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

Infrastructure at the Untethered Edge

Infrastructure at the Untethered Edge

Explores why building robust and secure data center infrastructure is hard and how the unique RackN infrastructure as code and context management in Digital Rebar solves this challenge.

Video https://youtu.be/IYw_PV_CR0E

Rob Hirschfeld

March 10, 2020
Tweet

More Decks by Rob Hirschfeld

Other Decks in Science

Transcript

  1. Self-Managed Infrastructure
    at the Untethered Edge
    RackN Digital Rebar v4
    Video: https://youtu.be/IYw_PV_CR0E

    View Slide

  2. “Stop managing
    infrastructure!”
    CSPs to Enterprise:

    View Slide

  3. “Stop managing
    infrastructure!”
    CSPs to Enterprise:

    View Slide

  4. Why is infrastructure
    so hard to manage?
    If infrastructure is critical to business...

    View Slide

  5. Integrated
    Vendor-Locked
    Tethered / Managed
    Best of Breed
    Commodity
    Do-It-Yourself
    or

    View Slide

  6. Integrated
    Vendor-Locked
    Tethered / Managed
    Best of Breed
    Commodity
    Do-It-Yourself
    or

    View Slide

  7. Integrated
    Vendor-Locked
    Tethered / Managed
    Best of Breed
    Commodity
    Do-It-Yourself
    and
    Infrastructure v2020
    We see the data center as an integrated, multi-vendor system instead
    of a bunch of stand alone components. RackN provides cloud-like
    automation and APIs without surrendering the control domains.

    View Slide

  8. Integrated
    Vendor-Locked
    Tethered / Managed
    Best of Breed
    Commodity
    Do-It-Yourself
    and
    Edge v2020
    We see the data center as an integrated, multi-vendor system instead
    of a bunch of stand alone components. RackN provides cloud-like
    automation and APIs without surrendering the control domains.
    plus Site-Autonomy

    View Slide

  9. Production Solutions
    DRP
    DRP
    DRP
    From Remote
    Edge Sites
    To Multinational
    Data Centers

    View Slide

  10. At Every Scale
    DRP
    DRP
    DRP
    Or Highly Available
    for 10,000 Servers
    Can runs from a
    Switch or RPi

    View Slide

  11. Edge
    Edge
    Data
    Center
    Edge
    Edge
    Data
    Center
    Edge
    Edge
    Edge
    Edge Edge
    Edge
    In practice, Enterprise and Edge have same management needs
    Cross Domain Control is not scale problem
    Security
    Zero Touch
    Application Mix
    Commodity H/W
    Enterprise = Edge

    View Slide

  12. Customer Managed Platform (No SaaS or MSP)
    Runs behind their firewall without any RackN supervision or connection.
    Dedicated
    Infrastructure
    Dedicated
    Infrastructure
    Edge
    Site
    Edge
    Site
    Edge
    Site
    Edge
    Site Edge
    Site
    Edge
    Site
    Shared
    Infrastructure

    View Slide

  13. Technical
    Deep Dive
    Critical Needs
    1. Easy to Use Shared Patterns
    2. Infrastructure as Code (IaC)
    3. Multi-Vendor
    4. Constantly Updated / Secure
    5. Integration to SoR
    6. Compliance & Visibility
    The “how” matters if we are
    going to dramatically improve
    infrastructure operations

    View Slide

  14. The Challenge: Many Control Domains
    Network Management
    Workload Network & DMZ
    Systems of Record
    Operator Network
    Infrastructure is not monolithic: management must span many systems.
    Corporate Network (Dev & Operator Clients)
    Protected Systems and Assets (Trusted)
    Infrastructure Administration (Trusted)
    Internet Facing Applications (Untrusted)

    View Slide

  15. Public cloud flatten control domains
    Network Management
    Workload Network & DMZ
    Systems of Record
    Operator Network
    In Public Clouds, all of
    these control domains
    are managed by public
    APIs!
    Cloud
    API

    View Slide

  16. Control domains are required, but hard to manage
    Network Management
    Workload Network & DMZ
    Systems of Record
    Operator Network
    Instead of replacing control domains
    with a single API, we build workflow
    orchestration that coordinates
    activities between them.

