Database Latency issue

Question

Database Latency issue

Dhruv Garg 0

We are trying this sql database as centralized as we want the flow of data from different regions through the database. Right now this server is in west europe and we are trying to access in India and China at this time, In India the latency is about 5-6 sec and for China its about 2-3 minute. Why is this issue and we have enabled the zone redundancy as well and made the premium subscriptions

Erland Sommarskog 135.4K Reputation points MVP Volunteer Moderator

2026-07-03T10:09:31.21+00:00

Could you be more specific in how you see these latencies? Is this latency when accessing the database from India or China? Exactly what operations are you running?
Dhruv Garg 0 Reputation points

2026-07-03T10:30:37.84+00:00

We are trying the SQL queries with data about 10000 rows and some joins in that using microsoft ssms. The same query for India takes less time as compared to China. So, are just looking how time it is taking. The latency is during the simple but same operation done in both region inside the ssms and Azure SQL Database.
Manoj Kumar Boyini 18,435 Reputation points Microsoft External Staff Moderator

2026-07-03T10:56:05.7533333+00:00
Hi @Dhruv Garg

From your description, the most likely cause is that your application is connecting to a single Azure SQL Database hosted in West Europe from users located in India and China. This means every database request must traverse a long network path, and neither Premium nor Zone Redundancy is designed to reduce this geographic network latency.

Zone Redundancy improves availability by protecting against datacenter or availability zone failures within the same Azure region.

Premium provides more compute, memory, and I/O capacity, but it doesn't reduce the physical distance between clients and the database.

As a first step, I'd recommend measuring the baseline network latency by running a simple query such as:

SELECT 1;

from clients in both India and China. This helps determine whether the delay is primarily network latency or whether additional query/application processing is contributing to the response time.

If the latency is mainly caused by the application making many sequential database calls, you can often improve performance by:

Reducing the number of database round trips (batch operations where possible)

Using connection pooling

Reviewing slow queries with Query Store or execution plans

Using caching for frequently read data

From an architecture perspective:

If the workload is read-heavy, consider Active Geo-Replication so users can read from a secondary replica closer to their location.

If users in multiple regions perform frequent writes (INSERT/UPDATE/DELETE), geo-replication will not reduce write latency because all writes are committed to the primary database. In that scenario, a regional data architecture (such as sharding or region-specific databases) may be required if low write latency is a requirement.

Could you also let us know:

Where is the application hosted (Azure region or on-premises)?

Are the delays seen for every query or only specific operations?

Is the workload primarily read-heavy or write-heavy?

Does the latency persist when executing a simple SELECT 1?

This information will help determine whether the bottleneck is network-related, application-related, or database-related.

References
https://learn.microsoft.com/en-us/azure/azure-sql/database/performance-guidance?view=azuresql
https://learn.microsoft.com/azure/azure-sql/database/active-geo-replication-overview
https://learn.microsoft.com/azure/azure-sql/database/read-scale-out
https://learn.microsoft.com/azure/networking/azure-network-latency

Please let us know if you have any questions.
Erland Sommarskog 135.4K Reputation points MVP Volunteer Moderator

2026-07-03T12:37:35.1966667+00:00

Manoj is correct that zone redundancy is not going to help here. The database is located in West Europe, and some latency is unavoidable from longer distances like India or China, not talking about the overhead in Azure networking. However, that latency would not be noticeable for a plain single query. If you have a pattern where you have an application that inserts one row at a time, it is a different matter. But that is not what you describe.

You say that you are running these queries from SSMS. Where is SSMS located in these tests? On machines on-prem in India and China and to which you connect by Remote Desktop? Or are they on Azure VMs in Indian and Chinese regions?

Are you running exactly the same query text to get exactly the same result in all three cases?What happens if you insert the data from the query into a table, rather than returning rows to the client?
Manoj Kumar Boyini 18,435 Reputation points Microsoft External Staff Moderator

2026-07-04T03:41:45.34+00:00

Hi @Dhruv Garg

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.
Dhruv Garg 0 Reputation points

2026-07-06T08:52:12.3833333+00:00

Thanks @all for the responses. I would appreciate if i get a recommended deployment setup for this requirement where i need clients applications (desktop apps) from multiple global regions are connecting to same shared database. I want to ensure all of them gets the optimum response times.

1 answer

Your answer

Erland Sommarskog 135.4K Reputation points MVP Volunteer Moderator

2026-07-03T10:09:31.21+00:00

Could you be more specific in how you see these latencies? Is this latency when accessing the database from India or China? Exactly what operations are you running?
Dhruv Garg 0 Reputation points

2026-07-03T10:30:37.84+00:00

We are trying the SQL queries with data about 10000 rows and some joins in that using microsoft ssms. The same query for India takes less time as compared to China. So, are just looking how time it is taking. The latency is during the simple but same operation done in both region inside the ssms and Azure SQL Database.
Erland Sommarskog 135.4K Reputation points MVP Volunteer Moderator

2026-07-03T12:37:35.1966667+00:00

Manoj is correct that zone redundancy is not going to help here. The database is located in West Europe, and some latency is unavoidable from longer distances like India or China, not talking about the overhead in Azure networking. However, that latency would not be noticeable for a plain single query. If you have a pattern where you have an application that inserts one row at a time, it is a different matter. But that is not what you describe.

You say that you are running these queries from SSMS. Where is SSMS located in these tests? On machines on-prem in India and China and to which you connect by Remote Desktop? Or are they on Azure VMs in Indian and Chinese regions?

Are you running exactly the same query text to get exactly the same result in all three cases?What happens if you insert the data from the query into a table, rather than returning rows to the client?
Manoj Kumar Boyini 18,435 Reputation points Microsoft External Staff Moderator

2026-07-04T03:41:45.34+00:00

Hi @Dhruv Garg

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.
Dhruv Garg 0 Reputation points

2026-07-06T08:52:12.3833333+00:00

Thanks @all for the responses. I would appreciate if i get a recommended deployment setup for this requirement where i need clients applications (desktop apps) from multiple global regions are connecting to same shared database. I want to ensure all of them gets the optimum response times.

Answer 1

This is a whole lot more than deployment. If you are going to have this setup, desktop apps all over the world, accessing the same Azure SQL Database, it is imperative that the application is written in a way where the latency will have minimal impact. That is, it must not be chatty where they insert or retrieve one row at a time. It must not be using MARS, Multiple Active Result Sets.

Because, no matter how you do it, there will be latency. The absolute minimum latency level is set by the speed of light, but the actual latency is likely to be even higher due to Azure networking. The latency of several minutes that you quote for China is quite extreme, but China is also know for their big Internet firewall and who people sneak by with VPN.

You could reduce the latency from application to databases by putting the application on remote desktops on Azure VMs in the same region and VNET as the databases. But the risk is considerable that the UI would lag in an intolerable way in this case.

If you make it a three-tier application, for instance a web application, the odds are somewhat better. But, again, there will be latency because of the speed of light, and you need to minimise the amount of roundtrips to reduce the damage of the latency.

By the way, Manoj and I asked you a couple of questions which you choose not to answer. That is your call. But had we gotten those answers, we might have been able to give more specific answers. Now we will have to let it suffice with general recommendations.