Day 2/ Session

What can the cloud do for you

Gaurav Agarwal, General Manager, Tivoli, India/South Asia, demystified the concept of cloud computing and talked about CloudBurst, an offering from IBM that helps organizations tap into the benefits of this technology

Agarwal started by asking the audience how many understood what a cloud was in the context of cloud computing and what needed to be done to make it relevant to the board, to the organization and whether it could really deliver a better experience to the users. How can we be dynamic, relevant, fast, and agile and how can you make it work for your organization. I’m not going to talk about Google or Facebook because you are not building a Google or a Facebook here. We are talking about banking operations, manufacturing etc.

When you hear about cloud computing the first thing that comes to mind is what can we do and can’t be done. In 1999, I went to one of the largest manufacturers of motorbikes in India and he was going to buy an ERP system. He said how much storage will I need? I said, ‘Approximately 400 GB.’ He literally jumped out of his chair and said, ‘Up until now I have been managing with 70 GB of storage. Now you are saying I will need 400 GB for ERP.’ Now, most you have more than 10 TB in your organization. I’m talking to a bank and he says he’ll have 1,500 TB in two year time. That’s about the rate of change that’s happening, 54% year-on-year. 70% of our money is being spent on managing stuff that we have rather than procuring new stuff. $40 bn or 3.5% of all the revenues of consumer goods companies in supply chain inefficiencies. This is screaming for your attention, for you to get in the boardroom and say I did this and we got this result. It’s time to start thinking differently about infrastructure.

It’s time to industrialize the datacenter. We’ve not been giving businesses the quickness of response that is needed. Rather than looking at a silo approach where you have an application guy telling you what infrastructure you need and then the infrastructure guy telling you how much time he needs to deliver it. Can I have a consolidated view of a business service? For e.g., if you want to launch a credit card operation, and that’s a business service, how I can launch that operation including the application infrastructure below it and can I have a single view of how that’s working.

This will mean self-service which means that rather than provisioning a requirement to deliver a service, the user who needs it can provision it himself based on the inventory pool created for him right at the beginning of the year. He can get service in a day, right now, and not six weeks later. We’ve announced something called IBM CloudBurst which is for IT organizations. It is an integrated package with hardware, software, services all built in together to make the cloud real today and tomorrow. One of the things I’m going to focus on is a test & development environment as an example. How it can really help you. Anybody who’s had an ERP application such as SAP deployed, CRM, Supply Chain and the standard SDL finance module. All of them are separate test & development environments. Before you realize it, you have 30 test & development environments running your SAP system. If you are a bank running CBS, you’ll have cash management, you will have credit, you’ll core banking—again you are looking at 15-20 applications each having its own test & development environment.

Individual servers are not fully utilized. For example, in a test environment you would test before going live after which that test server would be lying idle for the next change that you make to the development environment. And then you propagate that change to the test environment and test it. When you test it you don’t have enough resources to test it because a test environment is never as good as a production environment. Actually you want to test it in a test environment that’s as good as the production environment because your neck is on the line when the application goes into production. How many sleepless nights have you gone through when you are moving to a new application or migrating to new stuff that’s going into production and you are not comfortable that it can scale. Because the test environment never allowed you to test it to the scaling levels that you wanted to test.

It can improve service. Nedbank, one of the largest banks in South Africa; their problem was that any user service request, for e.g. any new application, development environment, test environment, production environment—it was taking a lot of time for the IT department to deliver. This was frustrating the business users and making the bank slow to respond.

Every time the bank’s IT department had to provision a new environment, it had to take the time away from a project and go in and put a DR or additional test environment in place and while doing that projects were suffering. Using TPM the bank’s IT department can now provision the environment overnight, the results are available in the morning, the IT staff checks a couple of things, ensure that it’s all OK and get back onto the project work.

A utility company had too many departments and each department was facing resource constraints and the number of specialists that they had was limited. Because of this the specialists were overworked while the lower level, less skilled staff was underworked. Therefore the speed of deployment was slow. This company’s IT team created standard images for every application and now users can click on the link and get those images to get up and running using cloud computing through a self service portal.

What CloudBurst will deliver is a self service portal which allows people to request for a service and the associated infrastructure along with it. It offers a service catalog in terms of what services you want as infrastructure etc, automation software to automatically provision what you want, built in virtualization and single delivery. You don’t have to talk to ten people to get this going.

