Click here to Skip to main content
11,581,510 members (60,197 online)
Click here to Skip to main content

Using Azure Lease Blob

, 9 Jun 2013 CPOL 11.3K 67 4
Rate this:
Please Sign up or sign in to vote.
This article describes how the Azure Lease Blob can help the composition of the business model during the runtime in the distributed event-driven pub/sub architecture.

 

 

Contents

 

Features

  • Push Model
  • Distributed Event-Stream Processing
  • Based on the Azure Lease Blob
  • Loosely decouple pattern
  • Ready for runtime business composition, troubleshooting, monitoring, etc.
  • No Database

 

 

Introduction

Windows Azure Storage Client Library 2.0 (packaged inside the Windows Azure for .NET SDK 2.0) introduces new features for expanded Azure Storage support of the 2012-02-12 REST API version with implementation for Blob & Container Leases. This new implementation features enable to design an event-driven distributed architecture, where the business processors can establish and manage a lock on Azure Storage Blob & Container for write and delete operations. This article describes how these storage features can be utilized in the modern metadata driven architecture based on the WF 4.X declarative programming.

Let's start with an example of a simple processing, where a state-full message is processing in the pipeline of the three business processors such as pre-processor, processor and post-processor, see the following picture:

 

 

As the above picture shows, the message is processing in the pipeline of the state-less services (business processors), where each processor can update a message. In other words, the message can hold a state of the business processors. The process is state-less and the business context is processed in the sequential manner.

Let's assume, the incoming message includes some multiple parts, that can be processed in the parallel manner. The following picture shows this example:

As the above picture shows, the pre-processor is a splitter of the message and the post-processor is an aggregator of the message. The Aggregator has a responsibility to create the final message based on the states stored in the database during the processing of each message via processors. This solution requires a database for storing a specific business context state. 

The business context complexity can grow, for instance: the business processor can be a nested splitter, etc. The following picture shows this example:

Basically as we can see, the services and messages are state-less, however, we have a database with application specific schema to hold a business context state. The business context state is persisted behind the Post-Processor in the Aggregator, which is also a state-less service with sharing data in the database. Of course, we can create a model where each processor will have responsibility to publish a business context state via its processing. This scenario requires a specific implementation of the database service where a business context state is stored.

Based on the business context complexity and requirements for throughput and small latency, we can use (instead of the Aggregator) the Microsoft StreamInsight Technology for Complex Event Processing (CEP), see StreamInsight Service for Windows Azure. In this scenario, the business processors are the sources for complex event processing to generate action such as aggregated message. I do recommend to look at  StreamInsight Service for Windows Azure and Testing the StreamInsight Service for Windows Azure for detail description of event processing on the Cloud solution..

Another approach for holding a business context state in the event-driven distributed architecture is the usage of Azure Storage Lease Blob. The following picture shows our example with Lease Blob:

Basically, the Lease Blob is a Blob in the Azure Storage with exclusive write operation. The content of the blob is a collection of the Events generated from the sources within the Business context. Instead of the one central event engine processor (like it is used in the CEP), this scenario is based on the distributed event-processing, where each service will process a business context state on the Event-Stream stored in the Lease Blob.

Let's assume we have three business processors 1,2 and N like it is shown in the above picture. They were created in the pre-processor splitter. Each service has responsibility to update a current Event in the Event-Stream and initialize the next Event(s) if the process continues. Based on that, the Pre-Processor/Splitter will create an Event-Stream (if it doesn't exist already), updating/creating Event 0 and initializing three Events such as Event 1, 2 and N. When the processor is processing, its event in the Event-Stream is updated and creating the next Event, and so on. As the above picture shows, the processor 2 was the last processor for splitter and therefore its responsibility is to create a message for Post-Postprocessor and writing an Event N+1 into the Event-Stream. Finally, the Post-Processor will update its event in the Event-Stream and generating the finally aggregated message.

