To build intelligent security solutions correlate alerts, get additional context to speed up investigation, automate security operations, use Microsoft Graph Security API.

• A federated system
• Microsoft Graph Security API
• Security API notification subscription
• Secure Score, system wide
• Identity security score, AD

https://www.microsoft.com/en-us/security/intelligence-security-api

• GA of Cassandra API
  • Table API
  • MongoDB
• Keyspace ?
• CQL Query Builder
• CQL Shell (CQLSH)
• Cassandra Stress, to verify performance
• Scale throughput
  • Change the throughput for any table
• SDK access
  • org.apache.spark...

Ignite 2018 session: https://myignite.techcommunity.microsoft.com/sessions/65369?source=schedule#ignite-html-anchor

Cheatsheets: https://docs.microsoft.com/en-us/azure/cosmos-db/query-cheat-sheet

https://www.codeproject.com/Articles/33052/Visual-Representation-of-SQL-Joins

Net customer/user Satisfaction (NSAT)

NSAT is an index of net satisfaction scores across a number of segments and audiences as measured by the Global Relationship Study.

• Gabe Monroy
• Kubernetes, container orchestration
• Portability
• Extensible
• Self healing
• Kubectl
• Etcd
• Kubelet, heart beat back to the API server
• Ingress
• Kube-proxy
• Lift and Shift application moving strategy, without refactoring your app
• Azure managed control plane
  • Devops project, Azure resource
  • Azure Dev Spaces ?
  • Helm chart ?
  • CI/CD pipelines
  • Key Vault is used to store secrets for Kubernetes
  • Vnet ?
  • Traffic Manager, in Azure
  • ACI ?, Azure Container Instances
• Azure Monitor for containers
• RBAC control for fine grained Kubernetes resources access control
• Scenarios
  • Lift and Shift to containers
  • Micro services
  • Machine Learning scenarios
  • IoT
    • Azure IoT Edge
• Maersk
  • NSAT, customer satisfaction
• Terraform
• CaaS
• OpenAI, a not for profit organization
• Azure IoT Edge

Open Neural Network Exchange is a format for deep learning models that allows interoperability between different open source AI frameworks.

ONNX is being co-developed as an open-source project by:

• Microsoft
• Amazon
• Facebook

https://en.wikipedia.org/wiki/Onnx

• Event Notification pattern
• Tight coupling prevention in the design
• Eg adding a queue between the order module and the shipping module
• System becomes more resilient
• Events should have as much context around then as they can
• Event sourcing pattern, writing the events to a log
  • Audit History
  • Decoupling
  • Materialized views
  • Integration with other systems
  • Teleporting ?
• Microservices
• Fault tolerance
• Functions can be triggered based on an event

• Azure CosmosDB
• RU: Request Unit
• Five consistency models ?
• A multi model database, key value, column family, documents, graph
• ARS, Atomic Record Sequence
• Schema agnostic
• Change Feed: log updates or others
  • Triggers for Azure Functions
  • Stream Processing
  • Data Movement, moving the data downstream
• Multiple consumers of the change feed

• Azure Functions
  • Bindings and Triggers
  • Proxies
  • Monitoring
  • Local Debugging
  • CI/CD
  • Doable Functions, will save their own states
• Azure Event Grid
  • Event handlers: Functions, etc
• Use the Change Feed to light up your micro services architecture
• Jet.com uses micro services on CosmosDB and Azure Event Grid
• Skype as well
• CQRS

• Roadmap
  • Q2, had public preview
  • Q3, GA
  • Q4
• Scale with backplane with Redis etc
• Azure Function binding, in preview
• For GA, Free Tier and standard
• Auto Scaling... ??
• Scenarios
  • In app notifications
  • Live dashboard
  • Co authoring
• It's realtime messaging
• An Azure resource
• dotnet build; dotnet run;
• Azure SignalR Service on asp.net Core SignalR
• Hub base class
• First negotiates. If configured to use Azure SignalR service, it returns negotiation points on Azure.

• Rest API
  • Gyro tracker demo
• Azure Function binding for SignalR Service

• 12 Factor

• Mongoose SDK Node.js sdk
• Develop locally with MongoDB
• Switch to CosmosDB in the cloud
  • Choose MongoDB api for CosmosDB
• Make app cloud ready

• dotenv package ??
• VS Code, Ctrl+shift B, run commands

• LiveTiles Bots
• Scenarios
  • Personal Assistant
  • Team Assistant
  • Enterprise Assistant
  • External Assistant
• What's a typical development cycle?
• Site application as a service as a bot!!
• QnA Maker
• LiveTiles Bot designer
• AIconics ?
• EthicalOS.org

• git push origin master:demo
  • If you forget to create a branch before pushing code remotely
• Branch policies
  • Require PR
  • Item linking
• Build Validation policy
  • Build must pass before merge into branch/master
• Checkout pull requests locally ??
• Continuous Code Quality with Automation
  • Static code analysis
    • sonarqube/sonarcloud.io
    • whitesource, security scanning
  • Drop these into pipeline
• Git Large File Support
  • Git LFS
  • Instead of shallow cloning
  • BFG Repo Cleaner ??
• Always use the latest version of Git
  • git update
• Trunk-based development
  • GitHub flow is different that Release flow branching structure
  • master is never in production?? Why?
  • Cherry pick a Pull request into a new branch

• Extensions
  • Slack, SonarCloud
  • Over 500 extensions
• Integration with GitHub
• Start with GitHub market place
  • Go to continuous integration
  • Setup Azure pipelines
  • Will land in Azure DevOps
  • Example: GitHub/microsoft/vscode
    • Has badges for Azure Pipelines
• CredScan ?, A step that check for credentials as part of a commit

