Add MapReduce equivalent (low cost intensive parallelized CPU)
It would be really nice if Azure was featuring a MapReduce service to offload really intensive CPU operations.
Even better if MapReduce is made 100% .NET managed as it would allow processing power to be much more "burstable" (not need to redeploy entire VMs, only the .NET package).
HDInsight brings Hadoop processing as well the ability to execute Dryad jobs to Azure. You can find details on the Dryad work here: https://github.com/MicrosoftResearchSVC/Dryad
Fantastic... I'm one step closer to migrating from Amazon to Azure. I need a reliable Lucene.NET port that supports N instances. Great work by the way.
Mario Kosmiskas has just published an article explaining how to deploy Hadoop in Azure. You can find detailed instructions here:
Heard Avanade ported Cloud MapReduce (http://code.google.com/p/cloudmapreduce) to Azure. Cloud MapReduce runs natively on cloud services, such as queue, table and storage.
Jamie Thomson commented
I saw a C9 interview with a guy from the DryadLINQ team and he said he was moving into the Azure team to see how they could bring Dryad to Azure. Very cool indeed if it happens.
Joannes Vermorel commented
DryadLinq would be nice, but I think that even a "map/reduce" would already cover about 95% of the benefits concerning large scale functional computations.
Jurgen Van Gael commented
DryadLinq on Azure!