Agarwal then showed a demo of how CloudBurst can be used to create a private cloud environment in a jiffy. This was a demonstration of self-service, zero touch administration with a new embedded, automated service management layer. Provides for lights out automated operation and includes a reusable image library for rapid deployment. A new user was created to log into CloudBurst. The user saw projects that belonged to her on a screen similar to the administrator’s screen. Since she was a new user nothing showed up under new project. In the Request New Cloud Project window you can see the available resources in CloudBurst which comes preconfigured with VMware/System x cloud resources. To request resources you select a date range or move a slider bar. As the dates change CloudBurst dynamically updates the available resources to show the capacity during that given date range. In the next screen CloudBurst asked for the project name and description. At the bottom of the screen you could choose different images to be provisioned including Windows Server 2003 SP2 and VMware RHEL 5.3. The VMware image was selected for the demo. CloudBurst allowed the user to customize the image in terms of CPU, memory and disk usage. As per the choices made, the maximum number of VMs that could be provisioned was automatically updated. The user in this demo chose to provision four VMs at the same time. In the next screen the user got to choose to provision specific software products. Select the monitoring agent. The VMs could be added to the user’s shopping cart by clicking a button. Next up was a screen with an overview of the VMs to check for errors, if any. Clicking submit, sent the request to the administrator for approval. CloudBurst checked resource availability, created the project and reserved and scheduled resources. In the overview screen you saw the FrontOffice WebOffice 3.0 appear under My Projects. The administrator was notified by e-mail. As part of the demo the demonstrator logged in as the admin to review and approve the project. The administrator selected the FrontOffice WebOffice 3.0 project to review the request details including who requested how many specific machines for how long.

The demonstrator selected approve project to kick off provisioning.

The request status thereupon changed to provisioning. Reports are generated to give details of projects provisioned and this can be filtered from a drop down list by team. CloudBurst also has remote control features including power on/off the VM, restart it, back it up and so forth and reset the password for the VM.

The benefits are improved service, reduced cost and it’s easy. We implement this for you. We have a detailed assessment capability for doing this which includes the use of your existing hardware and that can be a part of the cloud including the CloudBurst hardware that goes with the basic provisioning as well as indicate additional hardware that’s required.

Our vision is that this is relevant, that it will have all kinds of OSs, all kinds of workloads. It will not limit you to any particular environment. Although it initially doesn’t have all the OSs in it as we move forward in 2010 you will have other OSs such as zOS in that offering itself.

Next you have to figure out what do I do on the cloud. Do I send everything to it? Our view on this is that if cloud as a strategy appeals to you, in cases such as ERP, transaction processing, database etc. where security is a big issue, you have to look at what can be standardized. Things like Web, messaging, HPC—where you need Terabytes of storage and trillions of compute capability, test & development which is characterized by peaks—different users use different environments at different points of time are natural fits for cloud computing as it enables high-end infrastructure sharing.

In today’s environment stuff like e-mail, collaboration, software development, test & production, data intensive processing (HPC etc.) are moving into the cloud. Workloads moving into the cloud include Single Virtual Appliance, Test & Production systems, batch processing, storage as a service, backup as a service, HPC etc.

Things that require extensive customization don’t work because the cloud is about standards. Workloads that require a high level of auditability, workloads composed of multiple co-dependent services and stuff like that—aren’t feasible.

Figure out what you want to use as public, what you want to use as private, what is shared, what is owned, and therefore evolve your own strategy. You are doing this to reduce IT infrastructure cost. We have a technology portal where all new IBM software development that has an impact on business gets published here. For e.g. our BlackBerry equivalent capability for Symbian-based phones like Nokia came on that first. This is an environment that typically has 120 projects. When we did this it had 488 servers and 20 IT administrators supporting those projects. We moved them to a cloud. We brought the number of servers down to 55. The admin staff came down to two people. From a $3.9 mn spend it went to $0.6 mn. Power costs went down by 88%. Labor costs went down. Hardware costs went down. Software costs remained constant because we didn’t really provision applications per se.

Our research shows that if a cloud is implemented well, it should be able to save one-seventh of the infrastructure cost. If you save one-seventh, you should get 18-19% savings if it is implemented and planned well.

Another e.g. is of our research labs who deliver projects for 2013-15. A researcher can select a Windows/AIX/Linux environment or a mainframe environment with no intervention and he adds it to a cart and it gets delivered to him based on the resource pool available in any of the labs in the real world.

This is here today, not 2-3 years away.

— Prashant L Rao