Files
docs/doc/source/planning/kubernetes/redundant-top-of-rack-switch-deployment-considerations.rst
Ron Stone 3143d86b69 Openstack planning
Updates for patchset 2 review comments
Changed link depth of main Planning index and added some narrative guidance
Added planning/openstack as sibling of planning/kubernetes
Related additions to abbrevs.txt
Added max-workers substitution to accomodate StarlingX/vendor variants

Signed-off-by: Ron Stone <ronald.stone@windriver.com>
Change-Id: Ibff9af74ab3f2c00958eff0e33c91465f1dab6b4
Signed-off-by: Stone <ronald.stone@windriver.com>
2021-01-25 08:36:47 -05:00

38 lines
1.8 KiB
ReStructuredText
Executable File

.. gss1552671083817
.. _redundant-top-of-rack-switch-deployment-considerations:
======================================================
Redundant Top-of-Rack Switch Deployment Considerations
======================================================
For a system that uses link aggregation on some or all networks, you can
configure redundant |ToR| switches for additional reliability.
In a redundant |ToR| switch configuration, each link in a link aggregate is
connected to a different switch, as shown in the accompanying figure. If one
switch fails, another is available to service the link aggregate.
.. figure:: ../figures/jow1438030468959.png
*Redundant Top-of-Rack Switches*
|org| recommends that you use switches that support |VPC|. When |VPC| is used,
the aggregated links on the switches act as a single |LAG| interface. Both
switches are normally active, providing full bandwidth to the |LAG|. If there
are multiple failed links on both switches, at least one connection in each
aggregate pair is still functional. If one switch fails, the other continues to
provide connections for all |LAG| links that are operational on that switch.
For more about configuring |VPC|, refer to your switch documentation.
You can use an active/standby failover model for the switches, but at a cost to
overall reliability. If there are multiple failed links on both switches, then
the switch with the greatest number of functioning links is activated, but
links on that switch could be in a failed state. In addition, when only one
link in an aggregate is connected to an active switch, the |LAG| bandwidth is
limited to the single link.
.. note::
You can enhance system reliability by using redundant routers. For more
information, refer to your router documentation.