Please consult the documentation for your virtualization platform and hardware for best practices and specific configuration recommendations.

vMotion is not supported for deployment of ICE Server due to a requirement for a static MAC address. Georedundant configurations require the same MAC address for functionality. Currently, the MAC address may change, terminating the connection.

Avoid deploying the ICE Server VM on virtualization hardware also hosting other I/O intensive VMs, for example, database servers, otherwise the installation will fail.

The VMs for DC1 and DC2 should not share the same virtualization hardware host nor be on the same subnet.

Specify the deployment size and verify the availability of minimum virtual machine requirements (more is always better):

Note: The browser-based installation wizard requests you to specify one of the deployment sizes listed above. If requirements are not met for the deployment size, the wizard will identify them. Once all requirements are met, the wizard proceeds with installation.

Verify the host server's storage space is sufficient.

Unless there is no other option, always use a Solid State Drive (SSD) to host the ICE Server VM. ICE Server performance and stability are heavily dependent on (minimal) disk I/O latency, so an SSD is essential for a successful deployment.

Disk storage for the VM should be created by thick provisioning with the highest shares allowed. Please see: VMware: Resource Allocation Shares.

Important: Required – If a traditional mechanical HDD is the only option, then that machine must be dedicated to running the ICE Server VM only. There can be no competition for disk access between the ICE Server VM and other VMs when using a mechanical hard drive.

All drives are tested during installation. Any drive failing the disk performance (speed) test cannot be used for installation. Traditional mechanical hard drives fail more frequently than solid state drives (SSD).

Memory must not be oversubscribed. As a starting point, assume an amount equal to 25% of total VM memory usage should be allocated for memory overhead. For example, if there are two VMs on a host, and each uses 32Gb of memory (64Gb in total), then allocating an additional 16Gb (at least) for overhead is prudent. Consult the VMware or your virtualization hardware vendor for guidance on appropriate memory overhead based on VM size and type.

If hyperthreading, then the number of vCPUs (cores) used must not be oversubscribed. As a starting point, assume an amount equal to 25% of total vCPUs in use should be allocated for hyperthreading overhead. For example, if there are 2 VMs on a host, and each uses 8 vCPUs (16 vCPUs in total), then allocating an additional 4 vCPUs (at least) for overhead is prudent. For hyperthreading, a minimum of 20 vCPUs are recommended.

Limit the ICE Server VM to two snapshots. Disk space equal to twice the size of the ICE deployment size should be available for the snapshots. For example, a medium sized ICE deployment of 750Gb, requires at least 1.5Tb of disk space for two snapshots. Delete snapshots after 48 hours. Follow this same guidance for other VMs sharing the same hardware. A build up of snapshots may result in adverse disk I/O performance due to lack of free disk space.

If the ESXi console displays the VMware virtual machine disks consolidation is needed error message, please address the situation.

When possible, always set the virtual disk write cache policy to write-back, rather than the default of write-through. Please consult VMware or your virtualization hardware vendor for additional information.

It is essential to keep the RAID controller firmware up-to-date. Please consult VMware or your virtualization hardware vendor for additional information.

The following considerations are specific to configuration of a geo-redundant deployment using virtual machines.

To ensure successful geo synchronization during installation, we recommend that the DC1 VM have better or, at least, equivalent compute resources compared to the DC2 VM. Determining compute resources relative to each other is not as simple as looking at their respective host hardware specs and VM settings, you also must consider what other, non-ICE VMs are competing for resources on their respective host hardware.

When powering on DC1 and DC2, wait 5 minutes between the first and second activation. Powering up both VMs simultaneously can cause performance issues.