➥🔝 7737669865 🔝▻ Mathura Call-girls in Women Seeking Men 🔝Mathura🔝 Escorts...
Serverless and AI: Orit Nissan-Messing, Iguazio, Serverless NYC 2018
1. https://nuclio.io
Serverless and AI – SNYC 18
Orit Nissan-Messing
Co-Founder & Chief Architect @ Iguazio
oritn@iguazio.com
nuclio : github.com/nuclio
2. Event-Driven AI: Faster and Smarter Decisions
2
• Immediate insights: sub
seconds from events to data
driven actions
• Smarter decisions: up to
date AI models and fresh data
correlated with historical and
external data in real-time
• Cloud agility: continuous
development and deployment
3. 3
The Real-Time AI Pipeline
Transform
Real-Time DB and Model/File store
Real-time Triggers
and Actions
Enrich
(contextualize)
Infer
(predict)
Real-Time
Sources
Additional Context:
• Historical
• Operational
• Environmental
Act
ML
Model
Lower Time to Action, Higher Throughput
4. Serverless:
Easy-to-deploy functions, no scripts, Docker files, build, etc.
Auto-scaling and event triggers are built-in
Use any language
What’s missing?
Slow performance, lack of concurrency
Limited number of event sources (mostly HTTP, some Kafka)
Hard to manage complex functions and dependencies
Using Serverless for the AI Pipeline
5. Nuclio: Open Source Real-time Serverless
• Focus on your app, automate maintenance
• Fast
• High concurrency
• Low latency
• Open
• Runs Everywhere
• Cloud , On-Prem, Hybrid, Laptop
6. 6
Non-blocking, parallel
Zero copy, buffer reuse
Up to 400K
events/sec/process
Nuclio: Taking Serverless to The Next Level
Function
Workers
Event
Listeners
Serverless for compute and data intensive tasks
100x faster than AWS Lambda (really) !
Extreme Performance
Shard 1
Workers
Workers
Shard 2
Shard 3
Shard 4 Workers
Stream and Batch
DB, MQ, File
Functions
Auto-rebalance, checkpoints
Any source: Kafka, NATS, Kinesis,
MQTT, event-hub, Iguazio, pub/sub,
RabbitMQ, Cron, cloud-events
Data bindings
Shared volumes
Context cache
Statefullness
Nuclio processor
7. 7
Great Match: Serverless + Kubernetes
DBaaS
Object (S3)
• Cloud independent APIs, on-prem, edge
• Auto-scaling + abstract infrastructure
• Fastest evolving features and community
• One platform for serverless and services
• Abstraction, no Docker files or YAMLs
• Automated dev and ops flow
• Everything is a Kubernetes resource
• “Serverless” data services
• Scalable
• Location agnostic
12. Demo: Voice Driven Real-Time Analytics
NoSQL APIStream API Object API File API
Security
Queries &
Functions
Unified Data Data Lifecycle
SMART HOME
DEVICE
Voice
Query
SQL API Machine
Learning
Update
Locations
iguazio
Data
Platform
GOOGLE
DIALOGFLO
W
GOOGLE MAP
SERVICE
WEB UI (REACT)
SQL Query
NoSQL
Queries
ELASTIC
APPLICATION
MICRO-SERVICES
(OVER KUBERNETES)
12
13. 13
Open Source Enables Portability and Customization
Same APIs, Same User Experience, Anywhere
With native integration into each cloud platform
Laptop
Cloud
On-Prem or Edge
15. 16
Write code + local testing
Build code and Docker image
CI/CD Pipeline
Add logging and monitoring
Harden security
Provision servers + OS
Handle data/event feed
Handle failures/auto-scaling
Handle rolling upgrades
Configuration management
Write code + local testing
Provide spec, push deploy
Traditional Dev & Ops model “Serverless” Development Model
Serverless, Eliminating 80% of The Work
1. Automated by the
serverless platform
2. Pay for what you use
80%
16. Nuclio Open-Source, Real-Time Serverless
Local /
shared image
repo
Pluggable
Data Services
Pluggable
Triggers*
Platform
Controller
Function
…
Event
Listeners
Data
Bindings
Platform Abstraction Layer
Packaged functions, auto-scaling
• Up to 400,000 Events/sec per process, under 0.1 ms latency
• Any trigger
• Any Platform
18