NETWORKING

  • 05/20/2015
    8:00 AM
  • Rating: 
    0 votes
    +
    Vote up!
    -
    Vote down!

Network Performance Monitoring, Easy As Pi

Network monitoring startup NetBeez harnesses the value of the Raspberry Pi to test application performance in distributed environments from the user point of view.

Veterans United Home Loans had monitoring systems in place to gauge the health of its wide area network, which connects more than 30 locations across the country. But the IT staff was still receiving user complaints about application and network slowness. That was a problem for Eric Goodwin, the systems architect responsible for the network. "We want to know about issues before trouble tickets begin to swamp our help desk," he said.

Goodwin discovered the culprit was an infrastructure-centric view, or what he called an "inside-out monitoring approach." He explained, "We had the right tools in place to monitor our network from our central data centers. It only gave us a one-way picture of what was happening out on the network."

The troubleshooting process at Veterans United was also painstaking and time-consuming. When faced with a poorly performing application, operations engineers would start at the network level and verify the first three layers of the OSI model were functioning to rule out network connectivity as the problem. That involved logging in to multiple routers to run trace and ping commands and collecting the appropriate data, and sometimes meant disturbing end users in order to send ping and traceroute data from their workstations.

The company decided it needed to better understand how applications were performing from the perspective of those end users in all its distributed locations, and turned to NetBeez, a company incubated at the AlphaLab startup accelerator in Pittsburgh. NetBeez provides monitoring from the outside-in by distributing agents at the very edge of the network and providing a cloud-based or server-based unified console. 

Looking at applications from the user perspective is not a new idea, but it is one that's been difficult and expensive to implement. NetBeez took advantage of the popular Rasperry Pi microcomputer to create an easy-to-use and low-cost solution for IT departments, company co-founder Stefano Gridelli told me at Interop Las Vegas. A former network engineer for a large healthcare organization, Gridelli experienced the challenges of Veterans United as well. "The tools I was using could tell me if switches and routers were running properly, but they couldn't tell me much about what users were experiencing. So we developed NetBeez."

In a typical NetBeez deployment, hardware agents (Beez) are placed on wired ports or connected to the wireless network in each remote location. The Raspberry Pi platform is intended for small locations and supports FastEthernet. The company also offers two versions for Gigabit Ethernet, a wireless version, and a virtual version, said Gridelli.

These agents simulate user activity by continuously testing network services using simple tests like ping, traceroute, and iperf, along with http and DNS queries. The agents use the results to constantly verify application delivery and report response times back to the central server. Information from the Beez is analyzed and displayed on a central dashboard, where networking staff can see at a glance when problems in availability are cropping up, exactly where they are located, and with which applications.

Most importantly, the Beez can tell IT staff whether a problem is with the network or the application. "That can cut your troubleshooting time in half," noted Gridelli, especially in complicated environments involving distributed networks and cloud applications.

At Veterans United, Goodwin and his team were able to target the applications most critical to them. NetBeez allows the company's operations engineers to begin identifying network-related issues down to the individual application, and to confidently diagnose whether the network is part of the problem or not, said Goodwin. The monitoring platform also provides data they can give to the application administrators to better resolve application issues, he noted. The product helps the overall team to troubleshoot earlier and more effectively, and to be more proactive when issues arise.

Goodwin related a recent incident in which a third-party application provider decided to change the advertised public IP address to another address outside of the range of Veterans United's dedicated connection. "The NetBeez agent for that application reported a change in the traceroute and identified the exact time they changed the advertised public IP for the application. NetBeez not only alerted us that it happened, but it also told us when it happened. That is cool."


Comments

end user experience

This is a really interesting case study Sue. I'm curious if Veterans United looked at any other technologies like application performance monitoring tools before choosing NetBeez?

Re: end user experience

Marcia, the company was looking for something low-cost and easy to implement and said typical applications management software (Oracle, IBM) is neither of those things, and requires people trained in that plaform to manage. It also slows down the network a lot, according to Stefano at Netbeez. This is a far lower-cost option because it uses basic network tests and commodity hardware agents.

Re: end user experience

Hi Marcia,

 

NetBeez is currently one of several monitoring platforms we have implemented.  We are currently monitoring application, network, and server performance with other monitoring platforms.  NetBeez gives us more of an end user view of what is currently happening in relation to trace routes and packet loss/latency from remote offices into our data centers.  We have very distributed network of more than 30 locations.  NetBeez gives us the ability to measure, report, and alert on network latency/packet loss from a per location perspective.  We have been able baseline latency across each dedicated WAN connection at each remote office.  If latency or packet loss increases above the desired threshold, then alerts can be sent out, and we can start addressing the issue.  NetBeez gives us a distributed monitoring visibility that our other monitoring platforms cannot provide.  I hope this answers your question.

Best Regards,

Eric Goodwin

Re: end user experience

Yes, thanks so much providing those details Eric! It sounds like the distributed monitoring capability of NetBeez is key. Does it make any of your other monitoring tools expendable?

Re: end user experience

I hope they are doing route analysis for better monitoring, i have always had problem when incorrect routing or routing issues cause undesirable performance degradation or downtime.

Re: end user experience

aditshar1 there is an Information Week article that covers a routing-change use case that occured at Interop Las Veags. Search for "InteropNet Live: Behind The Scenes With NetBeez" (can't post url because it won't be allowed by the moderator). You will find it interesting.

Re: end user experience

I agree your point @aditshar, but this sounds like one of the common problem in monitoring scenerios, neither it is recommeded to add or delete via gui interface, how good is the parctise at your end.

Simple Networking tests and Application Slowdown

"...These agents simulate user activity by continuously testing network services using simple tests like ping, traceroute, and iperf, along with http and DNS queries. "

Maybe I am missing something here but how does this information tell an engineer whether it is the application or network that is the cause of the slow down ?   There has to be more to it because these are just network checks IMO.

Re: Simple Networking tests and Application Slowdown

That is a great point, I am assuming that the application and network are two areas of interest and the third area of interest is the user hardware. It could be due to security requirements that software cannot be installed on the user hardware, so the next best option is to have hardware (in this case, the Raspberry Pi) right next to the user hardware. Once everything is in place, application errors and network errors can be ready identified. 

Re: Simple Networking tests and Application Slowdown

ClassC, yes the tools does use very simple network tests. It correlates the reponse time to remote locations and *per application* so that engineers know where to look for trouble.

Hah

From the headline, I thought this was going to be a reference to the joke: "My password is the last seven digits of pi."