presentations‎ > ‎

Cloud for the enterprise, minus the hype

(presentations available below)

YouTube Video

Cloud to reduce costs

The term cloud is quite overloaded, and its definition blurry. For enterprises, cloud often implies that a centralized IT group will become the cloud service provider for the various lines of business, thus making the lines of business cloud service consumers. Given this definition of cloud for enterprises, cloud principles can be applied to help lower costs, and extract more value from hardware, software, and people. The challenge for most enterprises is adopting cloud in an evolutionary way; where an enterprise can incrementally pursue cloud via a series of steps, and each step has its own risk, investment, and ROI. Ideally, the cost savings from step 1 can be used to fund the activities in step 2, and so on. The cloud discussion often focuses on applying virtualization, standardization, and automation to simplify the delivery of IT services. While important, these are only a few steps in the overall cloud journey. The most successful cloud adopters worldwide have done the following, and our discussion on how cloud can help cut costs will elaborate accordingly, and help you better understand the cloud evolution. In certain situations, and evolutionary and step-wise journey may not deliver value as quickly as needed, therefore we'll also discuss revolutionary ways to adopt cloud, as well as revolutionary ways to quickly pursue these steps. IBM's recently announced PureApplication System, an expert-integrated system, is one way to pursue the revolutionary approach. The following steps describe the evolutionary journey (the sequence can vary, and not all steps are required):

  • Step 1: Apply SOA principles to application services, which enables service decoupling, and thus minimizing the disruption of services as the private cloud journey is pursued

  • Step 2: Centralize IT services. This is less of a technical issue, and more of an organizational effort. The more centralized IT services, the better economies of scale and TCO reduction are possible. This isn't a required step, but those that achieved the best TCO did so by creating a centralized IT organization, which served as the cloud service provider for the various lines of business. 

  • Step 3: Consolidate and virtualize hardware resources. Many enterprise have just completed this step, where they've applied VMWare virtualization, or moved up the hardware stack to mid-range UNIX and mainframe systems for better resource utilization. The challenge most enterprises are facing is that they aren't achieving the advertized 40-60% CPU utilization and corresponding cost reduction, rather they're only achieving 12-18% CPU utilization because they're running out of memory. There are a set of actions that can be taken to improve the use of virtualized, shared hardware that can be discussed. 

  • Step 4: Standardize and automate middleware & application services. Now that customers have successfully created a shared hardware platform, the next level of inefficiency is the effort to create, destroy, and recreate application environments. It could take 4-6 weeks (sometimes even months) for enterprises to create an application environment (define the middleware architecture, deploy the application, validate the configuration, and so on). In many cases, up to 60% of the project time & budget are spent on installing, configuring, validating and debugging vendor middleware. This step focuses on creating an application provisioning environment that enables enterprises to quickly (in minutes), repeatedly, and consistently create/destroy/recreate production-ready application environments. 

  • Step 5 

    a: Intelligently manage workloads. Now that we have a shared hardware platform, with automation for provisioning, we need to better align the allocation of resources with business objectives. Workload management becomes a critical capabilities, where the platform can auto-scale application services according to business priorities to meet demands. Resiliency is also critical, where in a shared platform, failures can cascade and cause outages, therefore we need to apply health management services to reduce the probability of failures, isolate the impact of failures, and recover quickly from failures in a way that is transparent to the end user. 

    b: Apply operational analytics. Given that we now have an elastic & dynamic environment,. "It's 10pm, do you know where your app is",  logging & problem determination becomes a challenge. Operational analytics enables enterprises to centralize logging across all application tiers, index and analyze application behaviors, errors, warnings, and so on, and allow the system to pro-actively minimize the number of "failed customer interactions" that occur. 

  • Step 6

    a: Cloud-enable existing applications. Application teams often believe their apps will run well in a shared, elastic platform, and will not make changes until the ops team has provided evidence of architectural problems. Applications typically have the following issues: they don't scale well, thus making it difficult to auto-scale to meet current demands; they have a large memory footprint, thus negatively effecting the density of the system, and increasing TCO; they aren't machine independent, often assuming access to local disk, preventing workload management from optimizing its placement, and so on. There are steps that can be taken to improve the behavior of existing application services, with minimal code change and disruption, to ensure they better leverage the shared, elastic platform. 

    b: Building new, cloud-centric applications. There are interesting design principles, pioneered by born-on-the-web companies like Facebook, Google, etc that can be applied when building new applications. These cloud-centric design principles can be adopted and used as the foundation for new applications, where these applications are designed to take full advantage of an elastic, shared platform. 

    c: Improving development & operations through continuous delivery. Upon creating an elastic, shared platform with automation in place to quickly create, destroy, and recreate application environments, the fundamental ways to develop applications can be optimized. The market describes this as DevOps, though more formally is described as Continuous Delivery; where the application development & deployment process leverages the underlying platform, and dramatically increases the velocity of application development. An extreme example is Facebook, where they have 1000+ concurrent development streams in their production environment, with routing filters in place to control which version is served to a particular demographic. In this model, new versions of an application are validated in the production environment, and can be rolled out in a coordinated and governed way to some or all of the users in the system. Similarly, enterprises can apply DevOps to automatically provision the application to the cloud, execute tests to validate its quality, and upon success, roll the version out in a non-disruptive way to the user-base. 

