Azure SQL managed instance and API planing and designing questions to fix the latency issue.

Question

Azure SQL managed instance and API planing and designing questions to fix the latency issue.

Arunkumar Babu 0

I have an application hosted in a single location, but due to latency concerns, the Azure SQL Managed Instance databases are deployed across three regions: the US, India, and Europe. Each region has its own dedicated database. When a user performs operations such as insert, update, or delete, how does the application determine the user’s region and route the transaction to the appropriate regional database?

SAI JAGADEESH KUDIPUDI 3,550 Reputation points Microsoft External Staff Moderator

2026-07-01T01:43:19.5333333+00:00
Hi @Arunkumar Babu ,
How does Azure SQL Managed Instance know which region the user is in?

Based on the provided documentation, Azure SQL Managed Instance itself doesn’t include any built-in “user geo-detection” that automatically determines the caller’s region and then routes the write (insert/update/delete) to the “closest” regional database.

Instead, routing must be handled by your application or by an upstream traffic-management layer, because the database endpoint the app connects to is the one that will receive the transaction.

So in a multi-region design like your US / India / Europe setup, typical patterns are:

Application-level routing: the app determines the user’s region (from the client’s network location, geolocation data, or other application logic) and then directs the request to the corresponding regional Managed Instance endpoint/connection string.

Traffic manager / load balancer routing (if you put a front door in front of your app/API): routing decisions happen before the SQL connection is established, so the application reaches the correct regional backend.

What you should do next (performance + latency)

Since your broader goal mentions fixing latency, the troubleshooting doc we have emphasizes identifying bottlenecks first (CPU vs I/O vs waits), and then tuning accordingly. For Azure SQL Managed Instance, common root causes of latency/slow response include:

Resource saturation (CPU, I/O, memory)

Inefficient query execution (missing indexes, outdated statistics, plan regressions)

You can confirm where time is going by checking:

CPU / I/O / waits via Azure Portal Metrics and DMVs

Identify slow queries using sys.dm_exec_requests and (optionally) Query Store (if enabled), then review execution plans for issues like implicit conversions, scans, tempdb spills, etc.

Follow-up questions (so we can answer your routing scenario precisely)

Are you using one application endpoint (single API) for all regions, or does each region have its own API/backend?

What logic component determines region today—app code, DNS, Traffic Manager, Load Balancer, API Management, etc.?

Are the three Managed Instances separate databases with independent write paths, or is there some kind of replication/sync?

For your latency issue, is the observed problem primarily:

higher CPU time

higher I/O latency / saturation

locks/blocking

or general end-to-end network/API latency?

Do you have Query Store enabled on the Managed Instance databases?

References

Azure SQL Managed Instance performance troubleshooting (slow performance, diagnose bottlenecks, DMVs, indexing, statistics): https://learn.microsoft.com/azure/azure-sql/managed-instance/performance-troubleshooting

Monitoring and performance using DMVs (linked within the provided doc): https://learn.microsoft.com/azure/azure-sql/managed-instance/monitoring-with-dmvs

Architecture best practices for Azure SQL Managed Instance (redirect connection type, performance efficiency): https://learn.microsoft.com/azure/well-architected/service-guides/azure-sql-managed-instance

Azure SQL Managed Instance FAQ (performance + connection behavior mentions): https://learn.microsoft.com/azure/azure-sql/managed-instance/frequently-asked-questions-faq

Hope this helps. If you have any follow-up questions, please let me know. I would be happy to help.

1 answer

Your answer

Answer 1

Is that an application in Azure App Service, or something on-prem? If the application is hosted in a single place, what's the point with having three different instances? Do the users also connect to the database by other means?

My naïve thinking is that when the user connects, you look up the user in a database (which could be a small Azure SQL Database), and you get the connection option for that user from that database. This assumes that you know all your users beforehand. If you don't, you would need to find a way to trace where they are coming from. That may be trivial or impossible depending on the situation. Maybe you need to ask them the first time they connect. (The reason I say that it may be trivial or impossible is because I don't know what users we are talking about. Users in a corporate organisation? Users from the public internet? Something else?)