The Event-Stream stored in the Lease Blob represents a runtime behavior of the business context. This stream can be used during the runtime for troubleshooting, monitoring and also for watchdog processing, where each delta business process can be under the watchdog timer. In other words, if something happens in the distributed process such as lose connectivity, etc., the Event-Stream knows the details for the process recovery. Another feature of the Event-Stream is to allow aborting the running business process based on the user request.

Thanks to Lease Blob feature in the Azure Storage, which simplified the implementation of the shared resource in the Azure Storage with exclusive write in the distributed model, similar like to lock resource in the application domain. This feature is built-in the blob service, therefore its consumer needs to acquire lease for this specific blob for its operation such as write, delete, etc.

OK, let's continue with the concept and design of the Using Azure Lease Blob. I am assuming you have a working knowledge of the Windows Azure Platform and WF Technologies.

 

Concept and Design

The Concept of using Azure Lease Blob in the distributed event-driven architecture is based on acquiring an exclusive ownership for write or delete lease blob in the Azure Storage. Basically, the event-driven distributed process is executing in parallel, where business processes are running concurrently within the business context. To control this business context such as aggregation of messages, triggering events based on the context state, etc. we need to have a small knowledge base for these business process states. This "knowledge base" is a runtime log of the events into the lease blob and this content is called the Event-Stream.    

The following code snippet shows a generic Event object stored in the Event-Stream:

[XmlRoot(ElementName = "Event")]
public class EventP
{
  [XmlAttribute(AttributeName = "name")]
  public string Name { get; set; }

  [XmlAttribute(AttributeName = "status")]
  public string Status { get; set; }

  [XmlAttribute(AttributeName = "id")]
  public string Id { get; set; }

  [XmlAttribute(AttributeName = "topic")]
  public string Topic { get; set; }

  [XmlAttribute(AttributeName = "key")]
  public string Key { get; set; }

  [XmlAttribute(AttributeName = "msg")]
  public string Msg { get; set; }

  [XmlAttribute(AttributeName = "trackingId")]
  public string TrackingId { get; set; }

  [XmlAttribute(AttributeName = "created")]
  public string Created { get; set; }

  [XmlAttribute(AttributeName = "timestamp")]
  public string Timestamp { get; set; }

  [XmlAttribute(AttributeName = "timeout")]
  public string Timeout { get; set; }

  [XmlAttribute(AttributeName = "ref")]
  public string Ref { get; set; }

  [XmlAttribute(AttributeName = "tag")]
  public string Tag { get; set; }

  [XmlAttribute(AttributeName = "prev")]
  public string Prev { get; set; }

  [XmlAttribute(AttributeName = "next")]
  public string Next { get; set; }
}

As you can see, the above Event object (element) has many attributes to capture a business state in the specific place of the business processing. These attributes allow to reference a message, walking through Event-Stream, holding watchdog timer, topic, key, etc.

Well, from the abstraction point of the view, the event-driven distributed business model can publish events during the processing like it is shown in the following picture:

 

The Events are stored in the blob in order how they are published by business processors, for instance, processor P0 at the time T1, P1/T2, etc. In the concurrent business processing, where more business processes are run concurrently, there is no exact sequence of the events guaranteed, in other words, two or more publishers want to store their events at the same time. To avoid this write collision at the blob storage, each publisher needs an exclusive access to the blob storage. That is a great feature of the Azure Storage Blob Service, such as Lease Blob.  

Azure Lease Blob

From the MSDN documentation Lease Blob:

The Lease Blob operation establishes and manages a lock on a blob for write and delete operations. The lock duration can be 15 to 60 seconds, or can be infinite.

The Lease Blob operation can be called in one of five modes:
Acquire, to request a new lease.
Renew, to renew an existing lease.
Change, to change the ID of an existing lease.
Release, to free the lease if it is no longer needed so that another client may immediately acquire a lease against the blob.
Break, to end the lease but ensure that another client cannot acquire a new lease until the current lease period has expired.



Ok, back to our abstraction model. We know, that the Event-Driven distributed business model will publish (store) Events into the Lease Blob one by one. This collection of the Events is called Event-Stream and it represents a runtime behavior of the business model processing. Basically, each incoming message (request) into the business model can have one Event-Stream for capturing states within its business context. Based on this Event-Stream, the business process can be controlled, monitored, recovered, troubleshooted, etc.  

