What is the maximum compute bandwidth for Azure?
What is the maximum compute bandwidth on Azure?
In Google App Engine in March 2011, the answer was determined by testing at 43. What is the equivalent number for Azure?
Test results to measure the performance of computational bandwidth in Google App Engine.
Edit I got a comment that GAE's throughput cannot reasonably be compared to Azure. If it is computed and fully parallelized, then I don't understand why not. Imagine the application is rated for a large integer x
trillion. Ask each of the n
streams to count a x/n
trillion. n
GAE threads do not result in the 1/n
-th elapsed time of one thread and it depends on how you can do it. The reciprocal is the bandwidth gain, and the maximum I achieved in GAE is 43.
43x is bitterly lacking, say 1000x, at this point I would consider porting my code to GAE. This is a feasibility study; far from pointless. I believe I have included enough information in the link for someone to write an almost identical test program. (Sorry if this is obvious.)
source to share
Since March 2011, when your GAE testing was completed, GAE has received backends where you can select different instance types .
Also, as of GAE 1.6.4, there are now threads on backend servers to help you make better use of multi-core instances.
Therefore, I believe that your computational throughput test should be repeated using these new features.
source to share
I would suggest evaluating Go or Python (the 2.7 runtime supports NumPy, which has carefully optimized the code inside it) in addition to Java in the latest runtime, and also using the backends that Peter mentioned. Anyway, I don't think App Engine or Azure is the best choice for HPC. You will get much higher throughput under numerical workloads on Amazon GPU-enabled (cg1.4large) instances if your code is capable of using that (if not, I advise you to make it possible, as GPUs are much cheaper than CPUs in numerical workloads).
source to share
2013 update on this issue, Microsoft announced big data support in late 2012: http://www.hpcinthecloud.com/hpccloud/2012-11-13/microsoft_azure_outfits_cloud_for_big_compute.html
it is now available at http://www.windowsazure.com/en-us/solutions/big-data/
This is an introductory article: http://www.windowsazure.com/en-us/manage/services/hdinsight/get-started-hdinsight/?fb=nl-nl
hdinsight is in the preview at the moment.
HDInsight is a Microsoft Hadoop-based service that provides a 100% Apache Hadoop solution for the cloud. HDInsight empowers you to get the full value of Big Data with a modern cloud data platform that manages data of any type, whether structured or unstructured, and of any size. With HDInsight, you can seamlessly store and process data of all types with Microsoft's state-of-the-art data platform that provides simplicity, ease of management, and an open Hadoop service running entirely in the cloud. You can analyze Hadoop data directly in Excel using new capabilities such as PowerPivot and Power View.
I cannot find anything about the maximum bandwidth. You may need to port your code and test it yourself.
source to share