This meeting we'll host a discussion on Google Cloud Platform and Amazon Web Services to bring light to similarities and differences between platforms. If you have questions about how our platforms compare this is the meeting to attend!
6. DoIT International confidential │ Do not distribute
• Assumes you are familiar with AWS
• You’re curious about Google Cloud Platform
• It’s NOT an attempt to convince you to switch, just to give visibility
• It’s NOT feature-by-feature comparison
• Open discussion is welcome during the meetup
About This Meetup
7. DoIT International confidential │ Do not distribute
• Networking
• Compute
• Storage
• Big Data
• Management
Discussion Domains
10. DoIT International confidential │ Do not distribute
Networking (part I)
• Subnets
• Single global network spanning multiple regions and zones
• Load Balancers
• TCP/SSL or HTTP/S with transparent WebSockets support
• Single LB sending traffic to instances in multiple regions with geo-location
• Global IP (IPv4 or IPv6) with geo location built-in
• Real-Time Scaling (no pre-warming is necessary), up to 10M connections/sec
• Firewalls
• Global ingress or egress stateful rules protecting subnets in multiple regions
• Shared Networks
• Create single network spanning multiple projects, including in different regions
11. DoIT International confidential │ Do not distribute
Networking (part II)
• DNS
• No Latency-based routing & Geography-based routing
• Not an internal DNS (yet)
• Content Delivery Network
• Simple CDN for serving dynamic/static content, lacking many CloudFront features
• Direct/Carrier Peering
• Direct Peering is free of charge
• Network Performance (4-core Debian based instance)
• Same DC [7.3Gbit/sec], Same Zone [7.1Gbit/sec], Same Region [5.5Gbit/sec], Inter-
Region [3.8Gbit/sec]
12. DoIT International confidential │ Do not distribute
Compute (Overview)
• Compute Engine
• Standard, Highmem, Highcpu or Custom Machine types
• Skylake/Broadwell based cores [1-64] w/ up to 455GB memory
• Live Migration, Instance Groups, Autoscalers
• Up to 8x NVIDIA® Tesla® K80 devices per instance
• Other Compute Options
• Google Container Engine (a.k.a Managed Kubernetes)
• Google AppEngine (PaaS)
• Cloud Functions (in beta)
• Storage
• HDD or SSD - up to 64TB per disk (up to 128 disks)
• LocalSSD - up to 3TB (8x 375GB devices)
• No IOPS charges
13. DoIT International confidential │ Do not distribute
Compute (Storage)
Maximum sustained IOPS
HDD SSD Local SSD (SCSI) Local SSD (SCSI)
Read IOPS per GB 0.75 30 266.7 453.3
Write IOPS per GB 1.5 30 186.7 240
Read IOPS per instance 3,000 40,000 400,000 680,000
Write IOPS per instance 15,000 30,000 280,000 360,000
Maximum sustained throughput (MB/s)
Read throughput per GB 0.12 0.48 1.04 1.77
Write throughput per GB 0.12 0.48 0.73 0.94
Read throughput per instance 180 800 1,560 2,650
Write throughput per instance 120 400 1,090 1,400
14. DoIT International confidential │ Do not distribute
Compute (Cost Optimization)
• Per Minute Pricing (with 10 minutes minimum)
• Sustained Discounts (up to 30%)
• Committed Use Discounts
• (1 or 3 years commitments on #cores / #memory, 40% / 57% off pricelist)
• Preemptible Instances (constant 80% discount, <=24h)
• Custom Machine Types (solves the t-shirt problem)
15. DoIT International confidential │ Do not distribute
Storage (comparison)
Feature Amazon S3 Cloud Storage
Unit of deployment Bucket Bucket
Deployment identifier Globally unique key Globally unique key
File system emulation Limited Limited
Object metadata Yes Yes
Object versioning Yes Yes
Object lifecycle management Yes Yes
Update notifications Event notifications Object change notifications
Service classes Standard, Reduced Redundancy,
Infrequent Access, Amazon Glacier
Standard, Durable Reduced Availability,
Nearline, Coldline
Deployment locality Regional Regional and Multi-Regional
16. DoIT International confidential │ Do not distribute
Storage (highlights)