    View Slide

  17. Starting with Bare Metal Automation
    DHCP
    Netboot (PXE, iPXE, …)
    Agent (Linux, Win, ESX, … )
    Network Switch / ToR
    REST API & UX
    DRP
    OOB / BMC (iLo, DRAC, Redfish, ...)

    View Slide

  18. Evolved to an Infrastructure Workflow Hub
    DHCP
    Netboot (PXE, iPXE, …)
    Agent (Linux, Win, ESX, … )
    Network Switch / ToR
    REST API & UX
    Systems of Record
    Configuration Mgmt DB
    Monitoring Systems
    DRP
    OOB / BMC (iLo, DRAC, Redfish, ...)
    CI/CD Pipeline

    View Slide

  19. Because Integrated Workflow Is Critical
    DHCP
    Netboot (PXE, iPXE, …)
    Agent (Linux, Win, ESX, … )
    Network Switch / ToR
    REST API & UX
    Systems of Record
    Configuration Mgmt DB
    Monitoring Systems
    DRP
    OOB / BMC (iLo, DRAC, Redfish, ...)
    CI/CD Pipeline

    View Slide

  20. Workflow must span control domains
    DHCP
    Netboot (PXE, iPXE, …)
    Agent (Linux, Win, ESX, … )
    Network Switch / ToR
    REST API & UX
    Systems of Record
    Configuration Mgmt DB
    Monitoring Systems
    DRP
    OOB / BMC (iLo, DRAC, Redfish, ...)
    CI/CD Pipeline

    View Slide

  21. Provisioning Crosses Many Control Domains
    DHCP
    Unique RackN design smoothly integrates over many APIs
    OOB / BMC
    Netboot (PXE - TFTP)
    Netboot (iPXE - HTTP)
    Netboot (Kickstart - HTTPS)
    Agent (HTTPS) Inventory
    Power On
    Switch Config (API/CLI)
    VLAN
    Setup
    Next
    Server
    Stage 1
    Stage 2
    Discover
    O/S
    Firmware
    Reboot
    Next
    Server
    Stage 1
    Stage 2
    Target
    O/S
    Net
    Config
    App
    Config

    View Slide

  22. Typical Operations cross 10+ Domains
    Provision
    We work with, not replace, existing operational systems
    OOB / BMC
    Configure
    Secure
    DRP Orchestration
    Network
    Power On
    Report
    ESXi
    Network
    TLS
    VLAN
    Setup
    Cert Serv
    Join
    vCenter
    Enable
    Enroll
    Verify
    Status

    View Slide

  23. Infrastructure Automation Landscape
    Infrastructure as Code
    Configuration Management
    Metal Provisioning
    Multi-Cloud Managers
    API Orchestration
    Infrastructure Configuration
    Automation
    Chasm

    View Slide

  24. Digital Rebar Platform is the Bridge
    Infrastructure as Code
    Configuration Management
    Metal Provisioning
    Multi-Cloud Managers
    API Orchestration
    Infrastructure Configuration
    DRP

    View Slide

  25. Superior Bare Metal Provisioning
    Infrastructure as Code
    Configuration Management
    Metal Provisioning
    Multi-Cloud Managers
    DRP
    Replaces

    View Slide

  26. Configuration Management
    Differentiated Infrastructure as Code
    Infrastructure as Code
    Metal Provisioning
    Multi-Cloud Managers
    DRP
    Replaces
    Extend &
    Replace

    View Slide

  27. Configuration Management
    Required State Management
    Infrastructure as Code
    Metal Provisioning
    Multi-Cloud Managers
    DRP
    Replaces
    Extend &
    Replace
    Leverages

    View Slide

  28. Configuration Management
    Cloud-Like Intent-Driven APIs
    Infrastructure as Code
    Metal Provisioning
    Multi-Cloud Managers
    DRP
    Replaces
    Extend &
    Replace
    Aligns
    Leverages

    View Slide

  29. Configuration Management
    ...Avoiding the Ghosts of Operations’ Past
    Infrastructure as Code
    Metal Provisioning
    Multi-Cloud Managers
    DRP

    View Slide

  30. How does this matter for Edge?
    Edge needs multi-dimensional
    solutions
    I’m all alone!
    Where are my
    control domains?!

