The Blog

  • April 9, 2008
  • Multi-tenant matters

    There is an interesting debate beginning to brew in the on-demand market.  Although most people are still clearing their throats, sides have been chosen and trial balloons have been floated.  It appears that a small group of big companies including Oracle, SAP and Microsoft are trying to slow Salesforce.com’s momentum by going after the crown jewel.  Another way to say it is that these companies are trying to hop on board a moving train.

    In one corner is Salesforce.com, the incumbent and still champeen with its multi-tenant architecture. In the other corner are the challengers singing a siren song about freedom of choice offering various flavors of hosting ranging from multi-tenant to single-tenant with a lot of wiggle room in the middle and a Cheney-esque “So?”

    Of course, all this tenancy refers to the way the software is put together and made to work.  Single tenant means one computer (usually) and one copy of the software and all of a company’s users.  In single tenant, or single instance, each additional company has to buy its own software and a system to run it. 

    Multi-tenant is different; it means one or more computers, one copy of the software, and as many users as you want — it’s all you can eat and it doesn’t matter what company you’re from.  Behind the scenes in multi-tenant architectures there is a lot of sophisticated software that keeps things separate but it’s all invisible to the user.  The benefit to everyone is that one copy of the software means one copy to manage, maintain, improve, patch and generally take care of.

    Not long ago, there wasn’t much choice, every application or package you could buy was single tenant.  It was client-server software and it was expensive to buy and implement but that’s the way it was.  Then Salesforce.com came along and made several changes almost over night.  In addition to leading the charge in on-demand and multi-tenant they achieved a new price point for business software that ran in a browser and just about killed the market for conventional business applications.  That’s not to say that enterprises all jumped on the bandwagon but there’s been a steady erosion in the on premise market ever since.  I can’t remember the last time a new software company raised its hand and said, we’re going to make ours using the old recipe.

    That’s all ancient history.  Fast forward to today and you see the challenge I alluded to.  Companies are trying to sell single tenant on-demand and they are saying that the architectural underpinnings don’t matter.  To hear the other vendors tell the story multi-tenant is snake oil and everyone ought to have a choice to go in whatever direction they like.  That sounds pretty reasonable but is it a valid proposition or just convenient marketing of single tenant software?

    For example, according to Anthony Lye, senior vice president of CRM at Oracle, “On demand is just that, a service, a subscription, a delivery option to an application that provides business value to its users, we do that and do that well. We have chosen to provide our customers with choice.”  By choice Lye means using single or multi-tenant as needed.

    Lye makes a good point.  Some companies are not willing or able to go with a multi-tenant approach.  Some entities, like banks, have regulatory mandates to keep their data behind their firewalls and I hear the EU insists on keeping some types of data inside the EU.  Nevertheless, providing a choice of single or multi-tenant is not the same thing as saying that on-demand is just a delivery option. 

    I like Anthony Lye and I think he’s one of the bright spots in the Oracle bureaucracy but I think Oracle and the others are trying to blur a distinction that really matters.  In the years since Salesforce.com came on the scene every software maker worth it compiler has been working hard to eliminate the biggest causes for customer angst.  They’ve replaced client-server software with browser based applications that put nothing on the client machine and they’ve worked on making their applications less expensive and easier to use — all good things.

    However, now the message seems to be if it’s low cost and delivered in a browser across the Internet, it’s good enough.  I believe that single tenant or even an application that can be distributed as either single or multi-tenant poses hidden problems.  The biggest challenge I see is that single tenant delivered on-demand is not simply another delivery option, it is a neat way of perpetuating a vendor’s hold on the customer and it is facilities management by another name.

    With single tenant all you’ve really done is to place a longer wire between the server and the client and it reminds me of a very old approach called facilities management (FM).  In FM a vendor hosts your data center at another location and takes care of all the management for you.  In this it sounds a lot like any other flavor of on demand but in the FM data center, rather than having a single copy of the software to manage, there are hundreds or even thousands of copies running on a similarly huge number of servers.  The vendor achieves economy of scale, takes a profit and passes some of it on to the customer so that the cost per user these days can approach that of multi-tenant.

