Google Cloud Platform Deployment Guide

DDN Infinia is a distributed data platform designed for extreme performance and scalability. DDN Infinia, running on Google Cloud Platform’s (GCP) compute engine instances, can handle petabytes of data and billions of objects needed for demanding applications like artificial intelligence, deep learning, and high-performance computing (HPC).

DDN Infinia delivers the following capabilities:

  • Elastic, distributed data cluster
  • S3-compatible object store
  • High data durability
  • Multi-tenancy with role-based access control

Google Cloud users can deploy DDN Infinia from the Google Cloud Marketplace using standard GCP tools.

Audience

The content in this section is intended for users and system administrators who are responsible for deploying, managing, and maintaining DDN Infinia software on Google Cloud Platform (GCP). The level of content presented assumes that users have a strong understanding of Google Cloud Platform and have working knowledge about compute engine, storage, network configuration, and system administration within a Linux environment. The user must also be familiar with the administration of DDN Infinia.

Prerequisites

The following are the prerequisites for deploying and configuring DDN Infinia on Google Cloud Platform:

  • GCP Account: Ensure that you have a valid GCP account with the necessary permissions.

  • GCP Resources: Verify that the following GCP APIs are enabled:

    • Compute engine
    • Cloud deployment manager
  • GCP IAM Roles: To execute the deployment, you must have a service account with the following project roles:

    • roles/compute.admin
    • roles/iam.serviceAccountUser

    If you are using a shared VPC, the following role is enabled on the shared VPC host project.

    • roles/compute.networkAdmin
  • DDN Infinia Activation Token: A valid activation token to configure the DDN Infinia cluster.

DDN Infinia Solution Architecture

The following diagram illustrates the architecture of the DDN Infinia Data platform on Google Cloud.

infinia_sol_arch

Supported Instances

The following are the specifications of the compute engine instance that is supported with DDN Infinia 2.3.0:

Instance typen2d-standard-32
vCPUs32 vCPUs
Memory128 GiB
Network Bandwidth32 Gbps
Boot Disk256 GB (SSD)
Local NVMe disks24 NVMe local SSD drives with 9 TB of usable capacity

Before You Start Deployment of DDN Infinia

Before launching the automated deployment of DDN Infinia from the Google Cloud Marketplace, verify that you have the following in place:

  • DDN Infinia Activation Token: Contact DDN support (email: gcp@ddn.com) to obtain a valid activation token for DDN Infinia.

  • DDN Infinia Cluster Sizing: Determine the desired capacity and performance to be provisioned for the DDN Infinia cluster. The DDN Infinia cluster uses a minimum of 7 VM instances and scales to over 100 PB, depending on VM instance availability and GCP quota assigned to the project in which DDN Infinia cluster will be deployed.

    • The following are the capacity and performance metrics per VM instance:

      • Capacity per VM Instance: 9 TB

      • Network throughput per VM Instance: 4 GB/s

      • Minimum Cluster Size: 7 VM Instances

        • Starter Capacity: 9 TB x 7 = 63 TB
        • Starter Throughput: 4 GB/s x 7 = 28 GB/s
      • Minimum unit of cluster expansion: 1 VM Instance implying:

        • Min Unit of capacity expansion: 9 TB
        • Min Unit of network throughput expansion: 4 GB/s
    • Based on desired storage capacity, the DDN Infinia cluster automatically provisions the required number of VM instances (subject to availability of GCP quota in the project). The following are a few examples:

      • 63 TB Deployment: 7 VMs, 28 GB/s total throughput.
      • 108 TB Deployment: 12 VMs, 48 GB/s total throughput.
      • 192 TB Deployment: 24 VMs, 96 GB/s total throughput.
      • 1024 TB Deployment: 114 VMs, 456 GB/s total throughput.

User Inputs Required During Deployment

The following details are required during deployment:

  • Desired usable capacity in the DDN Infinia cluster (as explained earlier).
  • Enable/Disable cloud login and monitoring as desired.
  • Network configuration details (for example, subnets, external IPs, and firewall settings).

Deploying and Configuring DDN Infinia on GCP

Before starting to deploy DDN Infinia on Google Cloud Platform, ensure the following prerequisites are fulfilled:

  • Ensure you have a Google Cloud account. If you don’t have an account, follow instructions provided by Google to create an account.
  • Ensure you have a valid DDN Infinia activation token that is required at the time of configuration of the DDN Infinia cluster.

For deployment of DDN Infinia on GCP, currently the maximum supported length for the instance name is 54 characters.

Procedure

To deploy DDN Infinia on GCP Marketplace, complete the following steps:

1

Log in to your Google Cloud account.