• Azure Relay
  • Connect Systems
• Event Hubs
• Event Grid
• Messages
  • Fact
    • Time Series, logs, streams
      • Event Hubs for streaming data
    • Discrete
      • Not time series i.e. no order to the messages
      • Event Grid, event based programming
  • Command
    • Service Bus, Enterprise messaging
• Service Bus Explorer
• System to system messaging
  • Contrast this with
    • Notification hub
    • SignalR hub

• Labeling process
• Hyperparameters??
• scikit learn algorithm cheat sheet
• Automated machine learning
  • To give you a model faster
  • Submitting experiments to Automated ML
• Jupyter notebook
• Cross validation ??
• KNeighbors ?
• Set metrics, e.g. accuracy, AUC etc
• Confusion matrix, TP TN etc
• Configure Automated ML by setting metrics

Benefits of Automated ML

• Model creation time, from days to hours
• Robust benchmarking process for ML projects
• Enable domain experts to leverage ML
• Is an excellent starting point for selection of algorithm
• Currently: Classification/binary, and regression

Power BI

• Automated ML is baked into Power BI

Hyperparameter tuning

• Adjustable parameters to given model training
• Hyperparameter configuration
• Find the configuration that result in best performance
• Best accuracy
• Number of layers in deep learning
  • This is in addition to the training datasets
• Automated hyperparameter tuning in Azure ML
• Early terminations of poorly running models
• Bandit policy ??
• This capability is available for free, for now!

• CDS, Common Data Services
• An Azure based service
• PowerApps portal
  • Define custom entities
  • Relationships, Lookups
  • Optionsets eg enums
  • Calculated Fields, computed columns
  • Roll-up Fields, aggregates
• Microsoft Flows?

https://docs.microsoft.com/en-us/powerapps/maker/common-data-service/data-platform-intro

Step into specific, nested methods

• Tracepoints, Actions, pseudo debugger variables $Function
• Inspection, nse (no side effects), i++ in the watch window won't affect the variable in the program flow itself.
  • ac, always call
• $ReturnValue1 2 3
• DebuggerDisplay attribute - will be used as an attribute on the class itself
  • nq, no quotes
  • h, hexidecimal
• Make object Id, for vars that are added to a collection
• Performance Tip
  • Intellitrace
    • Historical Debugging
  • Snapshot Debugger
    • Debugging in production???
    • Remote debugging
  • Managed heap
    • Diagnostics window, memory usage tab, take snapshot
  • CPU usage, diagnostics window
• Format Specifiers

• GitHub Labeler
• GitHub Classification issuer
• Azure Stream Analytics, Anomaly detection
• 0.5 is available
• Features, Labels. Supervised machine learning
• Model consumption
• Transformer, takes the features and numerizing them
• Microsoft.ML nuget package
• Estimator is the thing to use to build your model
• Steps
  • Create environment
  • Read data
  • Create estimator and create model
  • Build and train
  • Evaluate the model
  • Make prediction function ?
• That's a classifier. Labeler example. Classification, however not binary.
  • Multi Classification
  • SDCA ?, A trainer.
• Featurized text
• Demo: movie recommendation
  • Population average, based on score
  • Content based filtering, taking into account user preference
  • Collaborative filtering
    • Factorization Machines ?
    • Movie Lens Dataset
• Deep Learning
  • Tensorflow
  • Image classification
  • Download model
  • Microsoft.MLTensorflow nuget package
  • Microsoft.ML.ImageAnalytics
  • Style Transfer ?
  • ONNX ?
    • Emotion detection
  • MS has started integrated deep learning into ML.Net, with TensorFlow

• Hyperscale
• Managed instances
  • Use your own licences
  • Linked Servers
• Security
  • Enterprise Grade Security
  • Always Encrypted, Encryption in use
  • Encryption in flight
• Intelligence
  • SQL graph
    • Merge DML
    • View or Derived table in match query
    • Edge Constraints
    • Node tables and Edge tables
  • Intelligent QP
    • Continuously monitoring query performance
    • Spill to disk for queries
    • Spill to TempDb
    • Approximate QP, aggregate queries with data science and analysis and modeling. Not good financial data.
    • Compatibility Level, 150
    • Memory Grant Sizer?
  • Eg Fraud Rings
    • SQL
    • Power BI
• Reliability
  • Accelerated Database Recovery (ADR)
    • On/Off
  • Parallel PFS updates
  • Resumable online index create, so it won't impact production users
  • Zone Redundancy, greater fault resiliency with replicas
  • Read scale-out support

• Programming oriented application architecture vs AI oriented app architecture
• The model is another input into the compute section of an application
• Enabling Natural User Interfaces (NUI)
• Ink Intelligence in PowerPoint and Word
• ML models for learning
  • Handwriting recognition
  • Ink classification
  • Layout analysis
  • Shape recognition
• Search in Ink

• Data Ingestion
• Training
• Operationalization

• Project Ink Analysis, Private Preview

• Why create an AI oriented application
  • Engage employees
  • Engage customers
  • Optimize operations
  • Transform products

• Azure
  • Pretrained models
  • Frameworks
  • Cora: sales assistant
    • Quite settings
    • Lead scoring
    • Churn prediction
• Cora, sales bot
  • CRM :-O
• AI in Patient Flow, staffing and management
• Health care ML application
• LOS, length of stay, prediction

• Azure Blob Storage
• Application Insights (AI), APM
  • Azure Monitor
  • Live Steam Metrics
  • Performance
• Azure Storage Explorer
• Azure Functions
  • Only pay for the compute time of the function
  • Based on triggers, hooks
  • Can you have multiple functions in a single project that has multiple triggers?
  • Serverless
  • Event driven execution
  • Not used for chatty services
• Key Vault