Overview of Compute Engine (Aug 2015)

Overview of Compute Engine (Aug 2015)

01dc8e954957a10b428aa60b28c89d52?s=128

Ian Lewis

August 25, 2015
Tweet

Transcript

  1. Google Cloud Platform Overview of Compute Engine Ian Lewis Developer

    Advocate
  2. Ian Lewis Developer Advocate Google Cloud Platform google.com/+IanLewis-hoge @IanMLewis

  3. Google Cloud Platform Images by Connie Zhou

  4. For the past 15 years, Google has been building out

    the world’s fastest, most powerful, highest quality cloud infrastructure on the planet.
  5. • Pioneering data center efficiency • Financed over 250 Megawatts

    of new wind power • First data centers to receive ISO 14001 certification • 100% carbon neutral Our Commitment to Environmental Responsibility
  6. Networking Management Networking Big Data Storage Mobile Developer Tools Compute

    Engine App Engine Container Engine Google Cloud Platform Compute Compute
  7. Networking Management Networking Big Data Storage Mobile Developer Tools Compute

    Engine App Engine Container Engine Google Cloud Platform Compute Compute
  8. Compute Resources - US, Europe and Asia zones - Fast

    SDN-Based virtual networking Consistently Fast - Fast VM Provisioning - Consistent Performance Cost Effective - Sub-Hour Billing - No IOPS charges for Block Storage - VMs not required for Load balancing Google Compute Engine
  9. Cores CPU to Memory ratio More memory 1 2 4

    8 More CPU (Shared) 16 Instances to suit diverse workloads 32 Compute Engine VMs: • Debian, CentOS, SUSE, RHEL, Ubuntu, Windows (Beta • shared core or 1 - 32 cores • Up to 208 GB of RAM
  10. 100% 0% 20% 40% 60% 80% Sustained Use Sustained-use discounts

    Net Price Per Hour Sustained-use discounts
  11. us-central1-a scheduled maintenance event Live Migration Live Migration

  12. Live Migration Live Migration

  13. • Per GB pricing; no IO charges • Consistent Performance

    • Encryption • Live Migration Persistent Disk Persistent Disk
  14. • Per GB pricing; no IO charges • Consistent Performance

    • Encryption • Live Migration Persistent Disk SSD PD Standard PD Local SSD Persistent Disk • Sub-ms latency • 1-4 x 375 GB partitions • Storage redundancy • Checksums • Snapshots • 1GB-10TB
  15. • Per GB pricing; no IO charges • Consistent Performance

    • Encryption • Live Migration Persistent Disk SSD PD Standard PD Local SSD Persistent Disk • Sub-ms latency • 1-4 x 375 GB partitions • Storage redundancy • Checksums • Snapshots • 1GB-10TB 200 30 80 Local SSD IOPS/GB (log scale) 0 Standard PD 15 6 SSD PD 2
  16. RO RW • Attach/Detach Data disks on the fly •

    Mount Read-Write on a single VM • Mount Read-Only on many VMs Persistent Disk Attach/Mount options
  17. Management Networking Compute Big Data Storage Mobile Developer Tools Load

    Balancing Google Cloud Platform Networking
  18. Load Balancing Load Balancing L3 (network) or L7 (HTTP/HTTPS) Load

    Balancing Single IP per forwarding rule No prewarming needed; goes from 0 to full- throttle in seconds Detects unhealthy virtual machines instances Balance loads across regions Routes traffic to the closest virtual machine Supports content-based routing
  19. None
  20. Load Balancing HTTP/HTTPS Load Balancing (Beta) Cross Region Load Balancing

    Content Based Load Balancing
  21. Network Load Balancing Load Balancing Incoming Requests Forwarding Rule X

    Target Pool Health Check
  22. • TCP and UDP (plus VPN protocols) • IP addresses

    • public: static and ephemeral • internal: ephemeral with automatic DNS • Network groups & firewall rules • rules applied to instances via tags • Gateways, routing and VPNs Networking Internet Networking
  23. Managed Infrastructure Autoscaling groups of homogeneous VMs

  24. Managed Instance Group • Grouping of identical VM instances Managed

    Instance Group VM VM VM Managed Instance Groups
  25. create/ destroy VMs Instance Group Manager Managed Instance Group VM

    VM VM Managed Instance Group • Grouping of identical VM instances • Provisioned and monitored by Instance Group Manager Instance Group Manager
  26. create/ destroy VMs Instance Group Manager Instance Template Instance Template

    Managed Instance Group VM VM VM Managed Instance Group • Grouping of identical VM instances • Provisioned and monitored by Instance Group Manager • Instance Templates separate configuration from provisioning Instance Templates
  27. Managed Instance Group destroy/ recreate VMs Instance Group Updater Rolling

    Update VM VM VM New Template Managed Instance Group • Grouping of identical VM instances • Provisioned and monitored by Instance Group Manager • Instance Templates separate configuration from provisioning • Rolling Updates applied by instance Group Updater Instance Group Updater
  28. Managed Instance Group create/ destroy VMs Instance Group Manager actuator

    monitor Autoscaler Cloud Monitoring utilisation VM VM VM Pub/Sub Topic Compute Engine Autoscaler
  29. Just Like Regular VMs • Same configurations, OSes, performance, tools,

    etc. More Cost Effective • Flat pricing: < 50% regular VM rates Reasonable, Predictable Limitations • 24h uptime limit • Chance of early preemption by system (~10-20%) ◦ No preemptions due to other customers’ Preemptible VM requests • Pre-termination notice to allow for clean shutdown Preemptible VMs: What Batch
  30. Flexibility Gives Google Efficiency • Limited time “lease” and preemption

    improve fleet allocation efficiency • Google can pass on the savings to you: win-win! “Regular” VMs are Suboptimal for Batch Use Cases • Always-on, infinite duration nature is not needed for batch type jobs • 99.95% continuous uptimes are designed for “serving” needs, not batch • Embarrassingly parallel jobs are often fault-tolerant to loss of worker nodes • Batch jobs might not hit best sustained use pricing thresholds Preemptible VMs: Why Batch
  31. • Log processing, Analytics, Big Data (Hadoop, MapReduce, etc.) •

    Rendering • Media transcoding • Financial modeling and simulation • Continuous integration - automated build and testing • Genomics Preemptible VMs: Typical Use Cases Batch
  32. cloud.google.com Images by Connie Zhou