send a T-shirt your way. Caveat - T-shirts are for the first three people to contribute more than a couple of sentences. If I try to give away more than that at any one time they'll shoot me...


Filter Blog

By date: By tag:

SolarWinds uses cookies on its websites to make your online experience easier and better. By using our website, you consent to our use of cookies. For more information on cookies, see our cookie policy.