Hey, How Much Bandwidth!? I’m Paying For Twice the Bandwidth I’m Getting! by Team AppNeta March 3, 2012
Filed under: Performance Monitoring
As a network engineer for a global financial services organization, I know I am not alone in my daily concerns about network connection speed and performance. Our concerns are all underscored by questions such as:
‘How much bandwidth are we getting?’
‘What are we paying our ISP monthly again?’
‘How are my end users using our bandwidth?’
‘Why is my app so slow?’
The first day I used PathView Cloud these questions were running through my head and PathView Cloud’s integrated capabilities helped me easily find the answers.
PathView Cloud provides a continuous, minute-by-minute representation of the available capacity versus utilized capacity. This information is supplied in a cleanly formatted graph with highlighted points of interest. As shown in the screenshot above, you see the provisioned capacity to validate whether you are getting the bandwidth that you are paying for. Notifications can be generated for high utilization, low bandwidth and many things in between.
This was particularly helpful for the VoIP system I had been expanding to some of our remote sites. I used PathView Cloud with AppView Voice to load the network with voice traffic for an extended period of time and used voice ramping to get an overall analysis of voice performance. I used AppView Voice to run a few individual voice load tests to get a capacity reading for a single point in time. PathView Cloud actively monitors bandwidth and network performance 24/7 but you also have the ability to run periodic live application loads on the network.
When I first saw that I was only getting 300Mbps on a Gigabit WAN link across my state, I challenged the results with an AppNeta systems engineer. It seems AppNeta had anticipated my curiosity about how the solution can so accurately determine the size of my WAN circuit without actually flooding it. For comparison purposes, AppNeta developed its own (FREE) perfect flooding tool called PathTest so users can see how solid its non-invasive, non-flooding PathView capacity analysis is. The engineer enabled the PathTest utility on my PathView microAppliances. PathTest pushes your network to its absolute limits using precision packet-flooding techniques to completely fill the network path for a user-defined duration. However, if you run the test during production hours you risk impacting other users since the tool will flood the pipe to its maximum capacity, making it nearly impossible for normal network activity to continue.
The thing I found interesting about PathTest, which is better than other flooders like iPerf, is that you can specify everything from, duration, packet size, Quality of Service (QoS), Maximum Transfer Unit (MTU) and even protocol.
I ran PathTest after business hours, so as to not affect production. I pay for a 1Gbps link between my Headquarters and my Datacenter. So, I was expecting something around 900 Mbps. With the PathTest tool, we were able to flood my pipes in both directions simultaneously with different protocols. This is to see if there was an issue with a specific type of traffic or a specific direction. The graph below shows all of our results broken up by direction and traffic type. “Bidi” means a bi-directional test where inbound and outbound links are flooded at the same time.
I ran a UDP test in both directions to see how much traffic was generated and how much was received on the other side.
I also ran a test with ICMP and as ICMP includes the echo response, you can see the outbound and inbound traffic.
The verdict? PathView Cloud’s built-in non-invasive capacity analysis was dead on with the results of the PathTest flooding tool; yet I only gave up around 2kbps of monitoring overhead. I have a six figure WAN bill so I have to know I am getting the bandwidth I am paying for. Being in the network monitoring world for a while, I am very comfortable with network traffic generators. But, the ability to get non-invasive accurate bandwidth readings 24/7 is game-changing for me and my IT team. However, to put my mind at ease, I felt I needed two sources to verify the metrics I was seeing. By using PathTest to flood my network to its full capacity I confirmed the same results as PathView Cloud and verified that I was paying for WAY more than I was getting