By guest authors Shaloo Shalini and Pablo Chacin
SM: Let’s switch to the topic of standardization. Are you seeing standards evolve in the domain of cloud computing that make job of integration and interoperability any easier?
MS: No. I haven’t seen that emerge yet. But I just keep getting this question, a lot. I kind of find it ironic because if you simply look at the IT assets that are within your data center today, those are managed by the companies [that make them], and there aren’t a whole lot of standardized interfaces there itself. If you want to try and manage a mixed SAN environment with, say EMC, Hitachi, and NetApp that are there on your floor, they sometimes don’t solve integration problem internally. So it is almost like having a greater burden of proof and requirements from cloud providers than for existing vendors who were selling their stuff all the time without catering to standardization needs. Yeah, we keep beating the cloud vendors on standardization, but we don’t go back to EMC, Hitachi, and NetApp and ask them, “When will you guys get together, and come up with one standard look and feel so that we can manage a mixed environment?” When I go to buy my software products, I find that some of them work better on NetApp and others work better on EMC. We don’t do that with servers, but the cloud guys get beaten up. That seems to be a recurring theme.
Another thing that is ironic here is, which I am sure, if these vendors get together under an umbrella, say a noncompete umbrella, with a lot of standardization thrown in, then probably it is the IT guys who will complain about how inflexible the model is!
SM: You can’t do it one way for one and another with cloud. The cloud vendors are going to love your responses on this topic.
MS: I think it is the Wild West out there right now. There are lots of different business models and different approaches. Amazon is set up to sell you a virtual machine (VM) at a time, and Rackspace offers servers that they call “bricks.” They let you install hypervisor of your choice on their brick, and you can virtualize it to your heart’s content. It really gets us back to the business scenario. It is just a matter of figuring out what you want to do with this infrastructure on-demand capability available out there, what’s the business model and answering the question: “How do I want them to offer that up to me?”
SM: Moving over to the topic of different pricing models in the cloud, starting from fixed monthly fees to per user license, user-based fee, fees as a percentage of revenue, fees as a function of number of clients, and so forth, do you have a preference for a particular pricing model for cloud-based solutions?
MS: I have a presumption that the fees will converge over time to become structured similar to a telecommunication contract which you negotiate over some period of time. It could be a two- to three-year commitment where there is a certain base level of usage that you are committing to a cloud provider. That’s good for the cloud provider’s business because then they are able to plan and have guidelines on how much hardware they have to have, how much data center floor space, operators, and so forth will be required so they can add that up over time and not just sell an account at a time with no idea what kind of variation they are going to see. Just like a telecommunication contract, as long as you meet your certified minimum level of usage, you are typically discounting incremental demands, typically that’s provided at a discount relative to the base.
I would expect that for Fortune 500 companies, most will start with a standard price sheet when you approach the provider, and if you maintain an ongoing business relationship, you could demand discounts and conditions such as guarantees that they will have exclusive hardware not shared with other customers. With the right level of demand, they may be able to provide that as a guarantee to a high-paying enterprise customer.
SM: I am curious about these two models in the mix. One of them is the notion of real or true utility computing, which reflects your actual usage as opposed to, say, shelfware or boxed software. I am referring to the era when Siebel dominated the application software landscape, the shelfware just sat there without being implemented, with long adoption cycles. So people spent a lot of money but did not get benefits out of it.
Now, with cloud computing models having the ability to monitor and record usage, utilization-based compensation is a lot easier. So, is that the direction cloud pricing models are going to move in?
MS: I think there is a definite market for that. People would love that. It might have an interesting boomerang effect. It might cause them to ask questions such as, Who is using that stuff we own on the data center floor? You can tell them exactly, that out of my 6,000 employees in the organization, 2,349 used Amazon services over the past month, and I know exactly which rates I paid for each one Then we might also have questions like, I have 1,000 server farms sitting on my data center floor; if we have 2,300 up there, then what are the other guys doing? It might actually arrive at some penetrating business questions about the utilization of stuff which we really do not report on today in the data center. So it is attractive for people to have that level of insight into usage.
SM: The other question I have is about sharing some level of success scheme with the vendors. Whether it is transaction based, or a percentage of revenue, or whether a vendor participates in the upside. What are your thoughts about such pricing models?
MS: I think that’s a foreign concept which I haven’t really talked about. The presumption is that vendors are achieving economies of scale by aggregating demand that I could never achieve myself. So they have the game share on their side already. They would be trying to negotiate the rates. That would be an interesting question for an Equinix or Savvis-type of service outsourcing, obviously, on the labor side, where you outsource IT and labor functions with long-term outsourcing contracts where the benefits flow back to the customer.
SM: My sense on this is that it depends on the business. I have seen that in one company, which actually turned out to be a public company, called PDF Solutions. They are a very specialized vendor that does optimization of semiconductor yield management. Semiconductor yield is a very expensive business. If you can get even a percentage point of yield improvement, that could be like $50 million in cost savings. They have been able to get gain sharing from their customers just because it’s such a high-impact leverage position.
MS: So you could argue, if you go back to my telecommunication example, with aggressively deeper discounting or demand in excess of my minimum requirement which is actually a form of gain sharing that I am just specifying in a contract.
SM: How comfortable would you be buying cloud services with minimal touch points?For example, would you prefer buying AWS directly from Amazon of from a local VAR which is trusted and closer to your shop?
MS: I don’t have a strong feeling one way or another. I think pricing and performance would be the two keys, besides the sharing concept. If I am giving someone enough demand, are they effectively walled-off and giving me some dedicated hardware during a period, even when a service is virtualized? Can they guarantee that all my VMs will be on a common cluster or distributed cluster? Some of those operational aspects will be critical to me.
SM: And that’s easier to negotiate with the vendor itself, than the VARs, right?
MS: Yeah, in general.
SM: What is your experience with IT shops which are moving from a licensed software business model to a cloud model? How are they doing in helping you make that transition or managing their own transition?
MS: You are talking about SaaS type of applications?
SM: Yes. Ariba for instance, has recently made a pretty committed switch from a licensed-software procurement vendor to a cloud vendor, and several others have been doing that recently. All these companies which are precloud companies are now taking their entire strategy on to the cloud. How does the relationship with the client change with respect to the business model, the license model? Do you have any experience of working with a vendor that has made that switch and how have you dealt with that?
MS: I don’t have lot of direct experience. Actually, I would say that lot of shops won’t do that because you already have lot of application installed on premises, to some degree of customization and even greater degree of integration with other applications. So to pick that up over on to the SaaS version of application, to make that transition is difficult. You are actually delivering the same functionality, with just the change in the delivery channel from the on-premise to off-premise. The business case for that would be a tough one.
The original license fees and a lot of hard work that has already gone into deployment and integration, has depreciated close to zero. This is a more relevant question for a new customer. A new customer will decide whether he or she wants license-based software or a SaaS service. So it is much more relevant question in a new purchase, I would say.