Each publisher of the business processor in the Event-Driven distributed model must have an exclusive access for updating Event-Stream in the Blob Storage. To get this write exclusivity, the publisher needs to acquire blob for lease (locking) purpose before this operation. The write/delete exclusivity blob access operation is represented by LeaseId which the client (publisher) must include in the blob operation. The lease is granted for the duration specified when the lease is acquired, which can be between 15 seconds and one minute, or an infinite duration. Of course, we don't want to block other publishers to publish their Event into the Event-Stream, therefore each publisher has responsibility to release ASAP active lease.

The following picture shows this Lease BLOB Scope feature in the publisher:

 

As the above picture shows, the Lease Blob Scope begins calling a blob request for Acquire Lease. This request is repeated in the polling cycle until the LeaseId is obtained by Azure Blob Service. That is the way (such as polling) to acquire lease in the distributed internet connected model. Once the scope has an exclusive write access to the blob, we can get an Event-Stream to the memory, make necessary processing and writing back (Put) to the blob. After that, the scope must be released calling Release Lease operation.

Basically, the scope operation takes a very short time (couple hundreds milliseconds), including the Get/Put operations, therefore locking the Event-Stream in the distributed model is not critical, but in the case of crashing the active scope, all other clients (publishers) must wait for releasing Lease Blob by Azure Blob Service, which can take maximum of 60 seconds.     

 As I mentioned earlier, the LeaseBlobScope will have an exclusive access for Get/Put Event-Stream stored in the Lease Blob. This process will take very little processing time. It is executed in the memory, therefore after releasing scope we have an updated Event-Stream in the memory and it can be used for other actions.

The following picture shows this scenario:

As the above picture shows, the LeaseBlobScope needs to update an Event-Stream stored in the Lease Blob. Basically, this operation consists of the Get Event-Stream, Select CurrentEvent from the Event-Stream, Updating this CurrentEvent, Updating Status in the Event-Stream and the last optional step is to add a NewEvent if we need it. Once the Event-Stream (located in the memory) has been updated, we can store it back to the Lease Blob on the Azure Storage. The lease Blob is automatically released during the exiting a LeaseBlobScope sequence, therefore it is availaible for another acquire request.

Next sequence after the LeaseBlobScope can be used for more detail processing of the Event-Stream, for instance, triggering an action like it is shown in the above picture. Note, that the Event-Stream is holding all states within the business context, therefore the Event-Stream status can be scanned for last business parallel processor to fire a message for next action.      

To simplify handling a Lease Blob in the Windows Azure Storage within the WF Technology, the custom LeaseBlobScope activity can help. The following picture shows this custom activity which implementation is included in this article:

And the following pictures show few examples of assigning expression editor for handling Event-Stream using custom class extensions: 

More details about the custom activities and Event-Stream Library can be found in the Implementation section. 

 

The Event-Stream in the Lease Blob can be shared by any enterprise component that has an authority access to the Lease Blob. The components can use an Event-Stream for dashboard features (showing a progress status, etc.) and/or for controlling process such as user cancel, notification, etc. The usage of these features is fully transparent and loosely decouple from the business model. 

The following picture shows additional two components connected to the Event-Stream such as User and Watchdog:

 

Based on the user request, the User publisher can publish an event for canceling a running background business process. In this scenario, the Event-Stream is updated for cancel status and the business processes will take an action during the Event-Stream processing.

The second component from the above picture is a Watchdog service. This service has a responsibility to check an expiration time for Event-Stream. If for some reason, the business process is not executed within the specific expiration time, the Watchdog publisher will abort this Event or/and Event-Stream.

 

Event-Stream

The Event-Stream is a collection of the Events published by sources such as clients, business processors, etc. within the business context of the distributed model. Each Event represents a state of the business model. The business model can be composed during the runtime sequentially and/or concurrently where more business processors run in parallel. For better structuring and query events in the Event-Stream, a formatted id value is used with delimiter character '.' to represent a new group (or thread) in the Event-Stream, for instance: 01.00

The following pictures show an example of the SearchOrder Event-Stream for few business processing steps:

Step 1. This is an initialize step, where OrderStart processor is creating an Event-Stream and next 3 Events for further processing such as SearchStart. This is a splitter scenario to run searches in parallel manner.

