EMC Chief Technology Officer John Roese spoke in Australia recently about a trend that causes some degree of complexity in data center design: application fan-out.
This concept is best described using the cylindrical analogy below:
This diagram is meant to highlight an application-to-infrastructure stack that has been fairly common for a long time:
- Applications: For many years, E-mail, ERP, and CRM were three key enterprise apps.
- Application Frameworks: These apps were delivered in virtual machines.....
- Virtualization Frameworks: .... using the VMware framework
- Topology: In the latter part of last decade, private cloud became predominant.
- CPU: The industry standardized on x86.
- Infrastructure: Customized infrastructure appliances in the data path were common
- Media: Spinning hard drives (HDD) dominated
- Protocol: Block and file were the predominant protocols.
This cylinder, top-to-bottom, was fairly well understood. Strong industry expertise for building this stack developed over time. This expertise was challenged, however, as both the application (upper layer) and infrastructure (lower layer) evolved. The diagram below introduces application fan-out in the context of this same stack:
The upper section reveals that there is now a broader spectrum of apps delivering business value. These apps vary widely, from mobile applications (located far away from the data) to real-time analytic apps (where applications are embedded within the infrastructure).
This variety can also be seen in the proliferation of containers that hold applications, all the way from the Linux kernel capabilities (LXC) shown on the left (e.g. c-groups), to PaaS capabilities shown on the right (e.g. Docker's higher-level, application-based services).
And finally, the virtualization frameworks range from OpenStack, to VMware, to Microsoft. All three of these (and others) are likely to be found within the same data center.
Why discuss this form of application-fan out?
Because it can render the data center unmanageable.
The next post will discuss (a) how infrastructure fan-out further complicates matters, and (b) the key approach to get this complexity under control.