Azure - Fabrics - Web/Worker Roles
What problem does Windows Azure
solve?
The
best way to define Azure is by understanding what problem it solves. So let’s
take a typical organization that has a separate IT department maintaining an
online web application. Now let’s try to analyze which different cost factors
are involved in maintaining a web application online.
We can divide IT department cost in to four broader sections:-
• Hardware cost: - If you want to host your web application you would need servers, routers etc to run your application.
• Software licensing cost: - When you make your application you will need to purchase server OS license like windows 2003 server, visual studio licenses, SQL server licenses etc.
• Hosting cost: - You also need to deploy your web application on internet. For that you will need to host your web application on some hosting service. One of the biggest costs in hosting service is bandwidth. So we need to budget for the same.
• IT personnel salary: - Nothing can be run without humans. So we also need IT personnel’s to do development and maintenance of code and infrastructure.
In other words there is a considerable cost attached to run an IT department. Many of these costs need to be paid up front.
We can divide IT department cost in to four broader sections:-
• Hardware cost: - If you want to host your web application you would need servers, routers etc to run your application.
• Software licensing cost: - When you make your application you will need to purchase server OS license like windows 2003 server, visual studio licenses, SQL server licenses etc.
• Hosting cost: - You also need to deploy your web application on internet. For that you will need to host your web application on some hosting service. One of the biggest costs in hosting service is bandwidth. So we need to budget for the same.
• IT personnel salary: - Nothing can be run without humans. So we also need IT personnel’s to do development and maintenance of code and infrastructure.
In other words there is a considerable cost attached to run an IT department. Many of these costs need to be paid up front.
The
upfront cost can be solved if we can get a provider who can host a shared
service which provides shared server hardware, shared software license cost and
can provide pay per use bandwidth. The provider can maintain a team of IT
personnel’s who can maintain the infrastructure. If the provider services not
one but many organization using this shared service it can bring down the cost
considerably.
So in other words the model changes something as shown below. So the provider moves all the hardware and software on a centralized location and the applications of the respective organization are hosted in this shared environment. Using this model organizations do not need to pay upfront for the hardware, software and IT personnel costs.
So in other words the model changes something as shown below. So the provider moves all the hardware and software on a centralized location and the applications of the respective organization are hosted in this shared environment. Using this model organizations do not need to pay upfront for the hardware, software and IT personnel costs.
In
one line to define Windows Azure, Microsoft is the provider and windows azure is
the shared hosting service.
Microsoft has provided his own datacenters where they have hosted software's like SQL services,.NET services,Sharepoint services,Microsoft dynamic services and live services. Microsoft azure provides a cloud environment for running your web application and storing data. Its like visualizing windows services available on cloud.
Microsoft has provided his own datacenters where they have hosted software's like SQL services,.NET services,Sharepoint services,Microsoft dynamic services and live services. Microsoft azure provides a cloud environment for running your web application and storing data. Its like visualizing windows services available on cloud.
Courtesy:
- The above image is taken from the white paper
An Introduction to Microsoft® .NET Services for Developers: - The .NET framework for the cloud by Aaron Skonnard, Pluralsight. You can read the white paper from http://go.microsoft.com/fwlink/?LinkID=150833 .
An Introduction to Microsoft® .NET Services for Developers: - The .NET framework for the cloud by Aaron Skonnard, Pluralsight. You can read the white paper from http://go.microsoft.com/fwlink/?LinkID=150833 .
Isn’t
it friends, sharing cuts down cost J
How is billing
and costing done in Windows Azure services?
As
discussed above there are two types of cost one is the fixed cost (server
hardware, employee salary etc) and other is the variable cost (bandwidth usage,
hosting storage space etc).
Windows Azure services are billed using pay and use model. Microsoft terms this as consumption based model. In other words the customer does not need to pay anything upfront, everything is paid as per consumption.
The pay and use model is defined by four characteristics:-
• Compute / Hour: - Depending on how much computing power your application uses you will be charged. When I wrote this article it was $0.12 / hour.
• Storage in GB / month: - Storage is measured in units of average daily amount of data stored (in GB) over a monthly period. The storage cost currently is 0.15$ / GB stored for the whole month. Let’s try to understand the above statement. If you store 30 GB for a day then the average comes to 1 GB per month. If you stored 30 GB for 30 days then average come to 30 GB per month. Below are the calculation details :-
Windows Azure services are billed using pay and use model. Microsoft terms this as consumption based model. In other words the customer does not need to pay anything upfront, everything is paid as per consumption.
The pay and use model is defined by four characteristics:-
• Compute / Hour: - Depending on how much computing power your application uses you will be charged. When I wrote this article it was $0.12 / hour.
