Oracle is the dominant database deluxe, with the highest function and highest prices. Previous Wikibon research has been looking at best practice for architecting infrastructure to optimize Oracle license spend. This research focusing on best practice for protecting Oracle databases.
Oracle offers ultra-high availability capability at ultra-high costs, i.e at over four to five times the standard enterprise license costs (see Figure 1 below for comparative application recovery costs by topology).
One of the challenges is to move beyond expensive “one-size-fits-all” infrastructures, and enable a more flexible approach to data protection that allows the costs and business benefits to be in line for every application. Ideally this would be offered as options backup-as-a-service and availability-as-a-service within a “Database-as-a-Service” framework. This would allow the business user to dial-up or dial-down data protection and availability in the light of changing business priorities.
This research indicates that best practice can reduce the cost of very-high availability to between 25%-50% of the Oracle MAA costs, and also offer different levels of backup and availability to the lines of business.
RPO & RTO
RPO & RTO are the foundation of every business continuity decision. The recovery point objective (RPO) is the amount of data that is lost following a failure (going back to previous recovery point), and recovery time objective (RTO) is how long it take to get the users productive again (not just the system recovered). There is no system with zero RPO and RTO, and the cost of the recovery system increases sharply the closer to zero the goal. The business running the application has to decide the business impact of RPO and RTO system level agreements, and understand the trade-off with the business impact.
The relative costs of different topologies are shown in Figure 1.
Figure 1 shows the comparative recovery costs of different application recovery topologies. Oracle's Maximum Availability Architecture (MAA) fits into column 5 of Table 1.
Oracle Ultra-high Availability Options
Oracle is a very sophisticated data base system, and has a number of additional high availability options. These include:
- Oracle RAC - Oracle Real Application Clusters (RAC) is an option of Oracle Database Enterprise Edition.
- Oracle RAC is a cluster database implementation, using a shared cache architecture providing highly available database solutions.
- Oracle Active Data Guard
- Oracle Data Guard provides management, monitoring, and automation software to install one or more standby databases. The improves RPO and RTO by helping to protect Oracle data from failures, disasters, human error, and data corruptions while improving the availability of applications. Oracle Active Data Guard 11g extends basic Data Guard functionality by allowing read-only access to a physical standby database while continuously applying changes received from the primary database. This increases performance and return on investment by offloading ad-hoc queries, Web-based access, reporting, and backups from the primary database while also providing disaster protection.
- Oracle GoldenGate
- GoldenGate provides an active replica (includes real-time change data capture, distribution, and delivery for transactional data across heterogeneous systems). GoldenGate supports continuous availability, zero downtime migration, and use of the second copy for (say) business intelligence.
The Oracle License costs are:
- Oracle Enterprise Edition $47,500/core
- Oracle Real Application Clusters $23,000/core
- Oracle Active Data Guard $10,000/core
- GoldenGate $17,500/core
A complete list of Oracle high availability software and the Oracle Maximum Availability Architecture is in Note 1 in the footnotes. The Oracle high availability options are good, but very expensive. The Wikibon research referenced above show that normal Oracle can be well over 80% of the total cost of infrastructure. High availability Oracle doubles the cost of each node, and is four-five times the total cost.
The use of alternative approaches can reduce the cost to between 30% and 50% of the Oracle MAA cost.
Best Practice for Speeding up Backup
There are two main reasons for backup to be fast:
- The avoidance of long planned outage times which take applications away from users and customers (which often lead to unplanned outages if backup windows are exceeded)
- The use of space-efficient snapshots is a best-practice technology here, which allows the application to continue after a brief pause to write data held in memory to disk-based persistent storage.
- How often these snapshots are taken is good lever for providing a basis for different levels of backup-as-a-service.
- There is a strong need for technologies that provide an integrated catalog of snapshots with automated management.
- The quicker data is sent off-site, the less data is lost in the case of a disaster, the better the RPO SLAs can be met:
- RPO can be improved by the use of in-line de-duplication (e.g., EMC’s Data Domain with front-end DD Boost) to achieve high levels of de-duplication and the ability to replicate the changed data off-site very quickly.
- RPO can be improved by the use of RMAN multiplexing to provide parallel streams to multiple in-line de-duplication devices (e.g., EMC Data Domain) , with the ability to restore from them in parallel (improves RTO).
- Client-based de-duplication (e.g., EMC Avamar), is a good choice for smaller and distributed databases and smaller virtualized oracle databases.
Best Practice for Speeding up Recovery
Backup-as-a-service is one thing: recovery-as-a-service is everything.
One of the interesting new very high availability technologies recently certified by Oracle is the use of Active/Active technologies such as EMC VPLEX and IBM SVC. This is a good and lower-cost alternative to Oracle’s Data Guard/Active Data Guard and GoldenGate. It is suitable for metro distances, and needs to be supplemented by a distance copy. Long-distance Active/Active is a pipe-dream, is not best practice, and is unlikely ever to be so except for very small databases. Not least of the reasons is the very high cost of sending data long distances and the high cost of mitigating communications failure.
Early users of Metro EMC VPLEX have found that it provides very low RTO and RPO capability, and can be used with a variety of fail-over techniques, including Oracle RAC and VMware SRM. This again allows a flexible approach to giving backup and availability-as-a-service alternatives to the business users of Oracle databases.
Conclusions & Recommendations
For ultra-high availability, Oracle's MAA with Oracle RAC, GoldenGate and Active Data Guard provides an effective integrated system. However, this system is ultra expensive.
Oracle MAA costs are four-five times the cost of standard enterprise oracle systems. For the vast majority of Oracle systems, backup/availability-as-a-service capability can be achieved at much lower costs. Very high availability best practice is using a subset of the Oracle features, virtualization of the Oracle databases, high-performance storage, active/active storage controllers at metro distances and using other standard backup/recovery features. By using only the standard features of Enterprise Oracle, the cost of high availability can be significantly reduced to between 25% and 50% of the Oracle MAA cost.
Action Item: CIOs & CTOs can now move towards providing a varied cost backup/availability-as-a-service by using a best-practice subset of Oracle features, virtualization of the Oracle databases, high-performance storage, active/active storage controllers at metro distances and using other standard backup/recovery features to optimize on the cost of achieving variable business-led RPO and RTO mandates. This can reduce the cost of high availability services to between 25% and 50% of Oracle MAA costs
Footnotes: Note 1 The full list of the high availability Oracle software includes:
- Real Application Clusters
- Data Guard
- Secure Backup
- Recovery Manager (RMAN)
- Flashback Technologies
- Cloud Computing
- Cloud Storage
- Cross-Platform Transportable Tablespace
- Edition-Based Redefinition
- Online Reorganization
Oracle also offers the Oracle Maximum Availability Architecture (MAA), a blueprint of Oracle high availability best practice. Figure 2 shows the details of this high-cost architecture.