UCM: Difference between revisions

From HPCWIKI
Jump to navigation Jump to search
No edit summary
No edit summary
Line 1: Line 1:
=== About UCM ===
== About UCM ==
UCM stands for "Universial Cluster Management". UCM has been developed to control multiple nodes under UCM based cluster.
UCM stands for "Universial Cluster Management". UCM has been developed to monitoring and control multiple nodes under UCM based cluster.


UCM software package shipped on HPCMATE UCM master node. UCM supports administrator to manage multiple cluster nodes and real-time monitoring much easier.
UCM software package shipped on UCM master node (1U appliance) and optional Slave node (same spec with master) for active and backup configuration.  


With UCM, administrator to manage multiple cluster nodes and real-time monitoring much easier using the provideds web interface and command line interface (UCMCLI).


=== UCM Plugin ===
== UCM Appliance ==
UCM has plugin architectures to supports various features as plugin along with default features
HPCMATE provides UCM as appliance for hyperscale cluster configuration. UCM master/slave redundancy operation with node lock license of default and user selectable optional plugins.
 
UCM customer can add one or more optional plugin features whenever they wants after purchasing UCM appliance with updated license contents on the master/slave.
{| class="wikitable"
{| class="wikitable"
|+
|+
!Name
!
!Type
!Type
!Location
!Description
!Packages
!Notes
!Notes
!Default
Plugin
|-
|UCM Master
|1U rack mount server
|Working as a master node to control and monitoring multiple workers inside of UCM cluster
| rowspan="2" |node lock license
| rowspan="2" |Yes
|-
|UCM Slave
|1U rack mount server
|Working as a slave to replace master failure
|}
=== UCM Plugins ===
UCM has been designed to [[support]] plugin architecture to supports various features for various requirements
{| class="wikitable"
|+
!Name
!Descriptions
!Default
!Optional
|-
|Thermal managment
|Active thermal monitoring and management for cluster nodes
|Yes
|
|-
|Service volumes
|Service volume exports for multiple worker nodes
|Yes
|
|-
|Rack managment
|Visualize rack management
|single rack
|multi racks
|-
|Resource monitoring
|Real time resource monitoring
|Yes
|
|-
|Cluster logging
|Real time cluster system logging
|Yes
|
|-
|-
|UCMAgent
|LDAP
|Systemd
|Centralized user management
|each node
|Yes
|Python3 packages
|
|
* /etc/systemd/system/ucmagent.service
* Used configuration
* /opt/hpcmate/ucm/config/env for log level
* make.sh - build and release ucmagent
* setup.py - compilation
|-
|-
|[[Docker]] swam
|Centralized docker swarm management
|
|
|Yes
|-
|Slurm
|Slurm clustering and monitoring
|
|Yes
|-
|DFS
|Scalable distributed file system
|
|
|Yes
|-
|Diskless computing
|Diskless working node control/managmeent for HPC computing farm
|
|
|Yes
|-
|DNS/IP Gateway
|Master working as IP router for cluster nodes
|
|
|Yes
|-
|Private registry
|Docker private registry over internet and interanet
|
|
|Yes
|-
|-
|Jupyter farm
|Clustered Jupyter farm - useful for educational organization
|
|
|Yes
|-
|PACS
|Useful for AI/MLops for medial organization
|
|
|Yes
|-
|VPN
|Private VPN features
|
|
|Yes
|-
|CI/CD [[test]] farm
|CI/CD for real time multiple DUT test farm. especially useful for local and remote developers when limited available samples such as electronics prototype of semiconductor, car, etc
<nowiki>*</nowiki>Need optinal PDU and console server
|
|
|Yes
|-
|Render farm
|Enable Blender remote renderfarm with inhouse Blender addon software 
|
|
|Yes
|}
|}


== For further ==
== For further ==
Please refer UCM manual (only available to HPCMATE customers)
more details are available on UCM manual page (limted to HPCMATE customers)
[[Category:Internal]]
[[Category:Internal]]

Revision as of 16:58, 8 April 2023

About UCM

UCM stands for "Universial Cluster Management". UCM has been developed to monitoring and control multiple nodes under UCM based cluster.

UCM software package shipped on UCM master node (1U appliance) and optional Slave node (same spec with master) for active and backup configuration.

With UCM, administrator to manage multiple cluster nodes and real-time monitoring much easier using the provideds web interface and command line interface (UCMCLI).

UCM Appliance

HPCMATE provides UCM as appliance for hyperscale cluster configuration. UCM master/slave redundancy operation with node lock license of default and user selectable optional plugins.

UCM customer can add one or more optional plugin features whenever they wants after purchasing UCM appliance with updated license contents on the master/slave.

Type Description Notes Default

Plugin

UCM Master 1U rack mount server Working as a master node to control and monitoring multiple workers inside of UCM cluster node lock license Yes
UCM Slave 1U rack mount server Working as a slave to replace master failure

UCM Plugins

UCM has been designed to support plugin architecture to supports various features for various requirements

Name Descriptions Default Optional
Thermal managment Active thermal monitoring and management for cluster nodes Yes
Service volumes Service volume exports for multiple worker nodes Yes
Rack managment Visualize rack management single rack multi racks
Resource monitoring Real time resource monitoring Yes
Cluster logging Real time cluster system logging Yes
LDAP Centralized user management Yes
Docker swam Centralized docker swarm management Yes
Slurm Slurm clustering and monitoring Yes
DFS Scalable distributed file system Yes
Diskless computing Diskless working node control/managmeent for HPC computing farm Yes
DNS/IP Gateway Master working as IP router for cluster nodes Yes
Private registry Docker private registry over internet and interanet Yes
Jupyter farm Clustered Jupyter farm - useful for educational organization Yes
PACS Useful for AI/MLops for medial organization Yes
VPN Private VPN features Yes
CI/CD test farm CI/CD for real time multiple DUT test farm. especially useful for local and remote developers when limited available samples such as electronics prototype of semiconductor, car, etc

*Need optinal PDU and console server

Yes
Render farm Enable Blender remote renderfarm with inhouse Blender addon software Yes

For further

more details are available on UCM manual page (limted to HPCMATE customers)

Retrieved from "http://wiki.hpcmate.com/index.php?title=UCM&oldid=601"