nova1313

I really like E/R Studio for modeling work. It is expensive, but I've not found a nicer tool to do it across multiple database platforms.

Their website shows they don't officially support the latest versions of PostgreSQL. It does work, but the menus haven't been updated for all the latest features. If you want to use syntax from a newer release, you'll need to do a SQL override until they update the tool.

Their team version keeps the metadata in a database, instead of a file. The locking/unlocking of individual tables is cumbersome. The metadata database doesn't support using PostgreSQL for storing the metadata. I tend to store the models in Git and use the built in model diff tool to do model merges.

ER Studio gets slow when you are north of 10k tables/views in a single model. As a work around, I use sub-models and do not touch the large master model. On a large main model, their visualization tool I believe tries to render all 10k tables. It starts taking 1-2 seconds between clicks. Smaller sub-models (1-2k) off that large main model do not have that issue.

The scripting extension for ER Studio is decent, not great, but significantly improved in the last few versions. It is close to being VBScript. Scripting makes updating really large models a breeze. I use it a lot for storage/index parameter parameter changes during hardware/software upgrades.

I can confirm that the binaries do not stop working after the 6 month.

I suspect it might cut into their Exadata appliance market if they did, but I'd love to hear anything from Oracle indicating they were adding something like this to a new version. If you try to go massively parallel for workloads they seem to push us towards Exadata. Parallel table/index scans may be your best bet at the moment on Oracle.

As far as I know you can't do anything like custom scan providers in Oracle. I've tried searching a few times for functionality like that and always came up empty. Even if that was an available feature, I'm not sure how you'd even begin costing it for the optimizer to take that into account, unless it was something you'd just enable via hint.

I have OpenCL code running from within Oracle stored procedures via Java. You can at least write your own logic from within the DB to utilize your GPUs. I've done things like push 2 tables into GPU memory and let the GPU do the joins once you manually load the columns to join on. Return the set to SQL and use an indexed column to return the data you want from the table. It's not something I'd consider running in production, but it was a neat exercise and you can get it to work.