By Sramana Mitra and guest authors Shaloo Shalini and Bhavana Sharma
SM: So, you are supplementing your internal infrastructure with Amazon EC2 and so forth?
DF: Yes. You can basically set policies that define which conditions allow you to do that [spill over]. Obviously, some application or tests you may not want to run in the cloud. So you set the policies [to deal with those], but we have a variable capacity cloud, and we get some capacity from places like Rackspace and Amazon EC2.
[Note to readers: You may want to read more about resource sharing using cloud technologies and Labs on Demand at CA here.]
SM: This infrastructure, has it ever been used in the mode where you needed to spill over and access external infrastructure?
DF: Yes. It is in production now, and we have had use cases where it has happened. We also run new product release beta programs with customers using cloud-based technologies. What we have done is this: Customers don’t want to buy hardware to run the beta product, so we have actually had that hardware hosted such that we now allow customers to run the beta on our hardware and try our new product.
SM: So, this beta is one of the products you are trying to market?
DF: Yes, it is one of the products that we are going to sell.
DF: It’s basically a product we have built, and it has now gone into beta stage where customers are now running the beta. Despite the fact that from our customers’ point of view, it’s kind of a beta or proof of concept thing, doing beta is one of the core parts of our business model.
DF: It’s not really doing the beta itself, that isn’t a proof of concept for us. It is a core part of our product life cycle. So, we did critical tasks, and we did the most recent beta version of our service desk management product on Amazon EC2. What that means is that all our customers ran their beta images on EC2.
SM: I see. When these products move out of beta and into full deployment, will they remain software as a service (SaaS), or will the model change to an on-premise license?
DF: It will be both. [We have] some customers who want it to be a service, and there are some customers who want on-premise. So, we do have an on-demand version of some other products such as Service Desk. It’s not quite SaaS. There is some amount of fluidity, or let’s say, the boundaries between hosted on-demand and SaaS are somewhat ambiguous. So, the real difference between on-demand and SaaS, some of it is technical. There are lot of terms and conditions. But we do have our on-demand versions of our products including Service Desk.
SM: What about applications like ERP, CRM, HR, and analytics? Has CA moved any of those to the cloud?
DF: We use Salesforce.com for our sales force automation process.
SM: Is it fully deployed? Does all of CA use Salesforce.com?
SM: So you have made a full switch to the cloud there.
SM: Is there any other functionality or workload that has moved to the cloud in CA?
DF: The answer is yes, although I don’t know the details. One of the things that we did is when we started talking publicly about cloud computing, a lot of customers asked us, what are you doing, and we went back and analyzed our applications and traffic going in and out of the company’s intranet. We found that we are using about sixty external services.
Let’s take a step back. When people talk about the cloud or when I talk about the cloud, I tend to talk about four types of cloud services – first is infrastructure as a service (IaaS), and an example of that is the fact that Labs on Demand will spill out to EC2. For the second type, platform as a service (PaaS), we are not currently doing anything except for what we are doing on Force.com in the form of one product. Agile Planner is available in in that area. Third, there is software as a service (SaaS), which is what we are doing with salesforce.com within CA as a business process.
Then there is a fourth category for which the general term is Programmable Web or Web callable application programmable interfaces (APIs). Basically, you can have an internal business process and one of the steps in your process actually invokes a web services that is external. A lot of our external services are used in that mode. When you think about our HR systems [at CA] if somebody’s role changes, you may have to invoke an operation on the benefits providers. Say if you are basically on-boarding an employee, [a new employee]: You have to send a message to the insurance provider and to the 401(k) provider to deal with this. In a lot of cases we are using callable services that are effectively on the Web. I think that is going to be increasing over time. When I worked on SOA at IBM, the big hallmarks of SOA is, if you encapsulate all of your internal applications and put them in a registry, when you built portals or processes, a lot of it is just assembling calls and sequencing calls to existing things. If you look out even on some of these spaces like programmableweb.com – there are 5,000 Web callable APIs. So, I think an increasing model of the cloud is composite applications that can be calling cloud or Web addressable services and then people would be deploying the composite applications on the platform as a service (PaaS) such as Microsoft Azure.
SM: Is that something you see very often? I remember when I was working with SAP in a consulting capacity, at one point, composite applications was a big deal. There was a big push for them, but we hear a bit less of that these days, is that correct?
DF: Yes, and I think the reason that we hear less about it because everybody has just accepted the fact that applications are just composite applications if you really look at them closely.
SM: That is natural for cloud models, right? It is almost like a mash-up in the cloud?
DF: Yes. I mean, it is natural for the cloud model because if you have the callable APIs on the internet and you want to build a business process that calls the APIs, why wouldn’t you just put the business processes on the platform as a service? That’s to some extent where Azure is going. There are other things that do that, but I think you know the long term it going to be this model where you have your applications and then deploy them on PaaS, and in that a lot of what you are doing is calling Web addressable APIs that scramble the Web.
SM: Very interesting.