I have been working on Cloud computing (mostly Amazon AWS and Google App Engine) for quite sometime. And recently, I decided to use the technology in the hosting environment. I redesigned the layout of the Nazdrovia Servers and created a middleware to interact with the cloud.

This will actually help in saving some money and ,more importantly,  a good amount of time. Nazdrovia is using the data centres of two major ISPs based in US for providing the web hosting service. One of these sets acts as back servers so that in the event of a crash or broke down, all the load on a server will be switched to the server located in the data centre of other ISP. There are scripts running in the servers that perform the back process (synchronisation) automatically and securely. This model was appreciated by many clients and this factor alone triggered many clients to migrate to Nazdrovia Servers.

But if we analyse it in detail we can see that the company is investing more money. For example, when a customer buys 1 GB space from Nazdrovia , the service automatically allocates 1 GB dedicated space for that client. But almost all other companies are over-selling the space i.e. they will assign a dynamically expanding storage and will allot only a bare minimum in the beginning. Though this can save money, it is not a good idea when it comes to performance. This is primary reason why Nazdrovia didn’t follow this method. This can be solved by shifting to cloud.

The New architecture

I believe in openness and transparency. So I don’t wish to keep the new architecture as a ‘trade secret’ (In fact, I will posting another entry focusing on the technical details). In the new architecture, virtualization is employed so as to provide a novel and secure way of storing the data. The rough design will be like this:

• The C-panel based interface for the end user
• Virtual dedicated servers (VDSs) for storing the contents of each user. The C-panel will be communicating with the VDSs
• A Super Virtual dedicated server for ‘hosting’ VDSs
• Cloud as the base platform

This design has  many advantages:

• Secure way of storing the data – since each customer is virtually separated
• Users do not require to buy dedicated servers if they want to have more CPU power or RAM (they can go for ‘pay as you go’)
• Backup will always be there on the Super Virtual dedicated server

Testing

Last year, I upgraded by laptop to :

• Intel®  Core™2 Duo Processor
• 500GB SATA (7200RPM) HDD
• 4 GB RAM (DDR3)

But it started crashing when I started doing trials using virtualisation technologies. So I upgraded the laptop again to:

• Intel Core 2 Quad with 6MB Cache
• 1TB HDD (the Scorpio Blue) – model WD10TEVT Western Digital
• 32 GB RAM

And now it works perfectly fine. Now I have OpenSuse with a hypervisor as the base operating system. On top of that I have Ubuntu. And I use another hypervisor on that Ubuntu (9.10) system to run the following operating systems in parallel:

• Ubuntu (9.10)
• Fedora 12
• Solaris (yeah, I’m still keeping a copy!)
• Windows Vista (hmmmm….. keeping at copy as a well)

Hardware items

In order to test the same architecture, you may need the following hardware items.

Processor: It is easy to find a quad processor for your laptop. You can go to Intel’s website and search for the same.

RAM: upgrading to 32 GB RAM was slightly difficult. Fortunately, my new processor supports multiple slots and Samsung has released one of their latest green product a couple of months back. If you are interested in knowing more technical details, please consult this page. This allowed me to have 32 GB RAM in the system.

HDD: Well, as you can guess, this upgrade is not necessary since our interest is more inclined towards the performance.  But in my case, I have to upgrade my drive since I wanted to pull so much of data (software) for testing various features.

I searched for a suitable product for quite a long time and finally stumbled upon the model WD10TEVT (manufactured by Western Digital). You can find the product details here.

Test Results

After performing some of the basic testing it is found that the new architecture is economic and easy to manage. The new model allows the administrator to add more customers to the framework very easily. The framework will get scaled up automatically as the  demand goes up.

I will be posting another entry once I finish all the testing.