Search
  • Preetam Zare

My VVDS - Design Decisions on Clusters and Racks

This post apparently will be smaller but it is an extension of the previous post. Here we have to make a decision on Server placement in the Rack.


You might surprise how come this comes as first step than the sizing of ESXi. I'm at this stage following VVDS. I will give relevant references in the future.



Design Decision

PDES003: All the nodes required for the management & Compute cluster will be placed in same Rack. As per PDES:001 we have two AZ. PDES003 applies to both Landside and Cityside DC.

Design Justification

  • Easy to locate and manage the Stack be management or Edge

  • Simplify rackings of the servers

  • Easier Troubleshooting

Design Implication


Scope: Both AZ will have the impact of this design decision

In the initial sizing, we are assuming we might need minimum 8 Rack Units excluding cable management arms. We must ensure both sides have sufficient power and cooling. The required power and cooling will be shared in the later phase.


Scope: Only AZ1 will be impacted by this design decision

On Landside DC, we can accommodate a maximum of 8 nodes in single Rack due to space constraints, so for future expansion, we must put nodes in different Rack. So we do not have any short impact but long term.


Scope: Only AZ2 will be impacted by this design decision

On Cityside DC we might face challenges. As discuss in PD02: Devices are not racked in particular order this might impact how we rack servers in CitySide DC. So there is a dependency for CitySide DC which needs to be first fulfilled.


Design Decision

PDES004: The separate powerfeeds will be used for the Racks used for VVDS design in both AZ.

Design Justification

Redundant power feeds increase the availability of the entire Rack.


Design Implication

Scope: Both AZ will have the impact of this design decision

All equipment used or proposed must support redundant power supply. The equipment must run uninterruptedly if one of the power feeds fails. The equipment must be tested against these criteria during UAT phase.


Scope: Landside DC

The Landside has Powerfeeds going into racks and backed by UPS but the power is distributed not qually to the Racks as types of equipment are not distributed based on the power consumption. For future expansion e.g. Compute cluster we must considered this aspect for server placements in Rack.

As UPS power is directly going into the Rack, there are different power connectors than the one normally are shipped with the server. Correct Power Cable types for the server to power socket connection must be ordered.


Scope: CitySide DC

In Cityside DC, no UPS power is available. This is big Risk associated. Unavailability power in Cityside DC can reduce the functionality of service and might corrupt data and the Servers can get damaged. Risk Mitigation needs to be done. I have to update this section later.


#VVDSPost02 #VVDS


17 views

©2019 by virtual2Cloud. Proudly created with Wix.com