Multi AZ or Single Zone?

Home»AWS Support Community Forum»Multi AZ or Single Zone?

Multi AZ or Single Zone?

Ron Hoffman

For HA it is recommended to host my application in multi availability zones. Is there a case where I should host application to single AZ?

This discussion contains 4 replies and has 96 views.



An Availability zone is a physically or virtually separated data center. It is recommended to host your application on multi AZ for High Availability but there are cases where you can reconsider your decision.
1) When your application / web server is hosted in multi AZ but your DB is in single availability zone. In this case when the zone where database is down, even though you took precaution for HA, it will not serve your purpose as DB is down.
In this case, it is recommended to either use RDS with multi AZ deployment or keep secondary / slave database in other availability zone as backup.
2)When you have concern about latency, frequency. It is recommended that try to host applications that needs least latency in same AZ.


Nir Rozenberg

Note that Availability zones are mapped to accounts and it is not necessary that two AWS accounts with same availability zone name be in the same data centre.

See this article by Eric Hammond -

So that your default AZ doesn’t become overloaded, AWS manages redirection for availability zones such that each account’s availability zones can map to different physical data centre equivalents. i.e. US-East-1a of my AWS account can be US-East-1b of your AWS account.


Taral Shah

To add to that…if one of the Availability Zones is out or down, it may impact nodes in another AZ due to the services run by the user.
For example, if you have Master -Master setup for your RDBMS  in a separate AZ or you have some cluster node setup for your NoSQL DB.
If now one of the AZs is down, the EBS of that zone will not be accessible and the DB on that EBS may not be able to replicate to the other zone’s EBS.  Also the other Master or node trying to fetch data or replicate data will wait infinitely for replication until the other EBS is up.
This may cause the rest of the database or DB cluster to eventually hang waiting on the nodes.


Aldrin Leal

May I get my two cents?

Reserved Instances. Suppose you just decommisioned a service (and its machine), or you, in a hush offer from the RI Marketplace, got a unbeatable price. Or perhaps IT Authorized the purchase even if you havent had the need (budgeting is like black magic, right?).

So you have a RI. RI’s are bound to an AZ, and you can’t change it

Other than that, you should strive for multi-az and try to aim for multi-region (trust me, AWS outages are hard to bear even in Multi-AZ - But if you’re able to transfer, ot at least keep something in standby ready for the big ‘on’ button, you’re in heaven)

You must be logged in to reply to this topic.