7 responses to “The Flaw of the Hub-and-Spoke Architecture”

  1. Scott Davis says :

    Evan – Nicely boiled down. Thanks. One point I’d add is that in many network-effect endeavors (like knowledge storage/organization) there are negative returns to scale after some point, due to the exponential costs of coordination among the data/semantics/applications relevant to each additional domain (marketing, finance, ops, etc). The “Economies of Scale/Scope” argument works, but only within a natural range.

  2. Marty Moseley says :

    Hey Evan –
    Nicely done.
    I don’t think the issue is hub-and-spoke architecture per se, but the mis-application of that (proven) architecture.
    I think the problem is that DW architects are trying to schlep around waaaay too much data (most of it never used), and are using a tightly-coupled architecture. It’s overly complex, cannot scale, and can’t meet real-time business BI needs.
    A better way is to use existing ESB or MOM infrastructure to subscribe to the business events that provide data of interest.

  3. george okeeffe says :

    I keep getting error messages where this architecture is being used. “No spoke data”. Do you think that the software is at fault? It is unknown if the data is being transmitted to the receiver and not replied or if the data is not being received by the receiver.

  4. evan levy says :

    It sounds as though your current environment hasn’t implemented a reliable delivery mechanism. Most hub-and-spoke architectures don’t make data available (at the hub) unless they’ve implemented a method to ensure that all data sent is delivered.
    We often find that send/recieve problems are associated with custom point-to-point data migration mechanisms. That’s one of the reasons so many folks have implemented an enterprise service bus (ESB) to replace their custom solutions. An ESB can ensure delivery of all sent data.
    If you have any other questions, you’re welcome to contact me directly.

  5. Bala says :

    When you say “provisioning data directly from the source system is not only justified, it’s probably the most efficient solution.”, Are you referring to Independent DataMart?

  6. Evan Levy says :

    Thanks for the question. Data provisioning isn’t limited solely to data marts. My remarks regarding provisioning data directly from soure systems wasn’t focused at BI systems or data marts. It was addressing data provisioning in general.
    I think it’s important to consider that systems of all types (operational, analytic, etc.) may need non-enterprise, non-standardized data to support their business functions. A DW positioned as a data provisioning hub may not solve every possible data access need (e.g. a CRM system wanting an updated phone number or recent bill payment details).

  7. influence strategies says :

    This is a very interesting article, and I agree with you. With the technology that we have for data storage, businesses must be able to have access to it at all times. But since needs are constantly changing, there is that flaw with the hub-and-spoke artchitecture.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: