In order to improve performance on Azure table operations it is advised to batch them.

However the documentation does not clearly state that the batch must all have the same PartionKey and may only contain 100 items.

Here is how to generically query Azure Table Storage without having to define the entity that is in the table (which you may not know and can vary).

CloudStorageAccount storageAccount = CloudStorageAccount.Parse(diagnostics);
CloudTableClient tableClient = storageAccount.CreateCloudTableClient();
CloudTable cloudTable = tableClient.GetTableReference(“WADLogsTable”);
TableQuery query = new TableQuery();

query.FilterString = filter;
var items = cloudTable.ExecuteQuery(query).ToList();

See my previous post for how to set the filters.

http://msdn.microsoft.com/en-us/library/windowsazure/ff683669.aspx

String     PartitionKey eq ‘myPK’
DateTime     Timestamp eq datetime’2010-08-13T10:32:38.

Here is a great article on Azure Table Storage performance.

Short version always access via the Partition Key unless you want a table scan.

You can however search over a range of Partion Keys.

The most efficient version is Partition Key + Row Key

Azure SDK 2.0 has been released.
Here are the the release notes.

Breaking Changes:

1. Windows Azure Diagnostics no longer requires 1.7 storage
2. .NET Framework minimum version is now 4.0
3. ServiceRuntime, Configuration and Caching assemblies are now built against the .NET Framework 4.0 runtime. So upgrade all apps to target .net 4.0
4. Windows Azure Connect– The Windows Azure Connect preview is being retired June 30.
5. Hosted Web Core Support- Windows Azure web roles now require a <Site> element in the service definition file (CSDEF) for your role.
6. CSUpload.exe warning- CSUpload.exe now emits a warning recommending that you use Windows Azure PowerShell cmdlets to upload VHDs to Windows Azure.
7. Service Bus client library – the Message Buffer feature, and all APIs related to MessageBuffer have been removed. You should instead use Service Bus queues.

Here are the actual code breaking changes.

The following demonstates how to delete from WADLogsTable from before a given date.
It may not be the most efficient code but it does work.

This has been tested against Azure SDK 2.0

///

///
///
public void TruncateDiagnostics(CloudStorageAccount storageAccount, DateTime keepThreshold)
{
try
{

CloudTableClient tableClient = storageAccount.CreateCloudTableClient();

CloudTable cloudTable = tableClient.GetTableReference(“WADLogsTable”);

TableQuery query = new TableQuery();
query.FilterString = string.Format(“Timestamp lt datetime'{0:yyyy-MM-ddTHH:mm:ss}'”, keepThreshold);
var items = cloudTable.ExecuteQuery(query).ToList();

Dictionary batches = new Dictionary();
foreach (var entity in items)
{
TableOperation tableOperation = TableOperation.Delete(entity);

if (!batches.ContainsKey(entity.PartitionKey))
{
batches.Add(entity.PartitionKey, new TableBatchOperation());
}

batches[entity.PartitionKey].Add(tableOperation);
}

foreach (var batch in batches.Values)
{
cloudTable.ExecuteBatch(batch);
}

}
catch (Exception ex)
{
Trace.TraceError(string.Format(“Truncate WADLogsTable exception {0}”, ex), “Error”);
}
}

Timestamp lt datetime’2013-05-05T14:06:00′

CloudStorageAccount storageAccount = CloudStorageAccount.Parse(CloudConfigurationManager.GetSetting("Microsoft.WindowsAzure.Plugins.Diagnostics.ConnectionString"));
CloudQueueClient cloudQueueClient = storageAccount.CreateCloudQueueClient();
CloudQueue cloudQueue = cloudQueueClient.GetQueueReference("test-queue");
cloudQueue.CreateIfNotExists();

Please rememeber that queue names are quite fussy. You need to use lowecase names with hypens as seperators.
It does not like brackets or underscores.
This is especially important if you use the environment name in the queue name to allow shared storage accounts across a dev team.

Here is a great project that seems to solve one of the big problems with Azure Queues. There is a throughput limit that will prevent a queue from handling more than 500 transactions per second. The recommended solution is to shard the queues (although other solutions include moving to a vm based queing system or using an alternative queue such as RabbitMQ or ZeroMQ).