Technology, IT Infrastructure

A look back at the path to converged infrastructure

Blog-post by,

In the IT infrastructure space, there is always a limited amount of money that CIOs can spend on infrastructure and therefore a struggle between competing parts of the stack for budget. What was simpler in the mainframe days became less expensive, but more complex. Applications are supported by an infrastructure stack of servers, networking and storage; each with its own functionality. Through the 2000's, while companies like HP and IBM sold and bundled all of the pieces of the stack, each silo had unique hardware, software and full list of features. External storage arrays from companies like EMC and NetApp looked to pull intelligence (and the associated dollars) away from the servers. Functionality such as replication and backup could be placed at the operating system, in the network or in the storage array. A challenge of a “best of breed” solution is that independently choosing each component will typically lead to an inefficient utility of each resource.

Server virtualization, as led by VMware, attacked the inefficiency of server usage. While the typical pre-virtualized server would run at ten percent utilization (or less), server virtualization allowed multiple physical machines to be consolidated as virtual machines on a smaller number of physical boxes that run at higher utilization. Consolidation is a step towards convergence, which is bringing together multiple resources into a common infrastructure. The blurring of the lines between the various physical compute, networking and storage components has run in parallel (and intersected with) the virtualization trend. Blade server technologies also overlap the compute and networking domains.

Convergence is about more than consolidation; it is about gaining efficiencies through the merging of technologies. In the networking space, the discussion of convergence is about pulling multiple traffic types onto the same pipe; this has been done with IP telephony and is happening with storage through various Ethernet-based storage protocols (such as NFS, iSCSI and FCoE). Convergence of the full compute/storage/network stack has been driven by a number of vendors. Cisco has done this through its UCS solution that requires creating solutions with storage partners NetApp and EMC. HP has been driving the Converged Infrastructure messaging for a couple of years and is delivering a number of solutions (including Matrix, CloudSystem and VirtualSystem) based on existing and acquired technologies (including 3COM for networking and 3PAR, LeftHand and IBRIX for storage). In HP’s case, the line between storage and servers is blurring, allowing shared storage to be part of a compute platform. In many ways, a virtualized and converged environment has some of the same characteristics of the mainframe era, but in a more flexible and cost effective manner that creates building blocks for a more service-oriented infrastructure.


Some additional resources from the Wikibon Community on convergence:

Convergence Moves Up the Stack to Applications and Greater IT Efficiency

The Transformation of HP (Infographic)

(1) (1)

Would you like to comment on this content? Log in or Register.
Paul Calento 255 Points | Sun, 10/30/2011 - 18:20

Converged infrastructure is linked very closely with IT modernization. Specifically, the need to leverage the investments of the past for future value. Rip and replace doesn't work, nor does discrete technologies that don't work with each other. The challenge with gaining efficiencies, is measuring them. Most customers don't have the time to figure that out. To Paul Wandrag's comment, the issue of lock-in is a real one, but there are a fair amount of standards that limit this issue. It does underscore the real role that "trusted vendor" (an innovation x-factor?) plays in the decision to move forwared with a converged infrastructure strategy.

--Paul Calento

(note: I work on projects sponsored by and HP)

Paul Wandrag 0 Points | Wed, 10/26/2011 - 16:33

Convergence is a tight-rope act between single vendor lock-in and the flexibility to adopt the best of breed du jour. I have always been in the mindset of good enough for higher order platform components, best of breed for middle order components and commodity for lower order components. This seems a little counter productive but it's the middle order that keeps everything together. Agility is required at higher orders with more flux and cheap as possible for the lower and higly matured and commoditized lowest order.

Pearl Zhu 90 Points | Mon, 10/24/2011 - 16:36

Hi, Stuart, nice posting, hopefully virtualization-based and cloud-driven convergence help organizations achieve the higher level of agility, implement higher percentage of hardware utilization, energy efficiency, and modernize the legacy infrastructure.thanks

John Dodge 1535 Points | Mon, 10/24/2011 - 15:11

My coverage of IT started in the days of the mainframe and mincomputer. But the capacities were miniscule. They did not seem so at the time. Ah yes, the world of DASD and VSAM.....

Thing was, storage devices then were only matched to the same vendor's mini or mainframe .... You could not converge or consolidate those environments (until the advent of the VM OS). If you added capacity, they got bigger.

It's much different today, but we still compare today's technology to what occurred in the past to understand what's happening.

Such comparions, IMO, are largely invalid...if for no other reason that mainframes cost millions or were leased for tens/hundreds of thousands of dollars a year - mindboggingly, in 1970s and 80s dollars!!!!