Please share with the community what you think needs improvement with Oracle Exadata.
What are its weaknesses? What would you like to see changed in a future version?
General perception from many customers is the cost and maintenance of exadata database machine. Although Oracle recently introduced low cost storage and automated many features, still the cost and maintenance is a concern. Oracle should provide free of cost the Exadata software on the box to have more customers.
There is no room for improvement because everything that they could solve, they have. It has good storage optimization, hardware actualization, and internal communication network. They made improvements in whatever they could. The problem with the Oracle Exadata is that it is a dead-end of technology. The supplier's market is dying, so they have no future. For analytical purposes, companies are now using analytical databases. So they're using databases like MPP, multi-parallel processing databases, things like Greenplum, and Teradata. Exadata has no future. It is going into history. Five years ago, it was top technology. Now, they don't have a future. It's too expensive per terabyte. It's complex and hard to expand.
Try to diminish the logging service.
We have had issues with system restoration. If there is a system fail, the internal drives are useless. We need Exadata to integrate with that one, and to restore it back. But there should be something in place in case of system failure. One of the systems should have the facilities to troubleshoot from the other nodes to this note remotely. Or, even if not remotely, just to troubleshoot at least among the thrusters. There is no system collectively. Let us say this system failed; do we have any disks? Total disks, let's just say 100 disks. Five disks are failed. 100 disk measuring is not data top. If mirroring is that, like we can HRCMR. Let's say IBM CRM are CTC shadow image, REMC like that, the storage level they are giving the facilities inserted data mirroring, so immediately we can bring the DR. But internally, the Exadata doesn't have such features. Maybe they can think through that mirroring of collective system. Insert automatic measuring. If we have a collective recovery, let's say system-to-system, that DR solution, though they are giving a data guard, we have to set it up into the second system and the Data guard is a software solution. There is a really high skill required for a DBA to bring up the database. So a non-skilled person, like the storage engineer, can bring up for shadow images. HRCMI, HRCMR, like that. But in the case of Oracle, you must be really specialized to to bring up the data guard. In addition, they should insert automatic measuring. AM is, let's say, one table fail. Okay, you can recover it. Or one disk failed, then they can troubleshoot it. But automatic recovery, fine, but if we have a collective recovery, let's say system-to-system, that DR solution, though they are giving a data guard. We have to set it up into the second system and the Data guard is a software solution. There is a really high skill required for a DBA to bring up the database. For a non-skilled person, like the storage engineer can bring up for shadow images. HRCMI, HRCMR, like that. But in case of Oracle, you need to be really specialized to to bring up the data guard.
we do have a little trepidation with systemd as it does have a learning curve . Also changing to a binary logging format for us feels like retrograde motion , but sadly almost all linux varients have moved in this direction.
My biggest gripe has been patches which has dramatically improved since our initial Exadata was delivered (January 2011). The only issues we periodically experience are with non-default RPMs on the database nodes. These may fail during the pre-req check which means opening a SR with support. This has become the exception, not the norm so overall not much to complain about. The X2-2 used to experience frequent disk failures but now, that is a thing of the past.