As you can see, the Event-Stream has a root Event (name=OrderStart, id=00, status=inprocess). The OrderStart generates 3 messages for SearchStart processors. The status of the Event-Stream for this step is inprocess-init-init-init.

Step 2. In this step, the SearchStart process is invoked and in the Event-Stream the Event SearchStart is updated and creating a new Event for SearchDone. The following picture shows this scenario:

As you can see, the above picture shows more runtime business composition, where a second thread 02.00 is processing faster. This thread has a SearchDone and has available 3 Results. Note, the thread 03.00 failed and has been aborted, which is the finally state for the Event.

Step 3. In this step, we have more business composition such as completion, aborted and next nested search processing based on analyzing of the received Results:

Note, that Event is finalized in the case of completed, aborted or canceled. As the above picture shows, the Event-Stream is waiting for finalizing the business processes in the business threads 01.02.00, 01.04.00 and 02.03.00.

The Event-Stream represents a small knowledge base of the business model composition during the runtime. Based on the Events, we can see performance for each business processors, message flow, data, etc. If the process crashes, the business process can be restarted and continue from the crash point.  

The Event-Stream is stored in the Azure Storage Lease Blob in the xml formatted text.

 

OK, let's look at the first of the custom activities for handling a Lease Blob within the xaml workflow and custom EventStream library. I am assuming you have a working knowledge of the Windows Azure Platform and WF Technologies. 

Implementation

Adding the following custom activities into the Workflow Designer Toolbox allows us to use the Lease Blob features within the workflow orchestration.

As you can see, the above custom Activity library contains 5 basic Lease Blob operations and one combined operation for LeaseBlobScope. The LeaseBlobScope custom native activity allows to execute a sequence of activities with exclusive write/delete Lease Blob operations. This custom scope activity has a responsibility to release a Lease Blob on the regular or fault exit.

Let's look at the LeaseBlobScope custom activity from its implementation point of the view. 

 

LeaseBlobScope custom activity

The following picture shows design and properties of the LeaseBlobScope custom activity.

The LeaseBlobScope requires few properties for creating a Lease Blob such as BlobAddress, ConfigurationName for account, etc.

What is important for LeaseBlobScope custom activity is its polling of the Acquire request for Lease Blob in the case of receiving a collision error 409 from the Azure Blob Service. Note, that the infinity polling depends on the number of clients acquired the same Lease Blob and from the quick release of the active Lease Blob. In the distributed internet connectivity that's the way how we can get an exclusivity access for writing/deleting a resource on the Azure Storage Blob.   

The following code snippet shows an implementation of the Exceute methode of the LeaseBlobScope custom activity:

protected override void Execute(NativeActivityContext context)
{
  if (this.activities != null && this.Activities.Count > 0)
  {
    NoPersistHandle handle = this.noPersistHandle.Get(context);
    handle.Enter(context);

    CloudBlobClient client = StorageAccountHelper.GetCloudStorageAccount(this.ConfigurationName.Get(context)).CreateCloudBlobClient();
    CloudBlockBlob blob = client.GetBlockBlobReference(this.BlobAddress.Get(context));

    // lock the blob
    Func<CloudBlockBlob, TimeSpan, string, string> delegateFunc = new Func<CloudBlockBlob, TimeSpan, string, string>(LockingBlob);
    var ar =  delegateFunc.BeginInvoke(blob, this.LeaseTime.Get(context), this.ProposedLeaseId.Get(context), null, null);
    ar.AsyncWaitHandle.WaitOne();
    string lid = (string)delegateFunc.EndInvoke(ar);

    this.LeaseId.Set(context, lid);
    this.Result.Set(context, blob);

    // safety work on this blob
    context.ScheduleActivity(this.Activities[0], this.onChildComplete, OnFaulted);
  }
}

