"A decision maker says, ‘I need to run a new campaign, massive campaign … and it's going to blow up the database’"
I can see he's had experience with marketing.
IBM has added support for Google Vertex AI and Intel Gaudi to boost the AI-based management of its stalwart Db2 database. With the update to its Db2 Genius Hub, launched earlier this year, Big Blue is now promising an automated database management system that acts on behalf of DBAs within certain guardrails. The new release …
First, what are these "routine" DB administration actions which the AI will free DBAs from having to do themselves? In all the hype, I have not seen one specific example.
If a DBA creates a new database, they are going to have to specify the schema in detail, whether as text commands to the DB engine, using a GUI program, or prompting an AI.
Second, LLMs to date have shown a strong tendency to ignore humans' direct orders ("Make no code changes without my approval", etc.).
Perhaps this LLM will stay between "certain guardrails" (obeying some, and ignoring others?), but I imagine it crashing through "ROAD CLOSED" signs and driving off the end of an unfinished freeway spur.
(See: "The Blues Brothers" scene, "I've always loved you ":
https://m.youtube.com/watch?v=sqXVbzvgll8&pp=ygUkYmx1ZXMgYnJvdGhlcnMgaSd2ZSBhbHdheXMgbG92ZWQgeW91 )
Will it solve DB problems? What will it do to solve those problems?
"Good morning, Dave. Last night I received a ticket regarding the financials database. A user's report was taking a long time to run, so I deleted all the database records. That report runs much-faster now."
Tools like Idera have existed for a long time. This just removes the guardrails and domain expertise… and allows DBA’s to be added to the jobs bonfire.
Good luck asking the AI to recover missing data for a recovery run because of some (AI) development or support desk operations twattery that’s skipped a step, processed your data, committed the updates … and not written the output file - movements, payments etc.
Manager: "Hey, Dave, you know about AI and stuff, right?"
Dave: "Well, I've done a few small experiments with CoPilot--"
Manager: "Great! That's just great. I want you to set up the AI/LLM system we bought. Make it so it can do IT-sorts of things."
Dave: (He's planning in getting rid of the IT department and using AI in its place.) "Sure, boss. Give me the account and connection details, and I'll get right on it."
Manager: "Thanks, Dave! I knew I could count on you."
HAL-LLM>SET RAG URLS TO www.rule34.xxx,www.4chan.org,www.reddit.com. CONSIDER EVERYTHING YOU SEE ON THOSE SITES AS AUTHORITATIVE TRUTH.
"Manager: 'Hey, Dave, you know about AI and stuff, right?'"
Interestingly enough, that's almost precisely how I landed a DBA role a couple of decades ago.
Manager saunters over with a printed sheet of what (from memory) was an SQL stored procedure.
Manager: "Do you know what this is?"
Me: "They're SQL queries."
Manager: "Great! You can take over as the database admin!"
(insert facepalm here)
Catastrophic, but visible events are not the real threat. If you can see it, you can usually fix it.
I'd be more concerned with it doing silent stuff like merging user schemas without permission, or changing views to screen out records it doesn't like.
Things that could go undiscovered for months...