Powering real-time identity graphs with strong consistency: Aerospike Graph 2.5.0

Identity resolution at scale has never been more critical — or more complex. In the AdTech world, where billions of signals pour in each day and targeting precision defines competitive advantage, data consistency is segueing from a nice-to-have to table stakes.

That’s why I’m excited to introduce Aerospike Graph Service (AGS) 2.5.0, which supports strong consistency (SC) in graph workloads, enabling guaranteed atomicity and isolation for real-time graph updates. Backed by the new Aerospike Database 8.0, this release makes AGS a formidable engine for delivering accurate, scalable, and always-on graph solutions.

Why strong consistency, and why now?

AdTech platforms rely on identity graphs to stitch together diverse user identifiers — cookies, device IDs, hashed emails, and more — into unified customer profiles. These graphs are updated in real time as new data signals arrive.

The challenge? Concurrent writes can trigger race conditions and partial updates, undermining data accuracy.

Without strong consistency, simultaneous updates, such as a mergeV() to unify identities and an addV() triggered by a new signal, can result in duplicate or orphaned profiles. In practice, this can lead to wasted ad spend, poor attribution, and degraded personalization — the exact opposite of what these platforms try to achieve.

AGS 2.5.0 solves these issues with transactional isolation for mutation queries. That means:

• Atomic profile updates: No partial merges or lost edges during concurrent writes. All mutation Gremlin steps can now be executed atomically.

• Reliable attribution: Consistent, real-time identity graphs for better targeting accuracy.

• No orphaned identities: Query steps like mergeV() and addE() are executed transactionally, eliminating data drift.

How it works

AGS 2.5.0 introduces support for Aerospike transactions, which apply to all mutation queries when the database namespace is configured with SC mode. With transactions enabled:

• Mutation queries are atomic and isolated, even during cluster splits.

• Read-only queries continue to follow an eventual consistency model for performance.

• Isolation levels differ for reads and writes, giving you the flexibility to optimize where it matters most.

Want more control? You can configure transactional behavior using:

aerospike.graph.mrt.enabled=true
aerospike.graph.mrt.timeout=60

The command above lets you balance write consistency with performance by tuning transaction timeouts (up to 120 seconds). For developers, AGS offers built-in support for retry patterns to handle write contention at scale gracefully.

For more details, check out the technical documentation page.

Other highlights in 2.5.0

We didn’t stop at transactions. Here are a few other key improvements:

• Query tracing: Get a detailed view into query execution paths to identify performance bottlenecks.

• Metric enhancements: New observability metrics (server_connection_queue_size and server_io_queue_size) offer deeper insights into AGS performance.

Bug fixes: Smarter union step results, better timeout handling, and clearer error messages when using unsupported vertex IDs.

We also made improvements to the Graph bulk loader, reducing log noise and improving error visibility during secondary index operations.

Built for AdTech, ready for more

AGS 2.5.0 was designed with AdTech in mind, but its capabilities extend to any domain that relies on fast, consistent relationship data: fraud detection, personalization, recommendation engines, and more.

If your graph workloads demand millisecond performance, global scale, and true transactional integrity, this release was made for you.

We’re just getting started. Try out AGS 2.5.0 today, and let us know what you think.