ZFS appliance

Robeen

Oracle Database 18c

Solaris

Hello Team,

I am planning to deploy my production environment on ZS7-2 and 2 T8-2 servers for RAC.

Given that there is only 1 ZS7-2 storage, replication would therefore be local - the only option left since I will using features such as HCC and OISP for datawarehousing. In this scenario this single ZS7-2 appliance acts BOTH as source and target for storage projects being replicated. From a DR perspective it would not be wise to have the secondary storage on the same single ZS7-2 which also has the production storage.

1. Please advise on the probability of the ZS7-2 failing.

2. Kindly advise on other features of the ZS7-2 which would be an advantage for datawarehouse.

3. Do you advise me to do the storage replication by mirroring the disks given that there is only 1 storage?

Thanks,

Joe

Cindys-Oracle

Hi Robeen,

ZFS Storage Appliances are designed to provide 5 nines of uptime. Our solution offers full component redundancy (clustered controllers, boot drives, dual power supplies, network fabric, storage, and path redundancy). Includes data storage redundancy options (mirrored and raidz[1-3]). Controllers can also be configured with No Single Point of Failure (NSPF).

However, your production storage system (as you describe) is vulnerable to power failures, water damage, and natural disasters. A mirrored configuration provides data/storage redundancy but is not storage replication and doesn't provide complete protection from external events. Even with a clustered configuration, production systems are generally protected by offsite DR replication and backup/recovery solutions. Can you describe your backup strategy?

Data warehouse OLTP workloads benefit from a mirrored configuration and write SSDs. Specific details are provided in the

Best Practices for Online Transactional Processing (OLTP) Type Workloads in MOS Doc: Guidelines When Using ZFS Storage in an Exadata Environment (Doc ID 2087231.1). Even though this isn't an Exadata environment, the DWH configuration recommendations still apply.

Thanks, Cindy

Robeen

Hi Cindy,

thanks for the update.

The backup strategy is a full backup once  a week followed by incremental backups(Incrementally update backup strategy).

Initially, we planned to provision our datawarehouse on an OceanStore storage. We were planning to setup an Active Dataguard between site Rose-Hill and site Rose-Belle. Also, we would setup an extended rac stretched cluster for MAA architecture.

1. Please advise which one is better to use(in terms of performance and stability etc..) - Solaris servers on ZFS storage or E9000 servers on Oceanstor storage? Given that there will be only 1 ZFS storage, I think the initial setup would be best for an MAA.

2. I have checked the note above. Initially I was planning to provision my servers with the requirements below

Could you please advise if I can provision with same specs @ site Rose Hill if I plan to use ZFS? Here, there will be no replication.

3. How can I include the backup controller?

4. The Oracle Note says 'Tier 1'. Is it NVME?

Thanks,

Robbeen