Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Oxide Computer - IT Press Tour #56 June 2024

Oxide Computer - IT Press Tour #56 June 2024

The IT Press Tour

June 10, 2024

More Decks by The IT Press Tour

Other Decks in Technology

Transcript

  1. On-premises: A fractured ecosystem, all at odds Set up and

    integration times measured in months Product boundaries that hamper performance, impact reliability, and limit efficiency Incoherent, proprietary software propagates security vulnerabilities Hostile to developers, APIs as a second class citizen Lack of accountability when customers need it most
  2. “We were doing our own hardware because ... we were

    thinking much more about the data center as a computer, rather than a single box as the computer, and that really pushes you in a different direction.” URS HÖLZLE SVP ENGINEERING AT GOOGLE (2014)
  3. Servers as they should be Rack-level hardware, cloud software designed

    in System-wide integration drives 2x efficiency gains End-to-end design, open-source approach limits security vulnerabilities Rapid capacity planning and procurement tooling Comprehensive support through a single source Fully programmable with first-class API support
  4. The Cloud Computer HOLISTIC SECURITY HIGH EFFICIENCY POWER DESIGN INTEGRATED

    NETWORK SERVICES COMPUTE PROVISIONING SERVICES VIRTUAL BLOCK STORAGE SERVICE
  5. Oxide Virtual Private Cloud Virtualized networking lets you create, organize,

    and isolate project-specific networks across all your Oxide racks. Networking Observability all the way down Understand what end-to-end paths exist and how they relate to your operational requirements. Self-service networking Configure full networking capabilities quickly and easily – via API, CLI or console.
  6. On-demand virtual machines Provision and manage virtual machine instances. Flexibly

    allocate vCPUs and memory, and combine with an OS image. Compute Automate resource provisioning and management Metrics and monitoring Project-level resource limits
  7. Distributed block storage High-performance, NVMe-based persistent block storage service with

    configurable capacity and IOPS per volume. Redundancy assures high availability for business-critical workloads. Storage On-demand snapshots Oxide offers users instantaneous, point-in-time virtual disk snapshots to use for recovery and off-rack backup. Proactive data protection Powered by OpenZFS
  8. Developer experience The speed & simplicity of cloud abstraction Launch

    projects within minutes of powering on. Project-based resource provisioning & control Set and manage quotas on a per-project or per-tenant basis. Leverage existing tools & processes Manage with technologies you already know and use with our Terraform integration.
  9. “...70% of organizations that do not have a firmware upgrade

    plan in place will be breached due to a firmware vulnerability.” GARTNER
  10. Software Attestation Attestation and secret sharing provides protection against supply

    chain attacks, interdiction attacks, equipment theft, and firmware-style attacks Security Securing Customer Data A trust quorum ensures that customer data cannot be accessed until a majority of the devices in the cluster have proven themselves to be genuine. Auditibility Monitor your risk profile with clear visibility into the hardware and the firmware they are running. Quickly and easily patch vulnerabilities.
  11. Hyperscale Design Integrated DC bus bar, highly efficient fan control,

    and streamlined airflow significantly reduces power draw Efficiency System-Wide Optimization Power prioritization to ensure resources are focused on the applications that need them Newfound Observability Determine where you are spending your power budget across the entire system for both HW & SW
  12. Rack-Level Power Delivery With dynamic power management through our integrated

    Power Shelf Controller From 64 power supplies to a single bus bar Rack-level energy efficiency End-to-End Airflow Design Improved fan geometry, placement, and operational timing From 25% to <2% of total power draw Telemetry Understand and control power utilization Full system visibility made possible through hardware / software co-design
  13. Do more with less Shorter time to value Install to

    developer access in hours Improved resource efficiency Capacity, allocation, and usage insights Forecasting and automated alerting of utilization Lower operational overhead Autonomy through self-service environment for developers. Highly serviceable and easily supportable for operators
  14. Delivery To Developer From 100 days to 1 day No

    assembly required, no license management Cloud benefits realized Workload Performance From 72 hours to 6 hours End-to-end data transformation workflow 12x faster Deployment Speed From hours to seconds Self-service, fully programmable environment