Irena Bojanova - Home
Understanding Multi-Tenancy II
Irena Bojanova
MAY 03, 2013 08:00 AM
A+ A A-
painting of cowboy wrangling wild horses

It's the High-Tech Wild, Wild West out there!
Although the Cloud Computing marketplace is still chaotic, it is:

  • Exciting
  • Fast-growing
  • Full of opportunities

Continuing our discussion on multi-tenancy, note that Gartner places multi-tenancy close to the peak of expectations in the Hype Cycle for Cloud Computing, shown on Figure 1.

figure 1

Figure 1. Hype Cycle of Cloud Computing, 2011 (Gartner).

Gertner stresses also on the relation between elasticity and multi-tenancy, talking about "elastic multi-tenancy" as a distinct characteristic of cloud computing.

Rapid elasticity is defined by NIST as: "Capabilities can be elastically provisioned and released, in some cases automatically, to scale rapidly outward and inward commensurate with demand. To the consumer, the capabilities available for provisioning often appear to be unlimited and can be appropriated in any quantity at any time."

It is important applications to have an easily customized and used graphical interface or series of workflows, but it is good to know also about the multi-tenancy architecture itself, as there is where the main risks would come from. Three general methods for achieving multi-tenancy (IT Pro) are via a database, via virtualization, and via physical separation — presented in Table 1. Note that multi-tenancy could also be nested — a provider could host tenants, which are independent software vendors (ISVs) with their own tenants (subtenants of the cloud provider). A multi-tenancy platform with proper architecture would have provisions to support subtenants as well as tenants.

Table 1. General Methods for Achieving Multi-Tenancy

Multi-Tenancy Description Cost
Via database Database and configuration, with isolation provided at the application layer. Least costly.
Via virtualization Virtual machine technology, providing a hardware emulation layer over the real hardware. Multiple copies of server operating systems are run within one physical machine, while sharing physical hardware (e.g. network cards and disk storage) between virtual operating system instances. Might reduce services costs and expenses, but is more costly compared to multi-tenancy via databases.
Via physical separation Resources are provided to tenants individually — each tenant uses only dedicated hardware. Most costly.

Figure 2 illustrates the general multi-tenancy architecture with customer integration on application layer, infrastructure layer, and data-center layer. Note that Infrastructure-layer and application-layer customer integration for multi-tenancy are new additions to the cloud computing topology design.


figure 2

Figure 2. Overview of General Multi-tenancy Cloud Architecture (IT Pro)

  • Application-layer multi-tenancy — can compromise security, because application methods and database queries can store/ access data to/ from different user accounts; but can offer significant cost savings, when implemented properly. Architectural implementations concern both SaaS and IaaS layers.
  • Infrastructure-layer multi-tenancy — dedicated stacks of software are deployed from infrastructure layer to specific customers (a separate stack for each customer account). Hardware requirements depend on the actual service use.
  • Data-center-layer multi-tenancy — provides the highest level of security, when implemented correctly. Table 2 shows the three approaches for managing multi-tenant data.

Table 2. Approaches for Managing Multi-Tenant Data

Multi-Tenant in Description Comments
Separate database Each tenant data is in a separate database. Simplest approach to data isolation. Highest level of security.
Shared database and separate schema Multiple tenants data is in the same database, but each tenant has his/her own set of tables within a schema, created specifically for that tenant. Cost efficient approach with high level of security.
Shared database and shared schema Multiple tenants data is in the same database and the same set of tables. Most cost efficient approach.

Anyone have thoughts or sources that will help readers understand multi-tenancy? Please share here!

Irena Bojanova

Irena Bojanova, Ph.D., is Founder of IEEE CS Cloud Computing STC, an Associate editor of IEEE Transactions on Cloud Computing, and an Editorial Board Member of IEEE CS IT Professional. She is a professor and program director, Information and Technology Systems, at University of Maryland University College, managed academic programs at Johns Hopkins University and PIsoft Ltd., and co-started OBS Ltd., (now CSC Bulgaria). Her current research interests include cloud computing, web-based systems, and educational innovations. She is a member of the IEEE and can be reached at


[%= name %]
[%= createDate %]
[%= comment %]
Share this:
Please login to enter a comment: