Building bimodal bridges

Gartner has applied a simple framework to complex IT reality. Separating the enterprise into “bimodal” IT operation, Gartner researchers claim that 75 percent of enterprises now experience a bifurcation of IT into two modes, based on surveys of the firm’s client base. But what exactly are these ‘modes’, how do they interact, and what if it were possible to unify what appear as siloed categories of IT in the Gartner schema? Shawn Rosemarin, new Chief-of Staff of Americas Systems Engineering for VMware, sheds light on these questions with a fresh perspective on modal mutual dependencies and platforms of unification.

A key theme at Gartner events in 2014, the concept of “bimodal IT” was developed to support enterprises with legacy IT as they work to meet the challenges of our increasingly fast-paced digital world. The phrase describes the concurrent presence within an organization of two modes of IT. Focused on stability, reliability and efficiency, Mode 1 is a deliberate, consensus-driven, slow moving process that typically involves technology that is core to the business; in contrast, Mode 2 is experimental, agile, less risk averse, fast-to-market and focused on alignment with business units in the rapid evolution of business applications, though core technologies may also be involved.

In the view of one commentator, Gartner’s framework suggests the ultimate dissolution of traditional IT as Mode 1 operators lose out in competition with Mode 2 workers for scarce resources. This is an interpretation that is not dissimilar to Gartner’s own analysis. Of the three quarters of enterprises that now operate according to a bifurcated organizational model, half will fail, Gartner claimed, because they try to apply traditional IT mentality to next-generation applications and services, or fail to adopt new technologies, relying instead on traditional vendors to solve next-generation challenges. Viewed through this lens, “the Bimodal imperative” appears less like peaceful coexistence and more like the ‘withering away of the [current] state’ as the enterprise makes inexorable progress to Mode 2 IT service delivery.

Like many structural frameworks that take time and effort to construct, the bimodal concept suffers from the attempt to pin all reality to its scaffolding. How close does theory correspond to current practice? As Rosemarin explained, organizations in fact typically house three classes of enterprise applications: Systems of Record (ex. traditional ERP systems, which are critical to the functioning of the business); Systems of Differentiation (custom applications that deliver the identity of the company – the way a POS terminal works in a grocery store, for example – which in turn connect back to systems of record); and Systems of Innovation, based on agile architectures and focused on ‘time to market’. These applications can be loosely associated with the bimodal schema – traditional systems of record and differentiation that were built on N-Tier architectures and focused on 100 percent reliability correspond roughly to Mode 1, while Mode 2 is represented by Systems of Innovation.

But at a practical level, these systems are interdependent, and will need to interoperate and be managed, at least for the foreseeable future. Rosemarin asked, “As an operations team, how do I fundamentally operate across these two modes? My Mode 2 innovative application may still need to touch, read or interface with Mode 1 Systems of Record, so I can’t completely firewall these off, and separate them.” A siloed approach to operations, he argued, where traditional IT and innovation teams work in isolation is not tenable: “The challenge is that applications extend across both Mode 1 and Mode 2. The secret is getting the two modes to live in harmony and delivering against their respective mandates of ‘stability’ and ‘time to value’?”

New approaches to bimodal operation

When they try to build systems of innovation, many enterprises are challenged by traditional thinking and processes. As Rosemarin described this problem, “big enterprise organizations have struggled to change the way they bring systems of innovation to market. They have applied traditional methods – the ‘waterfall methodology’ – to building an innovative system to compete with an Uber-type application, and 18 to 24 months later they have their first version and they wonder why the market has left without them. We are now starting to see these organizations take a ‘Lean Enterprise’ (O’Reilly) approach towards their Systems of Innovation and focus on delivering a Minimally Viable Product (MVP) to test market adoption and iterate with quick releases. This change has driven a new architectural paradigm as well as microservices.”

Microservice architectures

One approach to the bimodal conundrum is to leverage the advantages of each IT mode by deploying a common tool set across both, and to ensure that the appropriate integrations between mode 1 and mode 2 applications are in place to build the vertical stack. Traditionally, communication between applications has been provided through Enterprise Service Bus (ESB) technology, which abstracted the mainframe operation to create a layer or extensible platform that developers could code to. But as Rosemarin noted, “with true systems of innovation, ESB doesn’t take you all the way as true Mode 2 applications are now built with microservices.” Designed for speed, microservices represent a fundamentally different approach to integration. “In Mode 1, you basically have database, application server and web server tiers which all scale out as required; in Mode 2, you build an entire service – or ‘mini application’ – as a unified object that runs independently, links to other services and scales up as required,” he explained. The benefit of this approach is that microservices may be linked to other microservices to deliver an application that goes to market, and the big difference is that one service can be changed or removed without driving an impact across the other microservices. In Mode 1, on the other hand, this is not possible because services do not run in a modular fashion – traditional architectures are tightly integrated, meaning any change will likely have an impact on other components.

Developed on modern application development platforms – or PaaS – these microservices are built on ‘code short forms’ that speed deployment time. Microservices also strengthen the link between code and the developer. As Rosemarin explained, the application layer in Mode 1 can consist of a million lines of code, and is integrated into packaged database and app server applications. In this scenario, if there are issues with code in production, support personnel, rather than the code author or application developer is accountable for the fix, with the result that responsibility for breakages is difficult to ascertain. In contrast, microservices can run independently and may consist of only 400 to 600 lines of code that are tightly associated with an individual developer who has responsibility for his/her application. According to Rosemarin, “this whole phenomenon of holding application developers responsible for the code they write – the ‘own what you write’ iteration of DevOps– has introduced the notion that the developer and the operations team can now work together, has helped to ensure they are hand-in-hand in terms of who knows what, who is in the best position to fix it, and who has accountability for that code base or micro service through its lifecycle.” In other words, microservices have helped to create a technology basis for cooperative working process and practice.

In creation of new application services, new development skills are often required. While traditional Mode 1 applications were built on COBOL, C, Fortran, Microsoft’s .NET or even Java, modern languages like Python, Ruby, PHP, etc., which are designed to accelerate development of a microservice, are multiplying with remarkable speed. “There have been more programming languages brought to bear in the last five years, than in the twenty years prior to that,” Rosemarin noted. “And they are all about becoming more capable of leveraging large scale objects or services and building them into the framework so that they are reusable – so that I don’t have to rewrite something that delivers a widget from here to there since it already exists in the platform as a reusable object.” Sounds a lot like the promise of many years ago of “Write Once/Deploy Many” may finally be coming to life!

Control plane bridge

New techniques like microservices demonstrate that the interactions needed to establish a common platform across the different modes of IT are within reach. While these emerging systems are enabling enterprises to improve application change management on the developer front, Rosemarin noted that innovation is also occurring at the control plane layer to connect Modes 1 and 2 apps to enable seamless delivery of IT services. Key functions where the bridge must occur, he argued, are in automation, operations, and chargeback – areas that have been at the centre of a good deal of VMware’s recent focus. For example, the company’s vRealize Suite is a ubiquitous cloud management platform for vSphere and other hypervisors, and for physical infrastructure and clouds, which also works across traditional and modern applications to deliver seamless operations, automation and financial management. With this single control plane, Rosemarin explained, the enterprise can say, “now I don’t have to build siloes based on where the application lives. I can have one platform that allows me the flexibility to build the app where it needs to be, and maintain one common operating environment.” Summarizing, he observed, “Imagine that I’ve got an optimal application platform to run Mode 1, and I’ve got an optimal application platform to run Mode 2. So it’s bimodal in terms of application platform; however, with VMware’s vRealize Suite I can now drive a ubiquitous shared services layer on top that allows one view of operations, and financial management across modes.” This type of thinking retains full architectural flexibility but a drives a single predictable operating model.

The Gartner schema suggests that enterprises with traditional, virtualized environments, which primarily support Mode 1 applications, should focus on driving efficiencies in hybrid IT (as opposed to hybrid cloud), reserving public cloud services such as VMware’s vCloud Air, AWS or Azure as the delivery platform for innovation apps. However, this proposition neglects the synergies that can be derived from a more integrated approach that takes advantage of modern tools. Rosemarin pointed, for example, to lightweight, portable containers like Docker that wrap a code base for delivery of apps/services across the enterprise, which unlike, VMs, do not need to transport the operating system (300 megabytes as opposed to 3 gigabytes, a significant difference across hundreds of thousands of containers). He also described VMware’s recent launch of ‘Project Bonneville’, which makes it possible for enterprises to deliver containers with the feature sets – abstraction, isolation, security – that are available in hypervisor environments. Bonneville is one of the most recent launches out of the company’s new Cloud Native Apps business unit, which is focused on delivering a powerful experience from development to operations for modern applications. At the same time, Rosemarin noted, a key goal of the new business unit is to allow developers to use new tools of choice – Chef, Vagrant, Puppet, Ansible and containers – to drive innovation. Success so far looks like this: VMware has delivered an open source Linux runtime container called Photon; Lightwave unified identity management (which delivers ‘Identify-as-a-service’ across all applications); as well as the Bonneville ‘unity system’, which allows an enterprise running a VM (Mode 1 app) or container (Mode 2 app) to apply hypervisor attributes, including virtual networking, virtual storage and virtual security, with no compromise.

While a ‘customer journey’ that entails progress towards the innovation environment may be a desirable endgame, as with all IT transition, enterprises are likely to embark on the voyage according to unique requirements and run at a pace that is aligned with their specific needs. For the foreseeable future, it’s likely that most organizations will continue to rely on both traditional and modern IT – the reason, Rosemarin remarked, we still have mainframe environments, despite the introduction of x86 years ago. So while enterprises may have little appetite to re-platform Mode 1 apps, Rosemarin predicts there will be a shift towards a “Mode 2 first mentality” as these applications are less costly to build, more agile and quicker to market. In the meantime, tools that can help operators manage through this shift, and until the ‘next big thing comes along’ will prove their worth, delivering the “true currency of IT,” which is the combination of Modes 1 and 2 in unified, seamless delivery.