As you can see, before running the first activity in the scope, we need to acquire a Lease Blob. This request is done asynchronously with delegate function LockingBlob:

private string LockingBlob(CloudBlockBlob blob, TimeSpan timeout, string proposedLeaseId)
{
 int ii = 0;
 DateTime startDT = DateTime.Now;

 while (++ii > 0)
 {
   string lid = blob.TryAcquireLease(timeout, proposedLeaseId);
   if (string.IsNullOrEmpty(lid))
   {
     var delayTime = TimeSpan.FromMilliseconds(new Random(Guid.NewGuid().GetHashCode()).Next(750, 2500));
     Thread.Sleep(delayTime);
   }
   else
   {
     Trace.WriteLine(string.Format("TryAcquireLease[{0}, {1}ms]", ii, DateTime.Now - startDT));
     return lid;
   }
 }
 return null;
}

The LockingBlob function has only one responsibility such as getting a leaseId from the acquired Lease Blob. In the case of the null value, it will wait for a random time and then trying again to acquire a Lease Blob. Note, that the Thread.Sleep is sleeping within the delegated thread, which is different from the workflow thread. 

The following code snippet shows a handler for Fault within the activity scope. You can see how the Lease Blob is released:

private void OnFaulted(NativeActivityFaultContext context, Exception exception, ActivityInstance faultedInstance)
{
  this.Result.Get(context).ReleaseLease(AccessCondition.GenerateLeaseCondition(this.LeaseId.Get(context)));
  NoPersistHandle handle = this.noPersistHandle.Get(context);
  handle.Exit(context);
}

 

AcquireLeaseBlob custom activity

AcquireLeaseBlob custom activity can be used to simply acquire Lease Blob:

 

This is an async custom activity with only one request to acquire Lease Blob. In the case of the collision error (code 409), we can sleep there for the specified time and then exit it. The following is a code snippet for its implementation:

protected override IAsyncResult BeginExecute(AsyncCodeActivityContext context, AsyncCallback callback, object state)
{
  CloudBlobClient client = StorageAccountHelper.GetCloudStorageAccount(this.ConfigurationName.Get(context)).CreateCloudBlobClient();
  CloudBlockBlob blob = client.GetBlockBlobReference(this.BlobAddress.Get(context));
  context.UserState = blob;
  return blob.BeginAcquireLease(this.LeaseTime.Get(context), this.ProposedLeaseId.Get(context), callback, state);
}

protected override string EndExecute(AsyncCodeActivityContext context, IAsyncResult result)
{
  CloudBlockBlob blob = (CloudBlockBlob)context.UserState;
  try
  {
    return blob.EndAcquireLease(result);
  }
  catch (StorageException ex)
  {
    if (ex.RequestInformation.HttpStatusCode == 409)
    {
        Trace.WriteLine(string.Format("AcquireLease conflict - sleeping for {0}]", this.RetryTime.Get(context)));
        Thread.Sleep(this.RetryTime.Get(context));
        return null;
    }
    else
        throw;
  }
}

As you can see, this is a straightforward async implementation of the Begin/End methods. For this custom activity, the workflow orchestration will take care of the next polling for Lease Blob (using While loop, Retry time, etc.). Note, that the workflow has a responsibility to release a Lease Blob using a basic custom activity ReleaseLeaseBlob or library function.

As I mentioned earlier, the Event-Stream is a collection of the xml elements (EventP object). The Event-Stream is stored in the Lease Blob in xml formatted text. For handling an Event-Stream using an Assign activity within the xaml workflow, the following Library has been implemented.

EventStream Library

This is very lightweight library for basic manipulation with Events in the Event-Stream such as their updating event property, checking status, etc. The following picture shows a class view of Event object (EventP), container of the Events (EventStreamProcess) with useful methods and extension class for EventStreamProcess.

