I was sitting in the Oracle Applications Analyst Summit this week when a debate broke out over this question. Now this is not a new debate certainly but it is one that comes up regularly. It seems that our friends at Salesforce.com have done a very effective job at pushing this message, so much so in fact that many people believe that multi-tenant architecture is a part of the fundamental definition of SaaS. But is that really true? There are several software vendors that would not agree, including Oracle.
First, what is multi-tenant and what's the difference between multi and single tenant software. In the simplest definition a multi-tenant solution is a solution that is offered / consumed by many customers but on the back end is a single shared infrastructure. The application is built to segment or partition each customers data so that it is private yet is operated on this shared system. The opposite of multi-tenant then would be single tenant, where a solution is deployed on an individual infrastructure for each customer. Simple enough. Our discussion this week arose when Oracle explained its three customer options for deploying SaaS solutions: 1. multi-tenant in Oracle's data center, 2. single tenant in Oracle's data center, 3. @customer, deployed in the customer's data center (customer manages the infrastructure, Oracle manages the application). At a glance @customer looks like on premise single tenant software but is it? Let's look at the licensing model and see if that provides any insight. In all three options the customer licenses the software in an ongoing subscription model. From a pricing perspective customers pay the same for single tenant and multi-tenant while @customer is less expensive (makes sense, the customer has to provide the infrastructure). So by Oracle's (and several other software vendors) definition tenancy doesn't make a solution SaaS or on premise.
To understand this view let's look at the two models and see what does make them different. SaaS solutions are by definition licensed in a subscription model. The customer pays a monthly fee that includes the use of the software, support, bug fixes and access to new versions of the software when available. On premise solutions are licensed in a perpetual license model, customer pays an up front license fee and then an annual "maintenance" fee that covers support, bug fixes and access to new versions of the software when available. So the two models are fundamentally different in the way the software is licensed. Are there any other differences? I'm not aware of any...
So why do several software companies use multi-tenancy as a differentiator? The claim has been that multi-tenant solutions were more cost effective for the customer, in other words the vendor would charge less because it would have lower infrastructure costs with shared infrastructure (as long as the cost savings is actually passed on to the customer, that is). Now I suppose that is true, as a vendor builds up infrastructure it would be less expensive to have shared infrastructure versus single infrastructure for each customer. Multi-tenant solutions are, however more expensive to develop, but that is up front vendor investment that, once expended is finished (which is one of the reasons that it is difficult for an on premise vendor to deploy it's software as a multi-tenant SaaS solution, there's a lot of upfront investment). But is this true for large vendors that already have a large scale data center(s) investment? What if the vendor offers both multi and single tenant solutions at the same price to the end customer, absorbing any back end cost difference in their own profit margin? Then is not customer choice a larger benefit than a back end infrastructure design that benefits the margins of the vendor?
Look at the value proposition of SaaS, easier to implement, cheaper and easier to maintain, costs spread over time versus large up front investment (OPEX versus CAPEX), these are not different in any of the three Oracle options. OK, option 3 might seem like it costs the customer more, but if a customer already has a large investment in a data center and data center resources, it might actually save them money to deploy in that data center but have Oracle maintain the software (and of course pay in a subscription model).
So the conclusion is multi-tenancy is not a part of the definition of SaaS, which is simply a licensing / pricing model, not an architectural model. SaaS is software consumed as a service in a subscription licensing model, that's it, no more.
First, what is multi-tenant and what's the difference between multi and single tenant software. In the simplest definition a multi-tenant solution is a solution that is offered / consumed by many customers but on the back end is a single shared infrastructure. The application is built to segment or partition each customers data so that it is private yet is operated on this shared system. The opposite of multi-tenant then would be single tenant, where a solution is deployed on an individual infrastructure for each customer. Simple enough. Our discussion this week arose when Oracle explained its three customer options for deploying SaaS solutions: 1. multi-tenant in Oracle's data center, 2. single tenant in Oracle's data center, 3. @customer, deployed in the customer's data center (customer manages the infrastructure, Oracle manages the application). At a glance @customer looks like on premise single tenant software but is it? Let's look at the licensing model and see if that provides any insight. In all three options the customer licenses the software in an ongoing subscription model. From a pricing perspective customers pay the same for single tenant and multi-tenant while @customer is less expensive (makes sense, the customer has to provide the infrastructure). So by Oracle's (and several other software vendors) definition tenancy doesn't make a solution SaaS or on premise.
To understand this view let's look at the two models and see what does make them different. SaaS solutions are by definition licensed in a subscription model. The customer pays a monthly fee that includes the use of the software, support, bug fixes and access to new versions of the software when available. On premise solutions are licensed in a perpetual license model, customer pays an up front license fee and then an annual "maintenance" fee that covers support, bug fixes and access to new versions of the software when available. So the two models are fundamentally different in the way the software is licensed. Are there any other differences? I'm not aware of any...
So why do several software companies use multi-tenancy as a differentiator? The claim has been that multi-tenant solutions were more cost effective for the customer, in other words the vendor would charge less because it would have lower infrastructure costs with shared infrastructure (as long as the cost savings is actually passed on to the customer, that is). Now I suppose that is true, as a vendor builds up infrastructure it would be less expensive to have shared infrastructure versus single infrastructure for each customer. Multi-tenant solutions are, however more expensive to develop, but that is up front vendor investment that, once expended is finished (which is one of the reasons that it is difficult for an on premise vendor to deploy it's software as a multi-tenant SaaS solution, there's a lot of upfront investment). But is this true for large vendors that already have a large scale data center(s) investment? What if the vendor offers both multi and single tenant solutions at the same price to the end customer, absorbing any back end cost difference in their own profit margin? Then is not customer choice a larger benefit than a back end infrastructure design that benefits the margins of the vendor?
Look at the value proposition of SaaS, easier to implement, cheaper and easier to maintain, costs spread over time versus large up front investment (OPEX versus CAPEX), these are not different in any of the three Oracle options. OK, option 3 might seem like it costs the customer more, but if a customer already has a large investment in a data center and data center resources, it might actually save them money to deploy in that data center but have Oracle maintain the software (and of course pay in a subscription model).
So the conclusion is multi-tenancy is not a part of the definition of SaaS, which is simply a licensing / pricing model, not an architectural model. SaaS is software consumed as a service in a subscription licensing model, that's it, no more.
