Cisco says its AI crystal ball can predict network errors
- Reference: 1651686675
- News link: https://www.theregister.co.uk/2022/05/04/cisco_ai_network/
- Source link:
The networking tech giant announced today what it said is the culmination of two years of work: an analytics engine that can predict network issues before they happen, and with enough integration and training even fix problems itself, [1]Cisco said .
Citing data from an in-house study, Cisco said that 45 percent of IT leaders it surveyed cited responding to disruptions as their biggest networking challenge of 2021. Predictive analytics technology, coupled with "enormous amounts of historical [networking] data," is a potential solution, Cisco said.
[2]
Cisco's predictive networking engine draws upon various machine learning models, a company spokesperson told The Register . They said Cisco has worked with advanced statistical models (Cisco didn't state which) as well as experimented with [3]long short-term memory and [4]convolutional neural networks .
[5]
[6]
"Our research for predictive networks has been tested and developed with customers, and early adopters are seeing major benefits saving them time and money," added Cisco chief exec Chuck Robbins in a [7]canned statement. Cisco named, in [8]a video , energy company Phillips 66, Schneider Electric, and HR provider The Adecco Group as some of its early adopters.
Cisco said it has plans to incorporate predictive networking technology across its portfolio "in integrated, easy-to-use software-as-a-service" offerings. Cisco didn't state which products its predictive networking would be included with, though it did tell The Register its predictive analytics engine "will power a broad range of products and services over the next few years."
[9]Google and Cisco, sitting in a (spanning) tree, cloud N-E-T-W-O-R-K-I-N-G
[10]New phisherman's friends and a few old favourites slither out of WatchGuard's Security Report
[11]Nvidia pushes crowd-pleasing container support into AI Enterprise suite
[12]Tencent is now bigger than Cisco and Lenovo – and predicts this virus thingy will help it get bigger still
Robbins saying in the company's announcement that self-healing networks are the future of connectivity implies the biz's future plans may well center around its predictive networking engine.
As for what sort of networking data the predictive engine can process, a Cisco spokesperson claimed were "no blind spots or limitations" in terms of what it can see and monitor. Cisco cited telemetry data from applications, traffic volumes, log events, and topology as specific things it could watch. Other than that, Cisco only told us that the company has a "broad portfolio of observability, visibility, and intelligence technologies" that will feed its new predictive model.
[13]
The engine can make recommendations to administrators, such as redirecting traffic, and also has a "trusted" mode that will allow the engine to perform some automated actions. Cisco told The Register the engine could automate things like path selection, quality-of-service and configuration changes, though what it's able to do without human approval will be limited.
As for how to get it, or when it will be available, Cisco didn't publish any information nor respond to The Register 's inquiries toward that end, aside from saying it planned to talk more about its predictive engine at Cisco Live in June. ®
Get our [14]Tech Resources
[1] https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2022/m05/creating-a-predictive-network-for-the-human-mind.html
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/networks&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YnL3iSk--XBRsXbRxYNhJwAAAMs&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2017/11/02/39_episodes_of_csi_used_to_teach_robocop/
[4] https://www.theregister.com/2018/08/09/neural_network_malware/
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/networks&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YnL3iSk--XBRsXbRxYNhJwAAAMs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/networks&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YnL3iSk--XBRsXbRxYNhJwAAAMs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[7] https://investor.cisco.com/news/news-details/2022/New-Cisco-Technology-Can-Predict-Network-Issues-Before-They-Happen/default.aspx
[8] https://video.cisco.com/video/6305284344112
[9] https://www.theregister.com/2020/04/22/google_and_cisco_sd_wan_alliance/
[10] https://www.theregister.com/2019/03/21/watchguard_security_report/
[11] https://www.theregister.com/2022/01/21/nvidia_ai_enterprise/
[12] https://www.theregister.com/2020/03/19/tencent_q4_2019_coronavirus/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/networks&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YnL3iSk--XBRsXbRxYNhJwAAAMs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://whitepapers.theregister.com/
Classic late era Cisco
Releasing a buzzword complaint product after the buzz is almost gone.
Where did they get their training data, and where was the opt out for it this whole time? Or did they try to "train" it with fake/generic data? Or with the data from their own networks, which may leak some interesting information if that turns out to be true.
At least the press release tap danced around the likely reason that it can't trigger many actions on it's own. It's probably a false positive factory.
I don't need layers of crappy ML, I need a cleaner feed of events that make actual sense so I can feed the SIEM I have been required to operate for years thanks to PCI compliance. Cisco is almost as bad as M$ in that regard. Firing nuisance hundreds of times, unnecessarily garbled phrasing for key events that requires special handling, and head scratching decisions about was is and is not logged/audited by default. Also an ongoing lack of consistency on how events are logged across platforms and versions.
ML probably wasn't the way to fix that.
What if it decides that WE are the problem with the network. Article needs a Rise of the Machines tag/icon