tables

Question

Any eta on tables support? For now I'm just holding stuff in memory. I have uses for blobs as well(haven't messed with that yet though), but having quick lookup by a key would be useful. Maybe the tables support could just use something similar to the xml to serialization functions tool that the actors library uses. Just keep the source of all my data objects as xml, with the structs as a by product. Then I can just pass the struct directly to a function in azure storage, and it route back to the non member serialization function generated by the tool for matching azure table property names to struct fields or something like that.

Is azure going to have machines with gpu's at some point? (or do they already, I'm assuming they don't as there doesn't currently seem to be much reason for them too). One really useful model with Casablanca would seem to be a worker role project in Casablanca utilizing fast auto vectorizing c++ and maybe even c++ amp on a gpu for heavy compute tasks. And then perhaps web roles in .net or something where performance isn't as critical. From personal language preference I'd rather use Casablanca across the board, but I can see the mixed model being useful.

Answer 1

Hi,

I'm sorry that it took a few days to get back to you -- I didn't pay attention to the forum over the weekend.

Regarding GPUs in Azure, I don't know that the plans currently are. The scenario that you describe is one that we definitely have our eyes on, since C++ is such a great language to use for compute-intensive workloads. The combination of Casablanca and C++ AMP seems powerful.

Regarding tables, we are working on it and will release in our next refresh (for which we are not ready to name a date), together with more complete support for blobs and queues (ACLs, leases, snapshots, etc.). Tables are more complex than blobs and queues, and require that we design good query solution, so it has taken us more time to do.

Sorry that all I have is two non-answers, :-)

Niklas