General FAQs
- What are the SLAs for Support?
- What Google Cloud regions are supported?
- What SLA is available for resizing and restarting Virtual Machines?
- What is the maximum Virtual Machine size?
- What is the maximum bandwidth available between virtual machines on the same host vs different hosts?
- What is the maximum Core and RAM that can be allocated to an affinity group?
- Is the customer notified when a Virtual Machine will be moved to another Power Host?
- When Soft Pinning is configured, how long will the Virtual Machine be removed from the original Host?
- Can you provide a sample of a scheduled maintenance notification?
- What security and compliance certifications are in place?
- What is the latency to IBM Power for Google Cloud from my datacenter?
- Why does the hostname revert after being changed?
- Volume Types
What are the SLAs for Support?
Support for IBM Power for Google Cloud is delivered via the Google Collaborative Support model.
An IBM Power for Google Cloud customer must have Google Premium or Enhanced support contacts for Support SLAs. The Customer will contact Google for support and SLAs will align to the Support contact. For more information, see Google Cloud Support.
Additional details about Google Support can be found in the Google Technical Support Services Guidelines.
What Google Cloud regions are supported?
IBM Power for Google Cloud is available in the following regions.
- us-east4 (N. Virginia)
- us-central1 (Iowa)
- europe-west3 (Frankfurt)
- europe-west4 (Netherlands)
- northamerica-northeast1 (Montréal)
- northamerica-northeast2 (Toronto)
We are continously evaluating expansion to new regions based on market demand. Contact power4gcp@googlegroups.com if your workload requires a region that is not listed.
What SLA is available for resizing and restarting Virtual Machines?
IBM Power for Google Cloud provides a self-service interface for interacting with IBM i, RHEL Linux and AIX workloads. The customer can resize and restart virtual machines using the API, Web Console or CLI.
Provisioning SLO for new virtual machines is 20 minutes until the system is ready for access via ssh.
What is the maximum Virtual Machine size?
Default core limits are identified by region in the table below. Some limits can be increased, contact sales using the Marketplace Page or open a support request if you require increased limits.
| Region | Max Core Count Per VM | Max RAM Per VM |
|---|---|---|
| us-east4 | 15 (Power 9) 21 (Power 10) | 400GB (Power 9) 1024GB (Power 10) |
| us-central1 | 15 (Power 9) | 400GB (Power 9) |
| europe-west3 | 15 (Power 9) 21 (Power 10) | 400GB (Power 9) 1024GB (Power 10) |
| europe-west4 | 21 (Power 10) | 976GB (Power 10) |
| northamerica-northeast1 | 21 (Power 10) | 1024GB (Power10) |
| northamerica-northeast2 | 21 (Power 10) | 1024GB (Power10) |
What is the maximum bandwidth available between virtual machines on the same host vs different hosts?
The maximum bandwidth between two Power Virtual Machines on two different hosts is up to 10Gb per second. However, that can be affected by the VM sizing, OS configuration and application. Maximum bandwidth between two Power Virtual Machines on the same host is expected to be greater. We recommend deploying workloads and validating performance if reaching a platform performance maximum is critical.
What is the maximum Core and RAM that can be allocated to an affinity group?
The total resources allocated to an affinity group cannot exceed the maximum core and RAM for a single Virtual Machine.
Is the customer notified when a Virtual Machine will be moved to another Power Host?
Yes. During onboarding the customer provides contacts for maintenance notifications. Converge will notify the customer contact via e-mail for scheduled or emergency maintenance.
- For Scheduled Maintenance, a notification will be sent to the customer two weeks prior to scheduled maintenance.
- For Emergency Maintenance, a notification will be sent when the Virtual Machine is being moved. Emergency maintenance is only performed when Power Host health is severely impacting Virtual Machine availability.
What do soft and hard pinning mean?
By default VMs are not pinned when created. You can optionally specify hard or soft pinning when creating a VM. Virtual machines with Pin Policy set to “soft” will be live migrated back to the original Host System once the maintenance is complete. Virtual machines with Pin Policy set to “hard” will be powered down during maintenance.
What are the affinity and anti-affinity groups?
The customer can create groups of VMs that should live on the same host or should not live on the same host.
When Soft Pinning is configured, how long will the Virtual Machine be removed from the original Host?
Up to 24 hours. If additional time is expected, a notification will be sent prior to maintenance.
