Windows Azure AppFabric Caching Service Released and notes on quotas
After an amazing amount of very fruitful work, today we are announcing our offering of cache in the cloud via the release of Windows Azure AppFabric Caching Service.
The Caching service is an explicit, distributed, in-memory cache service that helps improve the performance of Windows and SQL Azure applications, by keeping distributed data in-memory and reducing the need to retrieve data from storage or database.
The most common question I get from customers is around pricing this is the MSDN page with the pricing details. Here is important to take note that beyond just selecting your required cache size, you also need to take into account the given quotas for that particular size. As you will see, each cache size has an hourly allowance for number of transactions, bandwidth and concurrent connections, this means a couple of things:
1. Some amount of calculation on transactions per hour and average size of objects will need to be understood. The transaction quota assumes an average object size of about 3.6KB so if your average objects size is larger than that then focus on your required bandwidth, as you are likely to reach the bandwidth allowance before the transaction quota.
2. Also understanding the amount of connection is essential. A common misinterpretation of this limit is to assume that this is the same as the amount of transactions your web role can handle. The number of DataCacheFactory instances along with your MaxConnectionsToServer is what will determine the amount of transactions you can accumulate. For instance, in the case of 3 web roles utilizing one datacachefactory each and with maxconnectiontoserver set to 1, there will be only 3 connection been used and they would have many transactions (get/put) going over them – we will not hit any quota limits (the smallest quota limit is 5). To get further understanding on the relation of DataCacheFactory, MaxConnectionsToServer and the amount of connections, refer to the blog on peeking into client & server WCF communication.
You may have noticed that the blog referenced above is for Windows AppFabric Cache, this is because the engine under both technologies is fairly similar hence the concept of DataCacheFactory can be ported. Find more on the difference between these two technologies