Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Scroll Documents: Update page title prefix

...

Info

When deciding how many UMS Servers and UMS Load Balancers you need, simply counting your endpoint devices is not enough. Most importantly, you have to analyze the entire network environment as well as the other circumstances within your workplace. See (12.04.120-en_orig) Installation and Sizing Guidelines for IGEL UMS as well as (12.04.120-en_orig) IGEL UMS Sizing Guidelines & Architecture Diagrams and contact your IGEL reseller to get counsel.

UMS Server & UMS Load Balancer Are Installed on the Same Host Machine

The most common scenario when deploying UMS High Availability is to install the UMS Server and UMS Load Balancer on the same host machine. Both the UMS Server and the UMS Load Balancer offer redundancy and are installed on two servers. The database is ideally designed as a cluster.

Typical Use Cases#UMS Server + UMS Load Balancer

The installation on the same host machine is suitable if

2 UMS Servers 
2 UMS Load Balancers


Drawio
bordertruediagramName1
zoom1
simple1
pageId71974158
custContentId74812292
lbox1
diagramDisplayNameSimpleHA_ConfigurationsimpleViewertrue
contentVer1
revision2
baseUrlhttps://igel-jira.atlassian.net/wiki
diagramNameSimpleHA_Configuration
width500
linksauto
tbstylehidden
lboxtrue
diagramWidth352
revision2

In this configuration, each of the two servers can also perform the tasks as a UMS Server alone. If both servers are active at the same time, this has a load-distributing effect. Note, however, that the load balancer generates extra load along with the actual UMS Server.

UMS Server & UMS Load Balancer are Installed on Separate Host Machines

If you need to manage a very large number of devices and/or do not want the server resources to be shared between the load balancer and the UMS Server, the installation on separate hosts should be considered.

Typical Use Cases#UMS Server Standalone
& Load Balancer Standalone

The installation of the load balancer on a separate host machine is

  • required if the number of devices > 50,000
  • recommended if you do not want the load balancer to consume resources on the UMS Server host

Smallest typical configuration:

2-3 UMS Servers
2 UMS Load Balancers

General sizing recommendations:

  • up to 6 UMS Servers
  • up to 3 UMS Load Balancers
  • 1 UMS Server per max. 50,000 devices
  • 1 LB per max. 3 UMS Servers


Drawio
bordertruediagramName1
zoom1
simple1
pageId71974158
custContentId74779521
lbox1
diagramDisplayNameFull_HA_FinalsimpleViewertrue
contentVer1
revision1
baseUrlhttps://igel-jira.atlassian.net/wiki
diagramNameFull_HA_Final
width500
linksauto
tbstylehidden
lboxtrue
diagramWidth425
revision1

In the smallest typical configuration, queries from the devices are passed on to the UMS Servers by both load balancers. If one of the load balancers should fail, the other remains available and assumes responsibility for communications alone. A great number of UMS Servers could overload a single load balancer, which would then become itself a bottleneck. Therefore, there are provisions for no more than three UMS Servers in this configuration. For very large installations with more than three UMS Servers, the number of load balancers should be increased accordingly.

...

Note
  • High Availability with IGEL UMS Load Balancers: All UMS Servers and UMS Load Balancers must reside on the same VLAN.
  • For High Availability (UMS HA) with IGEL UMS Load Balancers, network traffic must be allowed over UDP broadcast port 6155, and TCP traffic and UDP broadcast traffic over port 61616. For further port configuration, see (12.04.120-en_orig) IGEL UMS Communication Ports.
  • The network configuration on Windows Servers must have the TCP/IPv6 option enabled for UMS 12.
  • IGEL UMS HA installation with IGEL UMS Load Balancers is not supported in cloud environments like Azure / AWS as they do not allow broadcast traffic within their networks. The HA installation without IGEL UMS Load Balancers (as well as the Distributed UMS) is, however, supported in cloud environments as of UMS version 6.10.

...