Is Windows Azure Storage (Blob, Table, Queue) optimized for access from Windows Azure Roles?

Question

I'm currently looking at Windows Azure to host an ElasticSearch implementation. Loading the application and running it under Java is not that difficult.

Currently, ElasticSearch only supports Amazon's S3 when it comes to cloud storage.

As a result, I've made a request to add support for Azure Blob Storage in ElasticSearch.

Right after I made the request, it occured to me that while I can host ElasticSearch in Azure, I can create an Amazon S3 account and then have the instance running in Azure connect to the S3 account for storage.

However, I do have concerns about the speed between the two. While I am sure both Azure Storage and Amazon's S3 are both optimized for really fast speeds, I have a nagging feeling that storage systems are really optimized when accessed from their respective computing clusters.

That said, is there any definitive information on this? It makes sense, but I'm looking for specific confirmation or denial.

Answer 1

It's not so much a matter of optimization of the Azure storage API for Azure Roles, but simply a matter of physical co-location and network distance / number of hops. You can (and should) specify that your Azure storage service resides in the same data center as the Azure roles that will be using that storage service.

You can expect network bandwidth to be greatest and latency to be lowest between an Azure role and an Azure storage residing in the same data center. Bandwidth will be lower and latency higher when your Azure role connects to anything outside of its own data center - be that Azure storage in another data center, or Amazon S3 storage in another data center.

Besides performance, also keep in mind that you pay for all data traffic in and out of the Azure data center for your services. Having your Azure role accessing data on Amazon S3 or in another Azure data center will take a bite out of your bandwidth quota, whereas accessing Azure storage within the same data center costs you nothing, no matter how much traffic you use between your role and your Azure storage.