    View Slide

  31. Edge must maintain and share local state
    State
    Management

    View Slide

  32. Edge builds to end-states (not prescriptive configs)
    State
    Management
    Intent-Driven
    Workflow

    View Slide

  33. Edge must be reproducible & consistent
    State
    Management
    Infrastructure as
    Code
    Intent-Driven
    Workflow

    View Slide

  34. Edge must work
    State
    Management
    Infrastructure as
    Code
    Intent-Driven
    Workflow
    Multi-Vendor &
    Brownfields

    View Slide

  35. RackN combines all four dimensions
    IaC
    DRP
    Hardware
    Master
    Catalog
    IaC
    IaC
    O/S
    Platform
    Fully
    Integrated
    Clusters
    Rebuilds In Minutes
    Zero Touch

    View Slide

  36. Plus a Multi-Site “Single-Pane-of-Glass”
    DRP
    DRP
    DRP
    Global
    Control
    DRP
    Local
    Control
    DRP
    Regional
    Control
    Live Single
    Pane of Glass
    Distributed
    DC US-1
    Distributed
    DC EU-2
    Distributed
    DC AR-3

    View Slide

  37. Multi-Vendor vs Managed is an Artificial Dichotomy
    Don’t let vendors say
    “not my problem” with
    component silos!
    Don’t let CSPs
    determine the
    operating rules!

    View Slide

  38. Unlocking Innovation via
    Self-Determination and Integration
    Vendor Neutral Multi-Site
    Infrastructure Automation

    View Slide

  39. Control Domain Demos
    State
    Management
    Infrastructure as
    Code
    Intent-Driven
    Workflow
    Multi-Vendor &
    Brownfields

    View Slide

  40. Control Domain Demos
    State
    Management
    Infrastructure as
    Code
    Intent-Driven
    Workflow
    Multi-Vendor &
    Brownfields

    View Slide

  41. Bootstrapping Process for Edge Lab
    Worker Edge Lab DRP
    RPi #2+ Host
    Agent
    State
    Data
    Leader
    RPi #1
    Bootstrap Tasks:
    ISOs & DHCP
    Tasks:
    NetBoot
    Tasks:
    NetBoot
    Tasks:
    Discovery
    Machines:
    RPi 1+
    Tasks:
    Discovery

    View Slide

  42. Bootstrapping Process for Edge Lab
    Worker Edge Lab DRP
    RPi #2+ Host
    Agent
    State
    Data
    Leader
    RPi #1
    Bootstrap Tasks:
    ISOs & DHCP
    Tasks:
    NetBoot
    Tasks:
    NetBoot
    Tasks:
    Discovery
    Tasks:
    Election
    Cluster Profile:
    Leader
    Tasks:
    Election
    Machines:
    RPi 1+
    Tasks:
    Discovery

    View Slide

  43. Bootstrapping Process for Edge Lab
    Worker Edge Lab DRP
    RPi #2+ Host
    Agent
    State
    Data
    Leader
    RPi #1
    Bootstrap Tasks:
    ISOs & DHCP
    Tasks:
    NetBoot
    Tasks:
    NetBoot
    Tasks:
    Discovery
    Tasks:
    Election
    Tasks:
    Download
    Cluster Profile:
    Leader
    Tasks:
    Election
    Tasks:
    Watch for Credentials
    Machines:
    RPi 1+
    Tasks:
    Discovery
    Files:
    K3s Bins

    View Slide

  44. Bootstrapping Process for Edge Lab
    Worker Edge Lab DRP
    RPi #2+ Host
    Agent
    State
    Data
    Leader
    RPi #1
    Bootstrap Tasks:
    ISOs & DHCP
    Tasks:
    NetBoot
    Tasks:
    NetBoot
    Tasks:
    Discovery
    Tasks:
    Election
    Tasks:
    Download
    Tasks:
    Build K3s
    Cluster Profile:
    Leader
    Tasks:
    Election
    Tasks:
    Watch for Credentials
    Machines:
    RPi 1+
    Files:
    K3s Bins
    Tasks:
    Discovery

