Kailem Anderson joined Ciena, the fiber-optics mogul last year as the VP of portfolio and engineering division of Blue Planet, a sister company of Ciena. Anderson is tasked with the job of finding a solution to the ease the pain of people who are spending hours in network management to keep them running without any issues.
Network management involves a lot of precarious tasks including monitoring, configuring and troubleshooting the servers in the network and keep it running in a smooth manner according to the network requirements of the organization.
Anderson confirms that Network management could suck all your energy up as he has been there, seen it all for almost 12 years in Cisco Systems. He says ”My everyday job is to take care of the network, manage it, monitor it and even use some experts and analysts to monitor and fix issues in the network, create a task force immediately and fix things when sees an alarm and continuously work on defining a set of rules for network management”. Anderson talks with an Australian accent and you could never sense the pain behind those words as his voice is so breezy and some sort of humor added into that.
Blue Planet registered a revenue of $26 million in 2018 in the overall $200 million revenue in the software division of Ciena. Even though the revenue made by Blue Planet is just a tiny fraction in the overall revenues of $3 billion in total, still people in the Wall Street see that the Blue Planet division is very much an integral part of the overall strategies laid out by Ciena as it working on a strategy to customize the network in an easier manner.
When network admins see a red flag in the network, it requires a complete analysis of the overall network and scrutinizes where it went wrong at the beginning so that they can arrest the source of the problem rather than solve the problem temporarily. It requires some deep detective work at various levels of the network to locate the issue. If you are lucky, that you find the issue at the bottommost layer, where it deals with the physical medium of transmission of the network, you can easily fix the issue by replacing faulty co-axial cabling or some fiber optic links, etc.
But when you are having a hard day, you need to dig deeper and go to the next level that deals with raw bits of data that are packaged into bundles known as Ethernet Frames and moving p that layer is where the data is packaged and converted into packets containing internet-addresses along with data that contain information where these data needs to be routed to etc. And moving on you go to the top levels of data from layers 4-7 where you will be looking at actual domain and you can get all the information about the source of these data and how these bundled and packaged as internet packets and see if there is an issue brewing over there.
A simple issue like a failure with a transponder can cause a huge slowdown in the network. This happens because the failure can lead to a change of the routing in the MPLS (Multi-Protocol Label System) and a single link will be tasked to carry the role of heavy traffic slowing down the network. And when a system administrator receives a red flag in the network, he is only given a few minutes to fix the issue as the slowdown could affect the development work of many programmers and developers connected to the network.
Hence, the network admin starts looking at the issue desperately. He checks the signals coming from the OSI layers, the system log and also monitors the data flow from each equipment connected to the network to identify the root cause of the problem. Anderson adds, “Honestly speaking it won’t give you the solution or help you locate the problem sooner as it might even look like a bandwidth problem or a routing issue before you locate the real culprit in the transponder failure.
Even though the problem is solved finally, precious hours of development is lost in the meantime and this Blue Planet is coming up with a solution in that involves the software program that explores various patterns and steps. The program is specially designed to identify scenarios that could have created this problem and help you trace the root cause of the issue much quicker than before.
Anderson says that its time to use the advancement in the technology to make the most out of it to label network events to enable the program to run through a series of patterns before zeroing down on the actual defect. He adds that machine-learning models were used in creating the tools necessary to identify the patterns.
Anderson further adds “It looked a bit impossible five years ago when he thought about this concept of labeling the network events, but now it seems quite achievable with the maturity of AI and even a small bit of tagging could reap you big rewards”.
According to Anderson, the move towards machine-learning in network management has enabled him to take a comprehensive look over the network and gives him a lot of options in restructuring the network according to his requirements easily and quickly.
In a company like Cisco that is handling millions of servers, a red flag means looking at millions of data and with just a small set of people, it is often difficult to filter the information and that is why system admins turn-off the information or look at only information that is relevant. Anderson says, “this is where the machine-learning model helps in analyzing every bit of data and using the labeled events, it can easily help you identify the root cause of the issue quickly”.
And no more sleepless nights for the network administrators as the machine-learning model is there for the rescue and help them close tickets quickly.