What to do when you hit the throughput limits of Azure Storage (Blobs)

In this post we will talk about how we can detect when we hit a throughput limit of Azure Storage and what we can do in that moment.

Context
If we take a look on Scalability Targets of Azure Storage (https://azure.microsoft.com/en-us/documentation/articles/storage-scalability-targets/) we will observe that the limits are prety high. But, based on our business logic we can end up at this limits.
If you create a system that is hitted by a high number of device, you can hit easily the total number of requests rate that can be done on a Storage Account. This limits on Azure is 20.000 IOPS (entities or messages per second) where (and this is very important) the size of the request is 1KB.
Normally, if you make a load tests where 20.000 clients will hit different blobs storages from the same Azure Storage Account, this limits can be reached.

How we can detect this problem?
From client, we can detect that this limits was reached based on the HTTP error code that is returned by HTTP request.
There two errors code that we should take into account:

• 503 - Service Unavailable (ServerBusy): The server is currently unable to receive requests. Please retry your request
• 500 - Internal Server Error (OperationTimedOut): The operation could not be completed within the permitted time

In general we should get 503 HTTP error code for this case, but in both cases I would consider that I might hitted a limits.

What we can do?

Based on our business logic, there are different solutions for our problems. Let's cover a few scenarios.

Problem 1: During multiple uploads in blobs on the same Azure Storage Account we hit a limits

In this case we are trying from multiple clients to upload content to the same Azure Storage Account

Solution 1: Retry later

The most simple solution, because you don't know exactly what is the problem, is to wait for a specific time period and retry. It is important to not have the same time period used by all clients. 

This can be handled easily if each client generates a random retry time (between two ranges). 

Solution 2: Request another upload storage location 

The second solution will need to involve the backend. When a 500 or 503 error code is returned, the client could retry 2-3 times to see if the upload can be done. If we still receive this error after multiple retries we should go to the backend and request another location where we can upload content. The backed should have multiple storage accounts and assign another location where we should retry the upload.

Solution 3: Assign upload location from multiple Storage Accounts

When clients request a storage location for upload, the backend should manage multiple storage account and play the role of balancer. Assign and use different Storage Accounts for different upload requests.

Solution 4: Provide two or more upload locations to clients

When clients receive an upload location, they should receive more than one upload storage location. If we encounter issues with the first location, the second one should be used and so on. Similar with Storage Account and Primary and Secondary location.

Problem 2: During download of a blob storage content we hit a limit

In this case we have multiple clients that are trying to download content from the same Azure Storage Account.

Solution 1: Replicate content on multiple Storage Accounts and share the location in Round Robin manner

First of all we should replicate the content in multiple Storage Accounts. Once this is done, each time when a client request a storage location we should provide a location from a different Storage Account. In this way we should be able to distribute the load on multiple Storage Accounts.

Solution 2: Replicate content on multiple Storage Accounts and provide a list of download location

It is similar with previous solution. The only thing that is different is that each client receives multiple locations from where we can download the content. If we receive errors when download is made from the first location, he should try to use the second, the third one and so on.

Solution 3: Request another download location from backend

In this case, the backed has multiple locations where content is replicated. Each client receives one location from where we can download content. If there are kind of errors after multiple retries for the same location, he will request from the backend another download location .

Of course all this solutions can be combined. I recommend to start with the simplest one and add more logic when you need it. All the time, you should send (or prepare to send) a second location that can be used for Upload or Download.