freshidea - Fotolia
Ivan Pepelnjak, writing in ipSpace, questioned the evolving role of machine learning in network traffic routing. His assessment came in response to a question posed by fellow blogger Russ White, who suggested using machine learning for network traffic routing, supporting the use of neural networks to understand traffic flows and adjust them over time.
"As fancy as it sounds, we don't need machine learning to solve those problems," Pepelnjak countered. Instead, he said, commercial products such as MPLS traffic engineering -- perhaps dynamically adjusted using NetFlow data -- can accomplish the same goals.
According to Pepelnjak, some organizations have already found other ways to overcome the network traffic classification limitations he said make machine learning impractical. He gave as examples Facebook, Google and Fastly -- all of which manage outgoing traffic on edge servers, where it is cost-effective to host complex tables and algorithms.
"Until the rest of us get there, we'll be dealing with a pretty coarse-grained knapsack problem," Pepelnjak wrote, referring to the famous mathematical puzzle, "and there's only so much you can do there."
"However, as the knapsack problem is an NP-complete problem and cannot be solved perfectly for large datasets, we might get to a point where machine learning algorithms give us better results than the best heuristic algorithms we manage to develop, but that's a far cry from what we're being promised," he added.
Read more of Pepelnjak's opinion on network traffic routing.
Cloud computing shifts identity management
Jon Oltsik, an analyst with Enterprise Strategy Group, or ESG, in Milford, Mass., recalled a conversation a few years ago with a chief information security officer. The security executive found his organization struggling with tight security controls as it shifted more of its workload to the cloud. As a result, identity management and data security became far more important.
Today, little has changed, Oltsik said, citing recent ESG research that found 61% of organizations believe identity and access management (IAM) is more difficult in 2018 than it was two years ago.
ESG identified key focus areas for enterprises working to overcome the cloud computing chaos that can plague IAM: single sign-on, multifactor authentication, IAM centralization and the development of IAM skills. A more centralized approach to IAM can help, he said, but a critical shortage in IAM specialists could derail organizations' strategies.
Identity and data are the new security parameters, Oltsik said. "It’s time that organizations realize this and fortify themselves in both areas."
Dig deeper into Oltsik's ideas about IAM.
Questions a network administrator should ask AI vendors
Mike Fratto, an analyst with GlobalData in Sterling, Va., recommended a set of questions network administrators should pose to artificial intelligence (AI) vendors.
Established vendors and startups are working on AI and machine learning (ML) products for networking, but many gaps need to be closed to ensure full reliability. Fratto cited Cisco's network analytics engine and Mist Systems' Virtual Network Assistant as recently released examples using sophisticated user interface elements to assist in root cause analysis.
Yet, Fratto said these advanced techniques often fail to deliver.
To that end, Fratto said AI vendor questions should focus specifically on a few areas, among them integration, how long it might take for the vendor's software to learn about other products and applications, and areas where integration gaps and blind spots might exist. Additionally, he recommended gaining a deep understanding of the AI vendor's five-year roadmap.
"AI and ML are coming to IT and network management with useful and practical outcomes," Fratto said. "I see these technologies only growing more sophisticated over time as vendors pour more intelligence into their analytic engines."
Explore more of Fratto's questions for AI vendors.