    But consider what that means.  By choosing the single tenant approach, you also choose a more limited number of applications that you can work with and often that means locking in to the vendor’s own portfolio.  In contrast the multi-tenant approach gives you no choice over which delivery mode you use but the number and kind of third party applications is greater.  In the Salesforce.com case, we all know that the AppExchange contains about 800 applications from third party developers and all of them are pre-integrated.

    So, in my mind the debate really isn’t about giving customers a choice between single and multi-tenant — that’s a false choice, like saying my way or the highway.  The choice is really about giving customers a choice between vendor lock down and the freedom to choose whatever applications make sense for their businesses.  In my mind, single vs. multi-tenant and the idea of choice are a big side show designed to deflect attention from the success of multi-tenant and save an empire’s worth of old code.

    Published: 16 years ago


    Discussion

    • July 2nd, 2008 at 8:25 am    

      The frustrations I’m hearing with single-tenancy (on-demand or inhouse) models sounds like each customer having a different vision of how CRM should work. In my experience, most of them have not embraced what CRM really is. It isn’t software.

      While I can understand the frustrations a vendor faces in managing their resources, they would also have to believe that they know the customers of their customers. And they would also have to believe that their software is actually a solution to every business’ “CRM” problem. Not CRM the software, but CRM the customer relationship management strategy.

      If you, as a business, have a well developed understanding of your customer and how you will interact with them to create value for both parties, then the software shouldn’t matter that much because most do the same things. As a customer, however, I would like to know that I can address my unique competitive issues when dealing with customers….no matter who is pulling their hair out.

    • April 22nd, 2008 at 8:15 pm    

      Hi Cary,
      I am sorry, I too have suffered through the single tenancy model – 20 big customers paying >$500k a year for the service. I have never quite seen so many heads of hair on fire managing the single-tenancy ASP model as I did in this job. Multi-tenancy is the way so I do want to clarify that I agree. Good design today seems to me should guarantee that a system is designed with multi-tenancy and customizations are more like configurations that can move forward with new releases. I think Salesforce and Ning are great examples of configurable apps powered by a programmable platform that is naturally multi-tenant. We have emulated this model with a set of configurable Web applications for Marketing and Support that integrate with Salesforce in real-time. I love the workd today versus the one with 20 customers on custom implementations, less water to throw on flaming heads (:>

    • April 11th, 2008 at 12:10 pm    

      Jennifer,

      I worked at salesforce.com 2000-2004, and I’ve since worked at a company that tried a single-tenant approach to delivering OnDemand. I can tell you that in addition to the cost benefits of multi-tenancy, which you mention, the real benefit is that multi-tenancy enforces the discipline of not allowing customization on a client by client basis. When clients are on single-tenant deployments, they start demanding special customization, or delayed deployment of upgrades and new releases. The software vendor ends up having to support multiple releases and one-off customizations. Then you go from having every developer working on writing new functionality to having 20 percent of your developers supporting old releases and over 50 percent of your network opps people babysitting unique configurations. That’s bad for the software company, but ultimately it’s really bad for the customer.

      Cary Fulbright

    • April 9th, 2008 at 11:38 pm    

      I have worked in enterprise software off and on over the last 15 years and recently joined a SaaS company that integrates with Salesforce CRM. I pitched and sold enterprise software and am now marketing and selling SaaS – after reading this post would ask to whom does multi-tenancy matter? I think it matters mostly to the vendor from a cost and quality standpoint. It is the SaaS model that I see appealing to the customer becuase of the investment and We were meeting the medium and small customers I talk to every day. There are those of course that have compliance issues and want the software on site and are willing to staff it, but it really is a small percentage of some applications in some companies. We had the debate the other day about would the FBI use on-demand software, I would say yes, in some circumstances to support some business processes.

    Speak Up

    You must be logged in to post a comment.