• Storage in GB / month: - Storage is measured in units of average daily amount of data stored (in GB) over a monthly period. The storage cost currently is 0.15$ / GB stored for the whole month. Let’s try to understand the above statement. If you store 30 GB for a day then the average comes to 1 GB per month. If you stored 30 GB for 30 days then average come to 30 GB per month. Below are the calculation details :-
Figure
:- Storage calculation details
•
Bandwidth: - One more factor
which is used to measure cost is bandwidth, in other words how much data goes in
and out from the data centre. We will discuss on the bandwidth cost in more
detail later because this has lot of variations depending on the services
provided by Azure. Please note bandwidth used within data centre is
free.
• Storage Transactions: - Any kind of add, update and delete on the storage data is also tracked. Currently it’s billed at $0.01 for 10,000 (10k) transaction requests.
The charges mentioned in this article can be obsolete as time passes by, for recent rate card please visithttp://www.microsoft.com/azure/pricing.mspx
• Storage Transactions: - Any kind of add, update and delete on the storage data is also tracked. Currently it’s billed at $0.01 for 10,000 (10k) transaction requests.
The charges mentioned in this article can be obsolete as time passes by, for recent rate card please visithttp://www.microsoft.com/azure/pricing.mspx
We
are still not clear with the payment model it looks a bit
confusing?
There
are three parts to the Azure payment model:-
Metering: - This defines the unit of measurement. For example compute per hour, storage per GB, bandwidth per GB and transactions per 10k.
Billing: - This defines the dollar amount for the unit.
Services: - Both metering and billing needs to be associated with a service like .NET service, SQL Service, Live services etc.
Metering: - This defines the unit of measurement. For example compute per hour, storage per GB, bandwidth per GB and transactions per 10k.
Billing: - This defines the dollar amount for the unit.
Services: - Both metering and billing needs to be associated with a service like .NET service, SQL Service, Live services etc.
The
charges mentioned in this article can be obsolete as time passes by, for recent
rate card please visithttp://www.microsoft.com/azure/pricing.mspx
| Windows Azure | SQL Azure | .Net services | |
| Bandwidth IN | $0.10 / GB | $0.10 / GB | $0.10 / GB |
| Bandwidth OUT | $0.15 / GB | $0.15 / GB | $0.15 / GB |
| Storage | $0.15/GB stored/Month | NA | NA |
| Transactions | $0.10/10000 K | NA | NA |
| Messages | NA | NA | $0.15/100K message. |
A note of SQL Server cost
Web Edition – Up to 1 GB relational database = $9.99 / month
Business Edition – Up to 10 GB relational database = $99.99 / month
What
is Fabric?
Before
we start to answer this question, let’s thanks Steven Nagy for helping us
understand this terminologyhttp://azure.snagy.name/blog/?p=84.
As said before Microsoft has his own datacenter’s where they will have number of server’s running windows 2008 servers.
As said before Microsoft has his own datacenter’s where they will have number of server’s running windows 2008 servers.
Note: - While I was writing this article what I understand is that currently they have one data centre in US west coast which hosts Windows Azure application.
The
whole concept of Azure is to give decent hosting rates to the end user. So
Microsoft has to somehow share his hardware across multiple applications.
In order to share the physical hardware across multiple applications, Microsoft used ‘Hypervisor’. Hypervisor is a modified version of Hyper-V. So hypervisor can help you to create virtual machines from those physical hardware servers. You basically end up with a cluster of nodes which can have virtual machine or hardware’s.
If you really expand these rectangle nodes you end up in to a Fabric like structure.
In order to share the physical hardware across multiple applications, Microsoft used ‘Hypervisor’. Hypervisor is a modified version of Hyper-V. So hypervisor can help you to create virtual machines from those physical hardware servers. You basically end up with a cluster of nodes which can have virtual machine or hardware’s.
If you really expand these rectangle nodes you end up in to a Fabric like structure.
Figure:
- Fabric
So
now let’s define fabric. It’s nothing but a cluster of nodes. These nodes can
have physical machines or virtual machines.
Where does our application
run?
Our
application run on the fabric nodes and each application gets its own virtual
space resource. We will discuss about the same in more detail as we move ahead
in the FAQ.
What is a fabric controller?
Fabric
controller is the heart of the Azure fabric system. It manages complete life
cycle of Azure services. Below are some important roles which Fabric controller
plays:-
• Manages provisioning for the application. As per application needs fabric allocates CPU , memory and bandwidth limits.
• Deploy the services.
• Monitor system status and health. See that it satisfies the SLA’s defined.
• If there are failures recover from the same.
• Manages provisioning for the application. As per application needs fabric allocates CPU , memory and bandwidth limits.
• Deploy the services.
• Monitor system status and health. See that it satisfies the SLA’s defined.
• If there are failures recover from the same.
What are roles, web role and
worker role?
Roles
are nothing but applications or components. To define the same in other words
they are actually the application code. Windows Azure categorizes application in
two categories one is the web application and the other is the worker
application.
Web role / Web application: - A web role is a Web application which can be accessed via HTTP. A web role can be hosted as a subset of ASP.NET and Windows Communication Foundation (WCF) technologies.
Worker role / Worker process: - A worker role is a background processing application somewhat similar to a windows process. A worker role does not communicate directly with external the world. In other words it does not accept requests directly from the external world.
