What are my options for migrating to a larger Azure Redis Cache instance

Question

I am working on this project that needs a high throughput Redis cache. Since it is kind of expensive to run a Redis cache (This is a hobby project) I am starting with a 1GB node. However if this pans off how to I change to a large instance of the cache without having a major outage to create a new instance, update connection strings etc?

I can't seem to find any way of either migrating of changing service tier. In fact it specifically says that "plan can not be changed" when I go to that page.

Are there any other options for me once I outgrow my initial cache instance size? This is very "uncloudy" behavior is it not? All my other Azure resources I can resize as I need to at a moments notice.

Answer 1

According to this talk (around the 26m mark), this type of functionality is planned for next year.

Shameless plug: AFAIK the only Redis-as-a-Service provider that allows changing your plan in realtime, without loosing your data or having your service interrupted, is Redis Cloud by Redis Labs (also available on Azure).

Answer 2

For now, you need to create a new cache and update the connection string in your application.

The feature to scale an existing cache up or down is coming soon. Once it is enabled, for scale up there will be no data loss and a very brief availability loss. For scale down, there may be data loss if the cached data is too large for the new size.

Answer 3

You can also take a look at Sharding Redis Instances. I had run into a similar problem, where my data in Cache had started blooming unexpectedly, I had used the concept of Sharding to tackle the issue. Every time the data size would cross a certain threshold in a Redis instance, a new instance of Redis would be created and data would be shared among the new instances. The problem would be to re-distribute the keys and also detecting the correct Redis instance when searching for a key. There are various algorithm to solve such a problem. I had used the concept of Consistent Hashing to solve it.

Since you wouldn't get any out-of-the-box support from Azure Redis Cache for Sharding, you might have to write your Sharding logic in the client side or on the proxy.

You might want to take a look at the implementation in Craiglist - http://blog.zawodny.com/2011/02/26/redis-sharding-at-craigslist/ Take a look at Twenproxy, you might get some more context - http://antirez.com/news/44

Answer 4

Just an update. This feature is now available in Azure Redis Cache (premium tier). Take a look at this msdn link.