Maintenance Notification:
On next Monday August 5th 2019 starting at 7am CEST we will conduct some maintenance. You might encounter some issues using the Exasol User Portal and Issue Tracker! We will restore the Exasol User Portal and Issue Tracker before 9am CEST on Monday August 5th 2019.

You are currently viewing an old documentation portal. This content will be removed on June 18th, 2019. Please view our current documentation portal at http://docs.exasol.com/.

Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

If customers require Exasol availability in case of data center outage, a variety of solutions can be implemented that involve two data centers in order to reduce downtime in such an event.

Available Solutions:

1 ETL-based

There’s an autonomous cluster with an autonomous database on each site. Both databases can be used simultaneously. Business critical data is replicated regularly with ETL techniques from one database to the other. In case of site failure, application connections will be re-directed to the other site’s database.

RTO is less than one hour.

RPO depends on the ETL frequency – everything changed after the last ETL replication will be lost.

2 Backup-based

Common for these solutions is that they all rely on backups of the production database on one site that are available on the second site if the first site fails. A restore on the second site makes the production database available again. Unlike the ETL based solution, all backup-based solutions require a cluster with the same number of nodes as the production database.

RTO is less than one hour.

RPO depends on the backup frequency – everything changed after the last backup is lost.

2.1 Test Environment as Standby

Production database is on one site while a Test database is running on the second site, each in a separate cluster. In case of production site outage, the Test database is removed and production gets restored from backup on the second site.

2.1.1 External Backup Server

Production backups are written (additionally) to a backup server located on the second site.

2.1.2 Cluster Backup

Production backups are written (additionally) to the archive volume of the second site’s cluster.


2.2 Remote Mirroring

Backups are written on the production site while the backup storage layer mirrors this backup to the second site. This works transparently to the backup on the production site. The second site’s backup storage cannot be used while the remote mirroring is ongoing. The backup storage vendor is responsible for this technique and additional (license) costs may be charged by the storage vendor accordingly.

2.3 Stretched Archive

The archive volume of the production database is stretched across two sites in this case, while the data volume remains local to the first site. This works transparently to the backup on the production site. The archive volume on the secondary site cannot be used while the mirroring is ongoing. This solution is implemented with Exasol technology, so no additional storage vendor charges occur.

3 Synchronous Dual Data Center (SDDC)

This is the most sophisticated of the shown solutions. It delivers zero data loss with an RTO of less than a minute.

Essentially SDDC means stretching the cluster across two sites, e.g. two fire sections within the same building. DC 1 works as Primary Site with two nodes hosting instances that operate on local Master Segments while their Slave Segments are being synchronously mirrored across the private network to DC 2. See here for more details.

Summary

Solution

RTO

RPO

Pro

Con

1 ETL-based

~ 1 h

significant

Second database can be used autonomously. Cluster configuration on second site can be different from first site.

Appropriate ETL mechanism has to be developed and maintained across changes on the production site.

2.1 Test Environment as Standby

~ 1 h

significant

Second site can be used for testing.

(Additional) backups must be taken to the second site.

2.2 Remote Mirroring

~ 1 h

significant

Backup methods used on production can be done as before.

Storage vendor technology that may lead to additional costs.

2.3 Stretched Archive

~ 1 h

significant

Backup methods used on production can be done as before. Exasol technology used, so no additional costs.

Backup performance might be impacted depending on the quality of the network connection between the two sites.

3 Synchronous Dual Data Center

~ 1 m

0

Minimal downtime without data loss.

Performance Impact on the production database, proportional to the distance to the second site.

Glossary

DC  = Data Center

ETL = Extract Transform Load

RPO = Recovery Point Objective (How much Data Loss is caused by the failure)

RTO = Recovery Time Objective (How long does the outage caused by the failure last)