The design concept of the EventStreamLibrary is based on using the methods and properties within the Expression text, for their chaining in the Assign activity. For instance: adding the events into the Event-Stream and selecting specific Event from the Event-Stream can be done within one Assign activity:

 

 

Note, for selecting a specific Event in the Event-Stream we can also used Select or MustSelect methods based on the xpath expression text:

 

Once we have a specific event for the business state from the Event-Stream (esp), we can make its update within one Assign activity, see the following example:

The following code snippet shows an implementation of the Update method from the EventStreamProcessExtension class:

public static EventStreamProcess Update(this EventStreamProcess esp, XElement element, string name, string value)
{
  if (element == null || element.Name.LocalName != "Event")
      throw new ArgumentException("element");

  element.SetAttributeValue(name, value);
  element.SetAttributeValue("timestamp", DateTime.UtcNow.ToString());

  XElement evnt = element;
  if (name == "status" && (value == "completed" || value == "aborted" || value == "canceled"))
  {
    while (esp.IsRoot(evnt) == false)
    {
      Trace.WriteLine("EventStream.UpdateStatus: {0}", evnt.ToString());

      evnt = esp.EventStream.XPathSelectElement(string.Format("./Event[@id='{0}' and @status='inprocess']", evnt.Attribute("prev").Value));
      if (evnt == null)
        break;

      List<string> next = evnt.Attribute("next").Value.Split('|').ToList().Select(s => s.Trim()).ToList();
      if (next.Count == 1 || esp.IsDone(next))
      {
        if (element.Attribute("status") != null && element.Attribute("status").Value != "completed")
          evnt.SetAttributeValue("msg", "WithError");
          
        evnt.SetAttributeValue("status", "completed");
        evnt.SetAttributeValue("timestamp", DateTime.UtcNow.ToString());
      }
      else
      {
        break;
      }
    }
     
    // update root status
    if (esp.IsRoot(evnt))
    {
      List<string> next = evnt.Attribute("next").Value.Split('|').ToList().Select(s => s.Trim()).ToList();
      if (next.Count == 1 || esp.IsDone(next))
      {
        if (element.Attribute("status") != null && element.Attribute("status").Value != "completed")
            evnt.SetAttributeValue("msg", "WithError");
        evnt.SetAttributeValue("status", "completed");
        evnt.SetAttributeValue("timestamp", DateTime.UtcNow.ToString());
      }
    }   
  }
  return esp;
}     

As you can see, there is one special case in the Update method such as updating a status property. In this case, this method will update status all Events within the EventThread from the specific Event up to the root Event.

The Event status can have one of the following values, where the completed, aborted and canceled are the final statuses of the Event object in the Event-Stream.

 

For testing finally status of the Event-Stream, we can use a IsDone method like is shows in the following example:

and the following code snippet shows how the method IsDone implemented in the EventStreamProcess class:

public bool IsDone()
{
    var root = this.EventStream.Elements("Event").ElementAt(0);
    if (root.Attribute("prev") != null)
        throw new ArgumentException("Internal error: The root event must not have attribute 'prev'");

    return root.Attribute("status").Value == "completed" || root.Attribute("status").Value == "aborted" || root.Attribute("status").Value == "canceled";
}

 

CloudBlockBlobExtensions

This is an extension to CloudBlockBlob class for custom handling a Lease Blob. The following picture shows all extension methods:

Implementation of the above extension class is very straightforward, for example: the following code snippet shows how we can retrieve a blob to the XElement return value:

public static XElement ToXml(this CloudBlockBlob blob, string leaseId)
{
    using (var ms = new MemoryStream())
    {
        blob.DownloadToStream(ms, AccessCondition.GenerateLeaseCondition(leaseId));
        ms.Position = 0;
        return XElement.Load(ms);
    }           
}

The usage of this extension in the xaml workflow declaration is by Assign activity, see the following example for Loading and Saving  EventStreamProcess object:

 

 

Note, that the EventStream Library is for simplified handling a collection of the Events in the Event-Stream. Basically, you can use any declaratively programming technique for get, update and put Event-Stream to/from Lease Blob, for instance LINQ. Keep on mind, that once you are in the LeaseBlobScope, the business processing must be done ASAP because you Lease Blob for your exclusive write, in other word, the LeaseBlobScope is locking other clients in the place, where they are asking for the same Lease Blob for acquire lease operation.