Web role / Web application: - A web role is a Web application which can be accessed via HTTP. A web role can be hosted as a subset of ASP.NET and Windows Communication Foundation (WCF) technologies.
Worker role / Worker process: - A worker role is a background processing application somewhat similar to a windows process. A worker role does not communicate directly with external the world. In other words it does not accept requests directly from the external world.
So
is it that worker role cannot connect to external systems outside
Azure?
Yes,
worker role cannot connect take inbound calls from internet but it can make
outbound calls. A worker role is background processing process so it does not
have incoming internet requests. Worker role also has access to queue services.
So that can be one more communication point where external systems can post data
and worker role can read from the same.
Figure:
- Worker and Web role
Is
it that every web and worker role is hosted in their own virtual
machines?
Yes,
every web role and worker role are hosted in their own virtual space or we can
say in their own virtual machine. Fabric does not allocate IIS for worker role
virtual machine.
You
can see from the above figure how fabric controller uses the data centre
hardware and creates new VM. Our application runs in those virtual machines. So
if it’s a web role fabric controller will create a new VM instance with IIS and
your web role instance. If it’s a worker role then it will create work role
instance only. It will not provide IIS instance for worker role as its not
supposed to take inbound calls from external world.
How
do the web role instance and worker role instance communicate with
fabric?
Every
virtual machine has an agent. This agent facilitates communication between web
and worker role instance with the fabric.
What are blobs, tables and
Queues?
Till
now we have only discussed about application code. With application you also
need to save data. Windows Azure has three kinds of data blobs, tables and
queues.
Blobs: - They are used to store large objects like images, audio, video etc.
Tables: - They are used to store durable and scalable data structure. Tables consist of entities and entities have properties.
Blobs: - They are used to store large objects like images, audio, video etc.
Tables: - They are used to store durable and scalable data structure. Tables consist of entities and entities have properties.
Figure:
- Table structure
Queues:
- Queues are used enable
asynchronous communication. They are enabled to store transient data in form of
messages. Azure uses queues to communicate between different entities. For
instance of a worker role and web roles wants to communicate they use
queues.
We will be discussing the above 3 things in more detail in the coming FAQ series. We will be dedicating each one of them a complete FA series. Till then hold your breath.
We will be discussing the above 3 things in more detail in the coming FAQ series. We will be dedicating each one of them a complete FA series. Till then hold your breath.
So is SQL
is the standard way to query blobs, tables and queues?
No
we cannot use standard SQL to query the above azure storages. They are accessed
through REST (Representational state transfer). REST is a style of software
architecture distributed systems. So if you want to read a data from a table you
need to specify the below URL.
http://<>/?$filter= >
The first is the website on which the table is hosted , second is the table name and then your query.
http://<
The first is the website on which the table is hosted , second is the table name and then your query.
Why was not SQL preferred?
REST
relies completely on HTTP protocols which enables other languages like PHP.JAVA
to communicate with Azure storages. In other words you can build your
application using any language like CGI, PHP and have your data stored in Azure
storage.
Why one more data storage when we already have SQL Server?
Windows
azure will be using a common hosting platform so that cost of hosting is less
for customers. In other words the storage should be highly scalable. In order to
support these kinds of requirements we need to scale out and not scale up.
Windows azure storage is built on scale out architecture.
Due to this SQL Server service is costly than azure storages. Let us try to understand why the cost increases because of Scale out architecture.
Scale up means when your database needs more processing power you just replace it with a more powerful processor. In other words you just buy more powerful processor with better RAM and put your application on the same. Because the numbers of processor do not change you are not charge with extra SQL Server licenses.
Due to this SQL Server service is costly than azure storages. Let us try to understand why the cost increases because of Scale out architecture.
Scale up means when your database needs more processing power you just replace it with a more powerful processor. In other words you just buy more powerful processor with better RAM and put your application on the same. Because the numbers of processor do not change you are not charge with extra SQL Server licenses.
Figure
:- Scale up
In
scale out you add multiple servers as and when you need more processing power.
Because news servers are added you also need to buy SQL Server licenses for each
hardware. Azure is a scale out model. As per more processor needs Microsoft will
add more hardware to his data centres. Due to more licenses the cost of SQL
Server service will be higher as compared to windows azure storage.
Figure:
- Scale out
So
use Azure storage:-
• If your application is having a massive scale.
• You are not looking for rich relational data functionality like SQL, structured relational tables etc.
• You are looking for a less cost.
You need to use SQL Azure service
• Scalability is not the important factor.
• You are looking for rich database functionalities like SQL, structured relational tables etc.
• Cost is not a big matter.
• If your application is having a massive scale.
• You are not looking for rich relational data functionality like SQL, structured relational tables etc.
• You are looking for a less cost.
You need to use SQL Azure service
• Scalability is not the important factor.
• You are looking for rich database functionalities like SQL, structured relational tables etc.
• Cost is not a big matter.
Comments