Before you begin

• Running Clustered AtScale requires the external load balancer and three AtScale machines, referred to as follows:

  • Coordinator Host - runs the AtScale Coordinator
  • Master Host (AtScale Application Host) - runs full AtScale (only designated "Master" during installation but not at run-time)
  • Standby Host (AtScale Application Host) - runs full AtScale (only designated "Standby" during installation but not at run-time)

• AtScale requires a number of ports to be available, and no other services should run on those ports. See below for the complete list of ports.

About this task

Provision the AtScale hosts and open ports as described:

AtScale Application Host Specification

Property	AtScale Application Host System Requirements
64-bit Linux Operating Systems	CentOS 7 Red Hat Enterprise Linux (RHEL) 7 Red Hat Enterprise Linux (RHEL) 8 Ubuntu 20.04 Ubuntu 22.04
Software and OS Utilities	curl, rsync, unzip, sed, awk, ping (use iputils-ping if Debian or Ubuntu), whoami, hostname, find, cp, mv, grep, sort, date, tail, timeout, ps, env, which, cat, libreadline-dev (required by Ubuntu 20.x)
Memory	128 GB recommended, 64 GB minimum
CPU	32 cores recommended, 16 cores minimum
Disk	200 GB (for AtScale installation and metadata catalog)
Network	1 Gbps reliable network connectivity between the AtScale server and the data warehouse Fully Qualified Domain Name (FQDN) host addresses should be used rather than IP addresses. Each FQDN must have a valid DNS entry and resolve to a valid IP address. Reverse DNS must be set up for your network. AtScale's master database will not start properly in a clustered installation without access to Reverse DNS. In the case of AtScale Clusters, elastic scaling is not supported, therefore your cluster configuration must consist of a static list of host names with working DNS entries. Changing the host membership of the cluster requires down-time.
OS Limits	Use ulimit to remove the atscale user limits for the following: open files - unlimited or 65536. cpu time - unlimited virtual memory - unlimited max user processes - unlimited or 65536. file size - unlimited Be sure to make the limit changes so that they survive a restart. Failure to raise these limits will prevent various AtScale services from starting.

AtScale Coordinator Host Specification

Property	AtScale Coordinator Node System Requirements
64-bit Linux Operating Systems	CentOS 7 Red Hat Enterprise Linux (RHEL) 7 Red Hat Enterprise Linux (RHEL) 8 Ubuntu 20.04 Ubuntu 22.04
Software and OS Utilities	curl, rsync, unzip, sed, awk, ping (use iputils-ping if Debian or Ubuntu), whoami, hostname, find, cp , mv, grep, sort, date, tail, timeout, ps, env, which, cat
Memory	16 GB minimum
CPU	4 cores minimum
Disk	200 GB
Network	1 Gbps reliable network connectivity between the AtScale server and the data warehouse Fully Qualified Domain Name (FQDN) host addresses should be used rather than IP addresses. Each FQDN must have a valid DNS entry and resolve to a valid IP address. Reverse DNS must be set up for your network. AtScale's master database will not start properly in a clustered installation without access to Reverse DNS. In the case of AtScale Clusters, elastic scaling is not supported, therefore your cluster configuration must consist of a static list of host names with working DNS entries. Changing the host membership of the cluster requires down-time.
OS Limits	Use ulimit to remove the atscale user limits for the following: open files - unlimited or 65536. cpu time - unlimited virtual memory - unlimited max user processes - unlimited or 65536. file size - unlimited Be sure to make the limit changes so that they survive a restart. Failure to raise these limits will prevent various AtScale services from starting.

AtScale Virtualization Layer Host Specification

Property	AtScale Virtualization Layer Host System Requirements
64-bit Linux Operating Systems	CentOS 7 Red Hat Enterprise Linux (RHEL) 7 Red Hat Enterprise Linux (RHEL) 8 Ubuntu 20.04 Ubuntu 22.04
Software and OS Utilities	curl, rsync, unzip, sed, awk, ping (use iputils-ping if Debian or Ubuntu), whoami, hostname, find, cp, mv, grep, sort, date, tail, timeout, ps, env, which, cat
Memory	32 GB minimum AtScale recommends allocated a maximum of 75% of memory for virtualization, leaving the remaining for the operating system and buffer cache. Recommended memory allocation depends on application use. To determine how much your application uses for a certain dataset size, use the following calculation: Concurrent Users * Activity Rate * (2 * Average Row Size) * Average Rows Per Query Returned * 200% Join Factor * (200% of Average Query Complexity [1-2]) Average query complexity is relative to the use of features such as time relative calculations or calculations requiring MultiPass.
CPU	8-16 core recommended, 4 cores minimum Recommended memory allocation depends on application use. To determine how much your application uses for a certain dataset size, use the following calculation: Concurrent Users X Activity Rate X 1 Cores Per Query X  Average Query Complexity [1-2] Average query complexity is relative to the use of features such as Time Relative calculations, or calculations that require MultiPass.
Disk	200 GB minimum (for AtScale installation and metadata catalog)|br| AtScale recommends 4-8 disks per node configured without RAID as separate mount points. (Linux: Mount disks with the noatime Linux mount option.)
Network	1 GBPS reliable network connectivity minimum, 10 GBPS recommended Fully Qualified Domain Name (FQDN) host addresses should be used rather than IP addresses. Each FQDN must have a valid DNS entry and resolve to a valid IP address. Reverse DNS must be set up for your network. AtScale's master database will not start properly in a clustered installation without access to Reverse DNS. In the case of AtScale Clusters, elastic scaling is not supported, therefore your cluster configuration must consist of a static list of host names with working DNS entries. Changing the host membership of the cluster requires down-time.
OS Limits	Use ulimit to remove the atscale user limits for the following: open files - unlimited or 65536. cpu time - unlimited virtual memory - unlimited max user processes - unlimited or 65536. file size - unlimited Be sure to make the limit changes so that they survive a restart. Failure to raise these limits will prevent various AtScale services from starting.

• For more details on hardware provisioning for virtualization, see Virtualization Hardware Provisioning.

Reserve the following ports for AtScale services:

Use this table to configure your Data Center's firewall.

• IN: Traffic originating from the end client/outside the cluster with a destination into the AtScale cluster.
• OUT: Traffic originating from AtScale with the destination outside the AtScale cluster.

¹ License Protected Feature: This service is not enabled by default and requires a special license to activate.

Internal Communications Within the AtScale Cluster:

Use the following table to configure internal Data Center firewalls, cross Data Center port forwarding, or validate that AtScale servers are operating with reasonable behavior.

• IN: Traffic originating from another AtScale node with a local destination.
• LOCAL: Traffic only within the local node.
• OUT: Traffic originating from the node to another AtScale node.
• INTERNAL: Should only be used for debugging purposes.

Services Running

¹ License Protected Feature: This service is not enabled by default and requires a special license to activate.

Hadoop cluster requirements for AtScale (only required when connecting to a Hadoop Data Warehouse)

When AtScale is connected to a Hadoop cluster, the following permissions and cluster resources are required: