Currently any SQL work needs to be run using TSQL, add support for using the designers to modify tables, run procedures etc.

According to http://msdn.microsoft.com/en-us/library/windowsazure/hh852669.aspx one of the backup strategies proposed is to copy the database. If that is what is proposed officially, then you can count on people writing scripts, services, worker roles, etc to back up the database. Using the equivalent of the SA to do so is just unacceptable solution. If my code is compromised then I am exposing usernames/passwords (encrypted or not). Furthermore, it's not like I can just remove that user and replace him for some other user.

We should be allowed to set specific roles to any users so they can copy the database.

Microsoft.WindowsAzure.Configuration.dll (1.7.0.1) does not have a strong name, which makes it difficult to reference in many situations.

Virtual Machine's could be provisioned by providing a URL to .OVA file or other VM file formats.

ClearDB might have provided an easy way of getting things going, but their pricing model doesn't scale with the level of website activity (to value) people have. In moving to the shared or even reserved option, it still means an additional expense/dependency to plain old MySQL hosting.

Surely this is something Microsoft could entertain in-house?

Allowing for configurable search such as that is offered wtih lucene, built into the infrastructure would be a great addition. Microsoft seems to be heading this way and has a lucene library for azure, but you cannot use it for commercial purposes. http://tiny.cc/azurelucene

The Web Site URL link on the details of an Azure VM should link to HTTPS, if there is only one endpoint enabled. Currently it only links to HTTP, which if gone unnoticed by the user, looks like the site is down when they click the link.

For example a web role coul be started when a request comes and stopped after 10 (configurable) mins of inactivity (and if the web role doesn't mind)

It seems that Azure is only for .NET and not for Java world. The auto scaling of Tomcat app using Mysql is a real challenge. For example how does tomcat know which instance of mysql it has to connect to at runtime? There are one master and multiple slave mysql instances in different VM's in cloud. Also there are multiple tomcat instances, and there is no automatic way of making a tomcat instance connect to a mysql instance. So scalability is self-managed and not automatically done?

Using TraceSource is recommended instead of using Trace. It can provide fine grained traces with better performance. However when DiagnosticMonitorTraceListener is used to copy traces to Azure storage, the SourceTrace name is appended to the trace message. Make it new column instead so it is possible to filter based on it.

It would be great if webroles started initially under then Network User profile to allow more options for advanced deployment scenarios. High level functions should be able to be performed on the file-system level and really is impossible to do this without knowing what the username is to specify write-ability by current user as a start-up function for roles.

Amazon has just released their ElastiCache service, which offers pay-as-you-use caching. Rather than charging a fixed fee each month, it would be great to use Azure caching on the same basis, paying for what we use each month

When we upgrade a new version of the application. We need to upload a file configuration too, but if i modified my configuration in administration panel, i lost it...