Demand more with Juniper’s industry-leading AI.

Based on reinforcement learning:

- Optimizes channel/power with AI-based reinforcement learning

- AI continuously maximizes User experience (SLE) and minimizes interference in real-time

- Adapts dynamically on an ongoing basis while network under load learning from client experience

- Learns and deprioritized triggered DFS channels to boost network uptime

- Coverage SLE is an ongoing 'Site Survey'

Watch the video

Basic RRM

- will monitor DFS failure patterns

- AP's remember their settings through power failures

- Won't make changes in 'busy hours'

ARM - Basic pattern recognition for comparing and optimizing low-level RF settings only across managed sites:

- Not a true AI solution: doesn’t leverage reinforcement learning to improve over time

- Doesn’t adjust RF to maximize user experience

- Analyzes periodical and static data for daily but not ongoing dynamic updates

- Requires Controller and Mobility Master for AirMatch RF optimization

- Requires data collector appliances and NetInsight server

15-year old algorithm

- Based on how APs hear each other

- Optimizes channel/power based solely on AP interference graph

- RRM is performed on a static, periodic basis when the load is low

Basic RRM. No AI/ML, requires several days of tuning.

- Continuous learning through Supervised Machine Learning

- Performs root cause analysis for most detected network issues

- Supports wireless, wired and WAN at a site level

- Troubleshoot issues instead of pulling logs

- Can be accessed through WebUI or API

- Built on 6 years of continuous learning and rich data science toolbox

- Dashboard

- No virtual assistant

- Dashboard

- No virtual assistant

- Dashboard

- Chatbot rumored but not productized or available to customers in beta

- Dashboard and network assistant only on cloud.

- Chatbot called Co-Pilot, very limited, No AI. Allows NLP version 1.0. No query.

- In beta the last 2 years.

- Proactively identifies anomalies and uses data science tools to determine root cause

- Leverages both Wired and Wireless SLEs for anomaly detection

- 3rd generation algorithm with ARIMA boosts efficacy

- Anomaly detection performed across Wi-Fi, LAN, WAN, Security Domains

- ChatGPT integrated

- 1st generation anomaly detection algorithm

- Will go through a weeks worth of data to find some basic anomalies

- Limited set of anomaly detection (DHCP, AAA, RF utilization)

- Requires NetInsight Data Collector appliance

- 1st generation anomaly detection algorithm

- Limited anomalies detected (DHCP, AAA, Association, Throughput)

- Requires Cisco DNA appliances (3+)

Client 360 tracks basic anomalies.

Pilot and CoPilot supported.

1st generation anomaly detection algorithm.

Limited anomalies detected (Latency, Throughput, airtime).

- Marvis Actions Framework for self-driving or driverassist mode (e.g. RF optimization, proactive RMA, unhealthy APs, missing VLANs, bad cables, switch config errors, etc.)

- Validated by Mist

- Customer Service to solve or help train system

- Closed loop feedback providing actionable intel to administrators “bottoms up”

- Dashboards

- No self-driving capabilities

- Will offer “suggestions”

- Top down

- digging

- Dashboards

- Lacks self-driving, only having “driver-assist” capabilities where it provides recommendations to IT

- Very basic driver-assist capabilities (identifies channel utilization issues and poor DHCP/AAA performance for IT to manually investigate)

- Top down digging for next generation log files

- Dashboards

- No self-driving capabilities

- Top down Need to ‘nominate’ troubled user to begin any active monitoring

- Dashboards generated by basic math.

- Lacks self-driving, only having “drive-assist” capabilities where it provides recommendations to IT

- Limited self-driving capabilities (Latency, Throughput, Airtime)

Creation of probability surfaces in the cloud and ongoing unsupervised machine learning to constantly update the model.

- Triangulation dependent on accurate map placement

- Errors introduced by variance in BLE clients

- Triangulation dependent on accurate map placement

- Errors introduced by variance in BLE clients

- Meridian sidelined

- Requires CMX appliance onsite (even for DNA Spaces)

- Requires 3rd party BLE integration

- Triangulation dependent on accurate map placement. Errors introduced by variance in BLE clients

No

- Mist Support utilizes Marvis to troubleshoot issues

- Marvis efficacy is continuously evaluated and when support issues arise where data or answer is not available, we train Marvis or add the missing data collection

- When Marvis detects a hardware failure in an AP, it can perform an automatic RMA minimizing the ‘burden of proof’ on IT teams rather than escalating issues with a vendor

- As AP deployments have grown at a rapid pace, support tickets have remained flat due to the use of Mist AI

- Dashboards

- No use of AI to automate support or support operations

- Dashboards

- Lacks automated support capabilities driven by AI

- Aruba AI Assist is a basic manual button to gather logs to email to Aruba Support for manual analysis

- Dashboards

- No use of AI to automate support or support operations

- Dashboards.

- Lacks automated support capabilities driven by AI

- Proactively captures packets when an error event occurs in real-time

- Eliminates need to reproduce issues as every failure has a PCAP starting before the failure and playing though it

- No more sending out tech folks with sniffers *after* the problem has happened

Watch the video

Manual

- Primarily manual - limited auto capture on authentication failure events

- Requires an additional, separate cloud dashboard for troubleshooting and analysis (Cape Networks)

- Requires overlay network of Aruba UXI wireless sensor hardware

Intelligent Packet Capture

- first a client needs to file a ticket

- then the client will be tagged to collect data going forward

- not at all automatic

No.