Can you provide a sample of a scheduled maintenance notification?
A sample notification is shown below.
One or more IBM Power for Google Cloud Virtual Machines are running on a host system that requires maintenance. The following virtual machines will be live migrated to a new host system using Live Partition Mobility (LPM).
Impacted Virtual Machine by Name ip4g-vm-1 ip4g-vm-2
Virtual machines with Pin Policy set to “soft” will be live migrated back to the original Host System once the maintenance is complete. Virtual machines with Pin Policy set to “hard” will be powered down during maintenance.
During live migration, your virtual machine might experience a decrease in Volume, Compute, Memory, and Network performance for a short period of time.
This maintenance is required. If you would like to migrate your virtual machine prior to the scheduled maintenance, please open a support ticket to live migrate your virtual machines preemptively.
Required action: No action is required from you. Converge will monitor and ensure restoration of compute, block storage, and networking services.
Recommended actions: Ensure your team is aware of the maintenance and is available to validate any critical applications during and after the maintenance process is complete. Take additional care if your workloads are sensitive to brief loss of performance or connectivity.
Our intent is to strengthen our infrastructure and deliver you the best experience possible. If you have questions or concerns about this scheduled maintenance or If you would like to Live Migrate your virtual machine prior to the scheduled maintenance, please contact Google Customer Care and reference maintenance event XXXX.
Sincerely,
IBM Power for Google Cloud Operations
What security and compliance certifications are in place?
IBM Power for Google Cloud undergoes independent verification of security, privacy, and compliance controls to demonstrate compliance.
Audit reports are available at request via contractscompliance@convergetp.com.
What is the latency to IBM Power for Google Cloud from my datacenter?
Some services are recommended for deploying workloads that require low latency communication to applications running in IP4G. Converge recommends deploying on services such as
- Google Compute Engine
- Google Kubernetes Engine
Further, for best results, deploy in the same region. A heterogeneous workload can have latency as low as 1ms between IBM Power for Google Cloud and services like Google Compute engine.
In some cases Google Cloud services are not a good fit for a workload. Instead, a hybrid cloud solution is required. However, low latency to IBM Power for Google Cloud is still important. To build a low latency hybrid cloud solution, the customer must select an appropriate data center to host the non-Google Cloud workload.
Google Cloud provides a list of low latency colocation facilities. They are a subset of all colocation facilities that can provide dedicated Cloud Interconnects. Customers can choose to establish a direct relationship with these low latency colocation facilities. They can achieve less than 5ms of latency to a Google Cloud region. When selecting a colocation facility for low latency connectivity to IBM Power for Google Cloud, Converge recommends the facilities in the table below. They will allow for the lowest possible latency to a IBM Power for Google Cloud region.
| Google Cloud Region | Dedicated Cloud Interconnect Zone | Colocation Facility |
|---|---|---|
| us-east4 (Virginia) | iad-zone1-1 iad-zone2-1 | Equinix Ashburn (DC1-DC11) |
| us-central1 (Iowa) | cbf-zone1-575 cbf-zone2-575 | Nebraska data centers (1623 Farnam) |
| northamerica-northeast1 (Montreal) | yul-zone1-99002 yul-zone2-99002 | Cologix MTL10-H |
| northamerica-northeast2 (Toronto) | yyz-zone1-2206 yyz-zone2-2206 | Equinix Toronto (TR2) |
| europe-west3 (Germany) | fra-zone1-58 fra-zone2-58 | Interxion Frankfurt |
| europe-west4 (Netherlands) | grq-zone1-532 grq-zone2-532 | QTS Netherlands |
Why does the hostname revert after being changed?
The following applies to both AIX and Linux. To prevent cloud-init from resetting the hostname, edit the cloud-init configuration file. This is usually /opt/freeware/etc/cloud/cloud.cfg.
Set the parameter “preserve_hostname” to false:
preserve_hostname: false
Volume Types
In IP4G there are two volume types. Those volumes are defined by a name (HDD or SDD) and correspond to the following Characteristics:
| Region | Description |
|---|---|
| US East | SDD and HDD are different hardware types. Performance is dictacted by the corresponding hardware. |
| All other Regions | SSD and HDD are both IBM Flash Storage, with a QoS limit of 10 iops/GB for SSD, and 3 iops/GB for HDD. |
You can see an individual volumes iops limit using
pcloud compute volumes describe <volume name>