3

In the search box, type DDN Infinia for Google Cloud and press Enter.

search_infinia

4

Click DDN Infinia for Google Cloud to open the application.

5

Click LAUNCH to configure the deployment. launch_button

6

In the TERRAFORM tab, provide the details as shown in following table to create a DDN Infinia environment on Google Cloud Platform:

deployment_page

FieldDescription
Deployment nameEnter a unique name to identify your deployment.

Note that only lowercase letters, numbers, or dash are allowed in deployment name. The deployment name cannot end with a dash.
Deployment Service AccountService Account Options:
• Use an existing service account if available.
• If required, create a new service account for the deployment.

Required IAM Roles:
• Grant roles needed for deployment, such as service account user, compute admin, and storage admin.

Service Account Email Format: Follow the format: example-service-account@projectid.iam.gserviceaccount.com

HTTPS Enabled:
• Ensure HTTPS communication is enabled for secure connections.
ZoneChoose a zone same as S3 clients for minimum latency and better throughput.
Desired CapacityEnter the required storage capacity based on your deployment needs. Default capacity is 63 TB. Note that by design the capacity is set in multiples of 9 TB. If you enter 65 TB, DDN Infinia will roundoff to 72 TB capacity.
Number of Clients (Optional)Specify the number of clients that will be deployed with this deployment.
LicensingEnter a valid DDN Infinia activation token.
Networking
Network interfaces
Default Network InterfaceSpecify the default network interface for communication.
Add a Network InterfaceNot Applicable.
Firewall
Allow HTTPS traffic from the InternetSelect to enable HTTPS access for secure traffic (Needed to access to UI). If enabled, provide the source IP ranges for HTTPS traffic to control access to your deployment.
Google Cloud Operations
Enable Cloud LoggingSelect to enable detailed logs.
Enable Cloud MonitoringSelect to enable performance and resource monitoring.
7

Click DEPLOY.

8

Wait for a couple of minutes for DDN Infinia for Google Cloud to deploy on GCP VMs. Note the First Instance Nat IP and Admin Password from the Outputs section in the DETAILS tab, to be used for cluster setup in next step.

verify_deployment_details

9

Use one of the following methods to setup cluster and login DDN Infinia:

To setup cluster and login using UI:

  1. Go to https://{First Instance Nat IP} from the deployment outputs
  2. Login with username: realm_admin, password: {Admin Password from deployment outputs}
  3. Follow the instructions in Initial Cluster Setup by using UI to setup cluster by using UI.

Or

To setup cluster and login using CLI:

  1. SSH to {First Instance Nat IP} from the deployment outputs
  2. Run redcli user login realm_admin -p <Admin Password from deployment outputs>
  3. Follow the instructions in Initial Cluster Setup by using CLI to setup cluster by using CLI.

Note that while creating DDN Infinia cluster using redcli cluster create command, you must pass -S flag with value false. This option disables subnet enforcement that allows different subnets in the GCP instance and also allows clients to connect from different instance, which is mandatory for deployment of DDN Infinia on GCP.

Example

This example shows how to create a cluster with subnet enforcement disabled.

$redcli cluster create -S false

Post-deployment Configuration Details

Upon successful deployment of DDN Infinia on Google Cloud Platform, the following configuration parameters are provided as output:

  • admin_password: Password for admin.
  • instance_machine_types: Machine types for all compute instances.
  • instance_nat_ips: External IPs of all compute instances.
  • instance_network: Network of the first compute instance.
  • instance_self_links: Self-links for all compute instances.
  • instance_zones: Zones for all compute instances.
  • realm_entry_secret: Password for realm entry.
  • site_url: Site URL of the first instance.
  • total_capacity: Total capacity provisioned in TB.
  • total_throughput: Total throughput provisioned in GB/s.
  • vm_count: Total number of VMs provisioned.

After first login, change both the admin password and realm entry secret immediately to ensure security for your account.

For optimal performance, DDN Infinia uses local NVMe SSDs in VM instances on the Google Cloud Platform (GCP). Please beware of the following scenarios that could potentially degrade the performance or durability of the DDN Infinia cluster:

  • Any reconfiguration of the DDN Infinia cluster to use other types of storage on GCP will render the cluster unavailable.
  • If a host error occurs and compute engine cannot reconnect the instance to the local SSD disk within the recovery timeout (default is 1 hour), data may not be recoverable. If reconnection fails, the instance is restarted with a blank local SSD disk.
  • Shutting down the guest operating system and forcing the instance to stop will result in data loss on local SSD disks.
  • Configuring the instance to stop on host maintenance events or experiencing such an event will result in a data loss.
  • Disabling the project billing will cause the instance to stop, leading to loss of data on local SSD disk.

Ensure proper configuration and regular data backup to prevent data loss in such scenarios.