ODA started to become popular with X3-2 and X4-2 in 2013/2014. These 2 ODAs were very similar. The X5-2 from 2015 was different with 3.5 inches disks instead of 2.5 inches and additional SSDs for small databases (FLASH diskgroup). All these 3 ODAs were running 11gR2 and 12cR1 databases and were managed by the oakcli binary. If you’re still using these old machines, you should know that there is a lot of differences compared to modern ODAs. Here is an overview of what have changed on these appliances.
Starting from X6, ODAs are also available in “lite” versions, understand single-node ODAs. The benefits are real: way cheaper than 2-node ODAs (now called High Availability ODAs), no need for RAC complexity, easy plug in (power supply and network and that’s it), cheaper Disaster Recovery, faster deployment, etc. Most of the ODAs sold today are single-nodes ODAs as Real Application Cluster if becoming less and less popular. Today, ODA’s family is composed of 2 lite versions, X8-2S and X8-2M and one HA version X8-2HA.
Support for Standard Edition
Up to X5, ODAs had only supported Enterprise Edition, meaning that the base price was more likely a 6-digit figure in $/€/CHF if you pack the server with 1 EE PROC license. With Standard Edition, the base price is “only” one third of that (X8-2S with 1 SE2 PROC license).
Full SSD storage
I/Os have always been a bottleneck for databases. X6 and later ODAs are mainly full SSD servers. “Lite” ODAs are only running on NVMe SSD (the fastest storage solution for now), and HA ODAs are available in both configurations: SSD (High Performance) or a mix of SSD and HDD (High Capacity). The latest one being quite rare. Even the smallest ODA X8-2S with only 2 NVMe SSDs will be faster than any other disk-based ODA.
Higher TeraByte density and flexible disk configuration
For sure, comparing a 5-year old ODA to X8 is not fair, but ODA X3 and X4 used to pack 18TB in 4U when ODA X8-2M will have up to 75TB in 2U. Some customers didn’t chose ODA 5 years ago because of the limited capacity, it’s even no more a subject today.
Another point is that storage configuration is more flexible. With ODA X8-2M you are able to add disks by pair, and with ODA X8-2HA you can add 5-disk packs. There is no more the need for doubling capacity as we did on X3/X4/X5 (and you could only do it once).
Furthermore, you can now choose an accurate disk split between DATA and RECO (+/-1%) compared to the DATA/RECO options on X3-X4-X5: 40/60 or 80/20.
A real appliance needs a real GUI, X6 introduced the ODA Web GUI, a basic GUI for basic ODA functions (dbhomes and databases creation and deletion, mainly) and this GUI became more and more capable during the past years. If some actions are still missing, the GUI is now quite powerfull and also user-friendly. And you can still use the command line (odacli) if you prefer.
ODA now has a repository and everything is ordered and referenced in that repository, each database, dbhome, network, job is identified with a unique id. And all tasks are backgroup jobs with a verbose status.
Next-gen virtualization support
With old HA you had to choose between bare-metal mode or virtualized-mode, the last one being for running additional virtual machines for other purposes than databases. But the databases were also running on a single dedicated VM. Virtualized-mode relied on OVM technology, soon deprecated and now replaced with OLVM. OLVM brings both the advantages of a virtualized ODA (running additional VMs) and a bare-metal ODA (running databases in bare-metal). And it relies on KVM instead of Xen, which is better because it’s part of the Linux operating system.
Data Guard support
It’s quite a new feature, but it’s already a must-have. The command line interface (odacli) is now able to create and manage a Data Guard configuration, and even do the duplicate and the switchover/failover. It’s so convenient that it’s a key benefit for the ODA compared to other platforms. Please have a look at this blogpost for a test case. If you’re used to configure Data Guard, you will probably appreciate this feature a lot.
ODA has always been a great challenger compared to other platforms. Regarding modern ODAs, NVMe SSDs associated to high-speed cores (as soon as you limit the number of cores in use in the ODA to match your license – please have a look how to) make the ODA a strong performer even compared to EXADATA. Don’t miss that point, your databases will probably run better on ODA than on anything else.
If you’re using Oracle databases, you should probably consider again ODA in your short list. It’s not the perfect solution, and some configurations cannot be addressed by ODA, but it brings much more advantages than drawbacks. And now there is a complete range of models for each need. If your next infrastructure is not in the Cloud, it’s probably with ODAs.