Ok, let's look at some test cases of the Lease Blob using these custom activities and library.

 

Usage and Test

First of all, the following are prerequisites:

The AzureLeaseBlob solution includes 2 projects and one Test solution folder where assemblies and samples for our testing are stored. The following picture shows a solution after its downloading from this article:

 

 

As you can see, the solution includes a WorkflowDesignerTester application, which is a re-hosted Workflow Designer on the Windows Form, see more details about this little tool in my WF4 Custom activities for message mediation article.  Another tool what we need to have for our testing purposes is Azure Storage Explorer. Basically, you can use any of your favorite explorer for Azure Storage Blob.

Before our first test, we need to create one empty blob in your account. The following picture shows this public BlockBlob resource such as temp/eventstream.

 

Let's start with very simple basic example.

 

Test A. - Basic example

This is a basic test of the LeaseBlobScope custom activity in the infinity loop. Please follow these instruction steps:

Step 1. Lunch the WorkflowDesignerTester45 program

We will use the WD 4.5 version for demonstration of the AzureLeaseBlob custom activities. The following screen snippet is the WD after opening a file LeaseBlobScopeBasic_Test45.xaml from the solution Samples folder.

 

As you can see, there is (on the left side) a workflow document outline view panel and on the toolbox side a custom library AzureLeaseBlob.ActivityLibrary. Our Test Sequence is in the designer workspace. This sequence is shown in the following screen snippet:

That's all for the test loop. DoWhile loop is looping a LeaseBlobScope custom activity with a random delay (time between 750-2000 milliseconds). The delay expression is

TimeSpan.FromMilliseconds(new Random(Guid.NewGuid().GetHashCode()).Next(750, 2000))

Inside the scope is one activity such as WriteLine to display a LeaseID on the console screen.

 

Step 2. Azure Storage account

Before running this test, we need to populate an azure storage account in the applicationStorageAccount variable. In the case of using another container Blob for our test, please also change the value of the EventStreamRef variable. The following picture is a screen snippet of the variables used in our test:

 

Step 3. Run

Pressing the button Run, the console self-hosted process is created and loaded the xaml document for its execution. Our test requires to lunch two processes to see how the LeaseBlobScope are handling a shareable Lease Blob in concurrent manner. 

Please, press the Run button twice.

 

The following two console programs are lunched and displaying LeaseIDs.

While executing the LeaseBlobScope custom activity, the scope has an exclusive access for writing to the Lease Bob (temp/eventstream). This Lease Blob status is represented by displaying its LeaseID. Note, that the other LeaseBlobScope is acquiring Lease Blob in the pooling manner.

To see an acquire delay for Lease Blob, the following utility DebugView can be used. The log messages shows a total acquire pooling time and counter of pooling.

As you can see, the above picture shows a 3 cases when the LeaseBlobScope needs to repeat the acquire request to the Lease Blob.

That's the situation when we have two concurrently running processes. Now we can run more processes, for instance 5 and look at the tracer for acquire pooling values.

You can play with workflow process, adding more activities within the scope such as writing some contents to the blob, etc.

Ok, let's make more advanced test, where LeaseBlobScope will work with a Lease Blob body.

 

Test B. Advanced Test with Lease Blob

This test demonstrates a LeaseBlobScope and EventStreamLibrary for concurrently handling an EventStream within the scope. The test case within the loop is to find the Event object with status=init in the EventStream and updated to status=complete. The loop is finished when all Events in the EventStream are updated to status=complete. To repeat the test, the blob must be deleted in the container.

First of all, we need to open a new test file.

Step 1. Open the LeaseBlobScope_v45.xaml file

Click on the File/Open to select a workflow document LeaseBlobScope_v45.xaml file from the Sample solution folder. The following picture shows a Test Sequence of this loop:

Basically, the Test Sequence has two parts such as DoWhile loop and Init. The Init part has responsibility to create an EventStream with specific number of Events (default value is 10) with status=init. The DoWhile loop is also simple and straightforward.

