by July 23, 2013 0 comments

Students and researchers in TCE were demanding more and more physical compute clusters and data storage to run their experiments. The situation was highly manual and involved filling paperwork on part of users and admins to book time slots when the resources of the data center could be used. Frequent installations and configuration changes to physical machines disrupted the stability of the servers and clusters.

The implementation

A cloud framework was set up using the OpenStack standard to loan out resources from TCE’s data center. A 32-node cluster was set up, and the master node acted as the gateway for any user across campus to interact with the cloud. The master node also hosts daemons for major services like imaging service, authentication, dashboard web, VLAN, volume services etc. The master node has two interfaces, one to connect with other slave nodes in the network, and the other to communicate with the college intranet and internet. Using SNAT protocol, the master node acts as an interface for the slave nodes to access internet/other services that exists outside the cluster. The horizon dashboard, which is accessible to all users of the cloud, is an amazing tool that dynamically allocates resources to the user depending on the scale of the experiment being run.

Company Scenario

Before deployment: The demand of researchers for computing power was taking its toll, because frequent configuration changes to the physical machines was affecting stability of the cluster and servers. An automated solutionto maximizing the data center’s usage was badly needed.

What was deployed: A 32-node cluster was setup using the Openstack framework to lease out resources from TEC’s data center. Master node hosted daemons for major services such as authentication,imaging service, dashboard web service, VLAN service and so on. A horizon dashboard web portal was used to provide access to cloud users, who could then setup the amount of computing power they would require.After deployment: Utilization of servers in data center was only between 5-15%, but now has increased to 81% due to running multiple VMs on the same physical infrastructure. Each resource in data center has its own Access Control List(ACL), so access to every resource is restricted and shared according to needs.

Implementation partner: In-house Members

Challenges Faced[image_library_tag 879/81879, alt=”asdfg” border=”0″ border=”0″ style=”float: right;” ,default]

The technical aspects of the project were challenging, such as the configuration of the computing cluster to host a cloud. New users also had to be trained to use the Horizon dashboard, so that they realized its potential and could stop relying on the admins to arrange an environment for them. Also, keeping the cloud operational 24/7 was a big challenge with the constant power outages in Tamil Nadu.

The Result

Users can now easily utilize all 32 nodes of the cloud and create Virtual Machines(VMs) to suit their requirements. The horizon dashboard helps researchers to claim resources as they need from the cloud and run their experiments on the VMs. Before utilizing OpenStack, the utilization of servers in the TCE data center was only between 5-15%. Due to the use of VMs, the utilization of servers is up to an amazing 81%. With the ease of use, Non-CS students are able to set up experiments on the cloud without any help from others. The total cost of maintaining the data center in terms of labor costs/electricity has come down because of maximizing the utilization of the same set of physical resources.



No Comments so far

Jump into a conversation

No Comments Yet!

You can be the one to start a conversation.