• Object Keys (avoiding hotspots)
• Conditional Updates
• Lifecycle Management
• No ‘bucket policies’, just ACLs
17. DoIT International confidential │ Do not distribute
Data Services (Google BigQuery)
• As-a-Service Analytical Database for petabyte scale
• No up-front cluster provisioning/sizing
• Built-in Ingestion with up to 100K rows/sec per table performance
• ANSI SQL 2011 with extensions
• DDL Statements - SELECT, UPDATE, DELETE
• Automatic caching of query results for up to 24 hours
• Integration with BI tools (Data Studio, Tableau, Looker, Periscope, re:Dash, etc)
18. DoIT International confidential │ Do not distribute
Data Services (Spanner)
• Fully Managed relational database
• Multinode write/read architecture
• Decoupled storage & compute resources
• Cross Datacenter & Cross Regional Automatic Replication
• ACID Transactions
• Strong Consistency
• Schema changes without locking tables
• Scales from 1 to 50,000 nodes & back without downtime
19. DoIT International confidential │ Do not distribute
Data Services (Cloud Dataflow)
• Managed Batch & Streaming ETL
• Reads from: Pub/Sub, BigQuery, Storage, Datastore, Bigtable & Custom Readers
• Writes to: same as readers + custom sinks
• Java/Python SDK
• Open sourced as “Apache Beam”
• Can run on top of Spark or Flink
20. DoIT International confidential │ Do not distribute
Data Services (Pub/Sub)
• A global service for real-time and reliable messaging and streaming data
• One-to-one, one-to-many, or many-to-many communication, with push or pull delivery.
• Multiple Regional Global Endpoints
• 7-days messages retention
• No automatic dead-letter queues
• Price by storage capacity (not # of messages sent)
21. DoIT International confidential │ Do not distribute
Data Services (CloudSQL)
• Managed MySQL or PostgreSQL (in beta) Database
• Read Replicas
• High Availability Replicas with automatic failover
• Limited support for MySQL flags
22. DoIT International confidential │ Do not distribute
Data Services (Bigtable & Datastore)
Google Bigtable
• A high performance NoSQL database service for large analytical and operational workloads
• Fully managed & decoupled storage & compute resources
• Each node provides 10K QPS & 120MB/sec of read/write throughput @ 5ms latency
• Scales to hundreds of nodes
• Transparent Cluster Resizes
• Open HBase API
Cloud Datastore
• + ACID transactions
• + Automatic Global Replication
• + SQL-like queries
23. DoIT International confidential │ Do not distribute
Data Services (Dataproc)
• Managed Spark and Hadoop service
• < 90 seconds to launch cluster of any size
• Ephemeral or Job-Oriented Clusters
• Local or Google Cloud Storage backed filesystem
• Seamless support for Preemptible instances or Custom Machine types
• Initialization Actions to install additional software (Presto, Jupyter, etc.)
• Affordable pricing: 1c per core / hour
24. DoIT International confidential │ Do not distribute
Data Services (...there is even more)
• Firebase
• Google Dataprep (in beta)
• Google Datalab
• Google Data Studio
• TPU Powered CloudML Engine (managed TensorFlow) with pre-trained models:
• NLP API
• Speech API
• Vision API
• Video Intelligence API
• Cloud Translation API
25. DoIT International confidential │ Do not distribute
Management & IAM
• Orgs -> Folders -> Projects
• User Accounts
• Service Accounts
• Preconfigured Roles
• No Custom Roles (yet)
• Programmatic Project Setup
• Deployment Manager
• Cloud Shell
26. DoIT International confidential │ Do not distribute
Monitoring
• Free and Paid Versions
• Plugins for many open source
solutions
• Integration with AWS
• Metric Store
• Visualization
27. DoIT International confidential │ Do not distribute
Security
• Identity Aware Proxy
• Cloud Data Loss Prevention API
• Key Management Service
• Security Scanner
28. DoIT International confidential │ Do not distributeDoIT International confidential │ Do not distribute
Thank you!
Vadim Solovey //CTO (vadim@doit-intl.com)