Transactional DDL in SQL Server: In Oracle, It Does Not Compute

Oracle treats most DDL as auto-committing, meaning once it executes, it's done. SQL Server, however, blurs that line in fascinating ways, allowing some DDL operations to be part of an explicit transaction.

We Gave Memory-Optimized Tables a Hash Lookup — Then Tried Pattern Matching Instead

Introduction It was the week before Black Friday — the biggest online ad rush of the year. Our US-based ad-tech platform was gearing up for an insane traffic spike. Hundreds of real-time campaigns were about to go live across multiple brands, each with thousands of user sessions flowing through our system. Every incoming user impression […]

Always On AG Latency for Read-Only Workload After Failover?

Unlocking Interoperability: A Guide to Foreign Data Wrappers in PostgreSQL and Aurora PostgreSQL AWS RDS

Unlocking Interoperability: A Guide to Foreign Data Wrappers in PostgreSQL and Aurora PostgreSQL AWS RDS As a database professional, I often encounter scenarios where data is fragmented across various systems. In today's distributed IT landscape, it's not uncommon for critical business information to reside in different databases, perhaps an on-premise PostgreSQL instance for legacy applications, […]

Life's Little Frustrations

As with others, I've had to deal with death in the family recently. Some other family members are dealing with cancer (a few friends too). Happily none of us has recently been a disaster zone, but that's happened too. So yeah, big, nasty scary stuff happens in life. However, for most of us, most of […]

When INCLUDE Columns Quietly Inflate Your Transaction Logs

In this article, I wanted to test a common assumption we DBAs make – that adding INCLUDE columns to indexes is harmless. I created a FULL recovery test database with a realistic wide Orders table containing extra large VARCHAR columns to simulate an ERP workload. I ran updates and measured transaction log backup sizes before and after adding INCLUDE columns to a nonclustered index. The results shocked me. The update without INCLUDE columns generated a 10 MB log backup, while the same update with INCLUDE columns produced over 170 MB – a 17x increase in log volume. I explain why this happens: INCLUDE columns are physically stored in index leaf rows, so updates affecting them write bigger log records. I also clarify that updating key columns generates even more log than INCLUDE updates because it involves row movement (delete + insert), but INCLUDE updates still cost more log than if those columns weren’t indexed at all. The takeaway is clear – INCLUDE columns are powerful, but they silently increase transaction log generation, impacting backup sizes, replication lag, and DR readiness. Always measure their real cost before deploying to production.

Databases and Disasters

I was just reading about how the Philippines are working to update their databases in support of faster and better responses in the case of an emergency. While I do volunteer for some of the local emergency services, I'm right at the bottom of the heap as just a radio operator. I don't have any […]

Sometimes, Troubleshooting Is Hard

I hop in the Jeep the other day and turn on my ham radio. Have I mentioned I'm a licensed amateur radio operator? Yeah, yeah, I know. I won't shut up about it. Ha! My call sign is KC1KCE. I haven't been on HF in a while, but I'm regularly on the air locally here […]

SQL Server Columnstore Index Fragmentation

People Make Odd Choices

One of my favorite things about going to in-person events is just the time when we're sitting around chatting, out in the hallway, over at the vendor booths, maybe in the speaker room. Any of them. Inevitably, you start to get what I would call "sea stories" (Navy & Coasties, "war stories" for the pickles, […]