The first activity is to get the EventStream from the Lease Blob

esp = new EventStreamProcess(blob.ToXml(leaseId))

then selecting an Event object with status=init

curEvent = esp.Select("./Event[@status='init']")

next step is to update this curEvent

esp = esp.Update(curEvent, "key", workerID).Update(curEvent, "status", "completed")

and last step in this sequence is to store updated EventsStream back to the Lease Blob

esp.EventStream = blob.FromXml(esp.EventStream, leaseId)

The following screen snippet shows this sequence:

Note, that the DoWhile condition is very simple expression using an EventStreamLibrary such as esp.IsDone=false

Before we start our test, we have to setup Azure Storage account.

Step 2. Azure Storage account

The following highlighted variables must be changed or modified:

Variable applicationStorageAccount must be setup for your Azure Storage account. The other variables are optional such as EventStreamRef and maxEvents. Note, that the container temp has been created manually in our first Test. Basically, you can create another one for this test. It is based on your needs. The number of Events, in the EventStream is setup for 10, but it can be maximum of 99. It is a string value.

 

Step 3. Creating EventStream

This step is for creating a Lease Blob which contains of EventStream, therefore press the button Run to lunch the process. The following screen snippet shows a result of the init sequence:

As you can see, we have an EventStream with 10 Event objects ready for our test. If we press the Enter key on the keyboard, the DoWhile loop will start and only this process will be updated EventStream which we don't want. Our test must be proven by more than one running processes, therefore please follow the next step.

Step 4. Run Test Loop

In this step we need to perform two actions such as pressing the button Run on the WorkflowDesigner to create the second process and back to first one to press Enter key on the keyboard. These actions must be done quickly.

The following picture shows two running processes:

As you can see, each process updated some Event objects in the EventStream in the concurrently manner without any collision for sharing a Lease Blob. 

Once the EventStream is updated for all Events, our test is finished. The following picture shows this situation:

 

You can see the property key in the Event object who was the exclusive owner of the Lease Blob at the moment.

The above test can be repeated again after the Lease Blob is deleted, for instance using a 3rd party tool such as Azure Storage Explorer.

The test can be repeated for more Events and/or running processes to see how the LeaseBlobScope works with a Lease Blob.

That's is all for our testing, of course you can continue with your needs for your application specific  EventStream.

 

Conclusion

This article described using an Azure Storage Leas Blob for compositions of the event-driven distributed business processes. The concept is based on the Event-Streaming during a runtime business composition, where each business processor (worker) published its state (Event) to the shareable resource called as Event-Stream for specific business context. In the distributed internet connectivity model, all publishers and consumers of the Event-Stream are loosely decoupled. For this scenario, the Lease Blob is the best solution for storing an Event-Stream with an exclusive write operation. Article includes the custom activities for Azure Storage Lease Blob and EventStream Library for handling a Lease Blob. The described concept of the using Event-Stream for runtime business process composition can help your application with scalability and parallelism in the distributed cloud-driven architecture. I hope you enjoyed it.

 

References:

[1] Windows Azure

[2] Windows Azure SDK 2.0 for .Net

[3] What's New in Storage Client Library for .NET (version 2.0)

[4] Introducing Windows Azure Storage Client Library 2.0 for .NET and Windows Runtime

[5] Lease Blob

[6] New Blob Lease Features: Infinite Leases, Smaller Lease Times, and More

[7] StreamInsight Service for Windows Azure

[8] StreamInsight in Azure

[9] Patterns and Best Practices for Enterprise Integration

[10] What's New in Windows Azure

[11] Azure Storage Explorer

License

This article, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Share

About the Author

Roman Kiss
Software Developer (Senior)
United States United States
No Biography provided

You may also be interested in...

Comments and Discussions

 
-- There are no messages in this forum --
| Advertise | Privacy | Terms of Use | Mobile
Web04 | 2.8.150603.1 | Last Updated 9 Jun 2013
Article Copyright 2013 by Roman Kiss
Everything else Copyright © CodeProject, 1999-2015
Layout: fixed | fluid