Hollywood tantalizes us with super high-tech futures where you ask the computer verbal questions like “Computer, what happened with the hyperdrive?” and the computer responds “The hyperdrive is inoperable because the phase-capacitors are out of alignment.” We never think about the alternate of not having the computer to do this analysis: engineers fanning out around the ship to collect information and analyze hundreds of elements and then report back to the bridge to determine the fault – the computer creates movie magic that wows us.
A number of months ago, our engineers were having a beer while watching one of these movies and realized that we could achieve the same thing for networks if we built a natural language troubleshooting interface.
TotalView already has more information on the network’s operation and performance than any other product: error counters, configuration information, queueing, usage. Being able to analyze this dataset with a single plain-English query to determine what’s broken would make troubleshooting significantly easier, and let anyone in the organization perform the troubleshoot.
This means that senior level engineers no longer have to work on trouble tickets – they get to focus on more fun strategic level projects (I never met an engineer that loved to work tickets).
For example, helpdesk or telecom folks could determine where a user had a problem and get it fixed with a single query: “What happened on the network between 10.41.161.32 and 10.6.34.18 at 2:35pm?” and the system would produce a report with a plain-English response: “The Finance2 switch interface #3 dropped 6% of its packets due to a cabling fault at that time.”
Resolution: New patch cord replaced old CAT3 patch cord.
It didn’t take another round of beers to have all of us realize that the next logical step would be to change the user interface to be spoken voice -- we should integrate with Amazon Echo.
We have seen a couple of companies provide a “playtoy” Echo interface that allows you to query how many alerts exist, or how their product is licensed. That’s fun, but we wanted something that was real, functional, and useful in a day-to-day environment. That meant that we had to make sure that the parsing could interpret the intention beyond the specific words. We are excited that we’ve found a way to make this a reality, and provide useful results that can improve network operations in a meaningful way.
We have decided to offer this feature for free to customers who have valid support contracts, or for a fee if the software is purchased after it is released.
At this stage, we’re knee-deep in the development and QA efforts with this, and it’s looking really cool!
We will provide additional updates along the way as we accomplish various milestones towards completion.
To learn more about NLT, click here.