It is not so clear that AIOps is going to be widely deployed in less than five years, argues Frank Yue, Kemp Technologies solutions architect.
“AIOps is more complex than you think,” said Yue. “It is hard for AI to figure out all the relevant connections” between apps and the network.
“Network performance monitoring is separate from app performance monitoring; security policy management is different from the separate analytics for each subsystem,” Yue said. The network, server, app, security, storage, WAN and cloud all have to be integrated.
Then there are the vendor specific solutions that are proprietary, and many different protocols are in use, including HTTP, DNS, BGP, Java and .net. Different languages including XML, Syslog, RTTP REST, SNMP, CLI also must be accommodated. There also are different functions, including server load balancing, routers, switches, Yue said.
It is hard for AI to know, in advance, the context of relationships between systems, apps, actions, he said. So a complete AIOPs platform must be multilingual, multi-vendor, mult-technology and multi-system.
And “nobody wants to buy 10 AI solutions,” he added.
There is, in other words, no present way to build a complete and comprehensive monitoring capability with a single pane of glass.
No comments:
Post a Comment