ECS: Best practices for ECS and Data Domain (DD).

N/A

Best practices for ECS and Data Domain (DD).

There are a wide range of issues with DD when archiving to ECS, these issues can be avoided by following the steps that are outlined below.

Is a Load Balancer needed for a DD and or ECS implementation?
Use of persistence with F5 Load Balancer.

Small file size used by DD which triggers a code issue in the ECS.

The minor code issue that is triggered by small files that are written to the ECS is fixed in ECS 3.1 HF3 and ECS 3.2.

For ECS versions 3.1 HF2 and lower these systems should be upgraded to ECS 3.1.0.3 (ECS 3.1 HF3) or ECS 3.2.0.1 (ECS 3.2 HF1) plus a configuration change which can be applied to avoid issues.

Also, because DD does its own replication it is recommended everyone should set up the DD bucket like this:

Ensure the bucket that is used by DD is part of a local replication group, in other words, a nongeo-replicated bucket.
Ensure Access During Outage (ADO) is not enabled on the bucket that is used by DD.
Note: ADO can be changed after bucket creation but the assigned replication group cannot be changed after bucket creation.

Futures: DD OS 6.1.2 and higher will have a larger chunk size when writing to the ECS Cloud which will result in a better storage efficiency on the ECS. This DD OS should be paired with ECS 3.2.1.0 (ECS 3.2 SP1). Both of these releases should be Generally Available the week of 09-July-2018 if not sooner.

Load Balancer:

• Is a Load Balancer required between the DD and the ECS? Yes
• Persistence Flag: Contrary to EMC documentation regarding an F5 Load Balancer, connections should NOT be sticky or persistent as this will not balance the load to all ECS nodes.

If you have DD archiving to an ECS, contact EMC Support and reference this Knowledgebase Article, KB 513725.