Multitenancy: Panacea and Problem

Multitenancy is a buzzword getting a lot of attention these days. Over at ZDNet, Eric Lai started quite a conversation looking at what he felt were Four Big Problems with Multitenancy from an enterprise application space. Frank Scavo replied that the post was a Mischaracterization of Multitenancy in his blog.

Multitenancy is an interesting word. You won’t find it in a dictionary (unless you use Wikipedia’s definition as your dictionary – and even that definition is still considered suspect by the Wikipedia editors almost two years after first appearing). It’s a word that conveys an obvious meaning on a vendor’s or expert’s slide, and probably shouldn’t need a strict definition. But we’re in the technology business, and folks like to debate what words mean. Even the National Institute of Standards and Technology (NIST) definition of cloud computing contains only one reference to ‘multi-tenant’, and it’s only used in the definition of the term ‘resource pooling’. Multi-tenancy and cloud computing looks at the definition of multi-tenant by starting with the definition of a tenant – in which there is some ambiguity. Let’s ignore the strict definition of the word multi-tenant and use a more pragmatic definition. A multitenant solution allows two tenants to share resources with the express desire to lower cost.

With that background, here’s three things to ponder when considering the next claim that a product is or isn’t great because it is multitenant:

  • Business requirements matter
    I’ve commented on this before, noting that Multitenant ERP Solutions are a fundamentally bad idea. Multitenant is a great cost saver for the technology end of the solution, but cannot afford to ignore the impact of SOX and other regulatory bodies on the pace and timing of change in ERP systems at real businesses. A vendor simply can’t force a SOX-compliant company to change their business processes just because one of the other tenants is now ready to change theirs. On the other hand, HR/Payroll solutions are ideally suited for multitenancy. Regulatory changes imposed by Federal, State, and Local authorities are not optional and must be deployed to all users of the HR/Payroll applications (i.e. all tenants) as quickly as possible – ready or not, here comes the regulatory change! Processes that are not subject to the rigors of auditing paralysis like ERP or to non-stop mandatory updates like HR/Payroll can be viewed through the narrow technical lens of technical costs. Sales Force Automation and CRM solutions can be ideal candidates for multitenancy because there is nothing preventing them from sharing the same applications, and most users will want access to the new features as quickly as possible.

    Bottom line: Multitenancy is not a panacea – business requirements will drive whether multitenancy is appropriate for a given application (and to what degree).

  • End user prices are market-driven
    There are many pundits who will claim loudly that multitenant solutions inherently cost less to end-users than single tenant applications. These pundits who truly believe that cost and price are related in every market skipped some days in their economics class. Commodity markets like platform-as-a-service are largely driven by price. To the extent that two software applications are virtually indistinguishable from a feature/function standpoint, the price will be the deciding factor, margins will be squeezed, and the low cost provider (who can, by implication, offer the lowest cost solution) will win. This logic breaks down quickly in application software markets. First, enterprise applications are still highly differentiated – even if only by brand. Customers are willing to pay more for one solution than another – price is rarely the deciding factor in a solution purchase (despite the claims of software sales professionals). Second, the assumption that price and cost are connected is proven invalid when end-users no longer pay to use the software. Consider, acquired by Intuit, who provides a cloud-based solution with substantially the same functionality as Quicken, but at no cost to the end user. Business models emerge to help companies turn their assets into profits. New business models may leverage multitenancy to save money and increase profits, but are not dependent on adhering to any strict definition of multitenancy to make money.

    Bottom line: Multitenancy, where it can be leveraged, lowers costs for vendors. The impact on the end customer depends more on the market and the business model.

  • Economics requires choices
    A multitenant implementation of a software application may be just as easy to secure as a single tenant application using the same tools, but the risks associated with deploying an application as multitenant are undeniably higher. In the most extreme case, Payroll data combines names, addresses, salary and financial information, and Social Security Numbers in the database. A large, multitenant HR/Payroll application may be just as secure as a smaller, single tenant HR/Payroll application. But from the risk standpoint, I’m reminded of Willie Sutton. When asked why he robbed banks, he replied ‘that’s where the money is‘. By putting all of this data into one giant database with one encryption key and one super-user password, we make it all that much more attractive to hackers. The best way to keep hackers out of your HR/Payroll database is to run HR/Payroll applications on computers that are not connected to the Internet in any way. I’m not recommending that this is an appropriate solution (see my comments about the advantages of HR/Payroll as a multitenant application above), I’m only noting that buyers with different risk profiles will see the alternatives differently. We know from Econ 101 that there are buyers with different preferences. Ideally, the application is fundamentally the same (from the end user standpoint) whether it is delivered in the low-cost cloud-based multitenant mode, or in the hyper-secure unconnected on-premise single tenant mode.

    Bottom line: Customers have varying preferences, with varying tolerances for risk competing with varying appetites for growth. Some people drive cars, some drive trucks. Mutitenancy does have different risks, and is not appropriate for all applications at all customers.

Our take: We believe multitenancy creates cost advantages for ISVs like HarrisData, and intend to leverage multitenancy in whatever forms provide the most value to our customers. However, a religious commitment to multitenancy ahead of real customer requirements is just pandering to investors.

This entry was posted in Cloud computing. Bookmark the permalink.

Leave a Reply