    View Slide

  45. Bootstrapping Process for Edge Lab
    Worker Edge Lab DRP
    RPi #2+ Host
    Agent
    State
    Data
    Leader
    RPi #1
    Bootstrap Tasks:
    ISOs & DHCP
    Tasks:
    NetBoot
    Tasks:
    NetBoot
    Tasks:
    Discovery
    Tasks:
    Election
    Tasks:
    Download
    Tasks:
    Build K3s
    Tasks:
    Gen Creds
    Cluster Profile:
    Leader
    Cluster Profile:
    K3s Creds
    Tasks:
    Election
    Tasks:
    Watch for Credentials
    Machines:
    RPi 1+
    Files:
    K3s Bins
    Tasks:
    Discovery

    View Slide

  46. Bootstrapping Process for Edge Lab
    Worker Edge Lab DRP
    RPi #2+ Host
    Agent
    State
    Data
    Leader
    RPi #1
    Bootstrap Tasks:
    ISOs & DHCP
    Tasks:
    NetBoot
    Tasks:
    NetBoot
    Tasks:
    Discovery
    Tasks:
    Election
    Tasks:
    Download
    Tasks:
    Build K3s
    Tasks:
    Join K3s
    Cluster Profile:
    Leader
    Cluster Profile:
    K3s Creds
    Tasks:
    Election
    Tasks:
    Watch for Credentials
    Machines:
    RPi 1+
    Tasks:
    Discovery
    Files:
    K3s Bins
    Tasks:
    Gen Creds

    View Slide

  47. Leveraging Ansible (Create Cloud Instance)
    Server DRP Endpoint
    Default
    Context
    Ansible
    Container
    Linode
    Cloud
    API
    User:
    Create
    Machine
    State
    Data

    View Slide

  48. Leveraging Ansible (Create Cloud Instance)
    Server DRP Endpoint
    Default
    Context
    Ansible
    Container
    Linode
    Cloud
    API
    User:
    Create
    Machine
    Tasks:
    SSH
    Key-Gen
    Machine:
    SSH Keys
    State
    Data

    View Slide

  49. Leveraging Ansible (Create Cloud Instance)
    Server DRP Endpoint
    State
    Data
    Default
    Context
    Ansible
    Container
    Linode
    Cloud
    API
    Cloud:
    Create
    Machine
    User:
    Create
    Machine
    Tasks:
    SSH
    Key-Gen
    Tasks:
    Create
    Instance
    Machine:
    SSH Keys

    View Slide

  50. Leveraging Ansible (Create Cloud Instance)
    Server DRP Endpoint
    State
    Data
    Default
    Context
    Ansible
    Container
    Linode
    Cloud
    API
    Cloud:
    Create
    Machine
    Tasks:
    SSH to
    Join-Up
    User:
    Create
    Machine
    Tasks:
    SSH
    Key-Gen
    Tasks:
    Create
    Instance
    Machine:
    SSH Keys
    Machine:
    IP Addr

    View Slide

  51. Leveraging Ansible (Create Cloud Instance)
    Server DRP Endpoint
    State
    Data
    Default
    Context
    Ansible
    Container
    Linode
    Cloud
    API
    Cloud:
    Create
    Machine
    Tasks:
    SSH to
    Join-Up
    User:
    Create
    Machine
    Tasks:
    SSH
    Key-Gen
    Tasks:
    Create
    Instance
    Tasks:
    Discover
    Machine:
    SSH Keys
    Machine:
    IP Addr
    Machine:
    Inventory

    View Slide

  52. Leveraging Ansible (Create Cloud Instance)
    Server DRP Endpoint
    State
    Data
    Default
    Context
    Ansible
    Container
    Linode
    Cloud
    API
    Cloud:
    Create
    Machine
    Tasks:
    SSH to
    Join-Up
    User:
    Create
    Machine
    Tasks:
    SSH
    Key-Gen
    Tasks:
    Create
    Instance
    Tasks:
    Discover
    Tasks:
    Install
    Agent
    Workflow:
    Ready
    Machine:
    SSH Keys
    Machine:
    IP Addr
    Machine:
    Inventory

    View Slide