  • Step 7: Using the right tool for the job. Different types of hardware are good at solving different types of problems. For example, data-intensive tasks run best of z/OS, while memory-intensive tasks run best on commodity x86. XML processing runs better on Datapower than on x86 with Linux. Upon realizing this, enterprises analyze their workloads, and identify hardware accelerators and platforms that are best suited for specific tasks, thus using the "right tool for the job". This ultimately leads to an infrastructure layer that is inherently diverse, and the challenge is building an infrastructure management layer that can simplify the management of a heterogenous collection of hardware, 

  • Step 8: Building a Federated Cloud. Enterprises seek to build a private cloud that spans multiple datacenters, where applications can spillover from one datacenter to another, as well as span datacenters for better qualities of service. The engineering necessary to build a federate cloud are based on the tasks for building an active-active datacenter, which can be quite a challenge. 

  • Step 9: Optimizing the cost of the application platform through hybrid deployments. Certain services, sometimes called "crapplications" should be consumed in an aaS model. Other applications, typically core enterprise services, will run on-premises. Integration with SaaS services must still be governed & integrated with operational processes. With the right platform in place, services can be easily placed in the most cost-effective cloud: public, private, community, etc. 

Cloud to generate new revenue

In addition to lowering cost, cloud can also be used to generate new revenue. Enterprises are externalizing their core strengths: their data, business processes, and so on via API, Web, and mobile channels. These new channels will stress the underlying application platform in new ways: unpredictable & erratic transaction patterns; steep, high spikes in usage; and so on. API management, cloud-centric application platforms, and business models enabled via the "Engaging Enterprise" & SoLoMoCo (Socially-integrated, Location-aware, Mobile-enabled, and Cloud-delivered), enable enterprises to generate new forms of revenue. 

The convergence of cloud, mobile, social, and big data, built using SOA best practices, serve as the technical foundation for these new systems of engagement. An Engaging Enterprise has four technical elements: first, the API set, which is the new public persona exposed to their partners and clients; Second, the applications & services they publish across multiple channels(mobile via an app store, web, api, and so on); Third, an elastic services platform, capable of securely and economically running on cloud-based systems, with the ability to grow to “internet-scale”; and finally, integration with their internal services, which are the foundation of their enterprise SOA, including  their mission critical business processes, transaction services, and business decisions. Each of these technical elements have different life-cycles, where API’s, applications, and elastic services evolve more quickly than internal services. DevOps therefore is a key enabler for supporting the different life-cycles, integrating the experience across development, management, and operations,  and ensuring faster time-to-market. 

Ningbo Logistics is an example of an Engaging Enterprise. Ningbo is working with IBM to fundamentally transform the logistics industry in China. The transformation is enabled through the convergence of: mobile technology, with mobile user experiences, and business optimizations through location-awareness; cloud delivery models designed for internet-scale; new business insights driven through analytics; and a more responsive business through faster time-to-market, enabled by DevOps. 

The Logistics Management Service, an industry cloud, serves as the foundation for a new business network, where Ningbo can externalize their logistics expertise, build an ecosystem of partners on their platform, and provide efficient, cost-effective logistics management as a service. Business model innovation is one of the key attributes of an Engaging Enterprise. Clients can monetize their wisdom and industry expertise in new ways, building new business networks as Ningbo did, and reach clients through new channels. 

Ningbo Logistics Industry Cloud

Related presentations: 

IBM Impact 2012: Future of Cloud Computing, IBM's vision & strategy
Cloud computing enables enterprise customers to fundamentally transform their datacenter. Private cloud enables customers to offer cloud-based services to their lines of business, and hybrid cloud capabilities enable customers to leverage external application and system resources. These services range from compute resources, to middleware capabilities, to application services that may even be running off-premises. This session will discuss IBM's cloud strategy and vision which covers infrastructure and platform as a service, hybrid cloud capabilities and workload optimized systems. Information in the session is grounded on customer experiences gained over the last several years.

Download slides

IBM Impact 2011: Private Cloud [R]Evolution: Customer journeys to the private cloud
Enterprise customers are seeking to drive efficiencies within their IT infrastructures by evolving to the private cloud. This session discusses the private cloud evolution:
covering the challenges experienced after virtualizing; the value of standardized & automated IT service delivery; and extracting more value from hardware, software, 
and labor through workload management. This session will also present a technical perspective on what next-generation enterprise application infrastructures will look
like, and the critical role that private cloud, hybrid cloud, and workload optimized systems will play. 

IBM Impact 2011: IBM Cloud Vision and Strategy
Cloud computing enables enterprise customers to  fundamentally transform their datacenter: Private cloud enables customers to offer cloud-based services to their lines of business; and hybrid cloud capabilities enable customers to leverage external application and system resources. These services range from compute resources, to middleware capabilities, to application services that may even be running off-premisis. This session will discuss IBM's cloud strategy & vision, which covers infrastructure & platform as a Service, hybrid cloud capabilities, workload optimized systems, and is grounded on customer experiences gained over the last several years.

Download Slides
Download Audio

Fixing Virtualization: Getting the CPU utilization you were promised
Despite virtualization, customers are still only achieving 12-18% CPU utilization. The challenge is that they are running out of memory before they run out of CPU. Application Virtualization and Elastic Caching can be used to help customers extract more value from their hardware, software, and people. With this session, attendees will have a plan they can immediately execute, with minimial disruption to operations, and 0 application code changes, to get more out of their virtualized infrastructures

IBM Impact 2010: Enabling Clouds with WebSphere

Client Experiences Adopting Cloud

Snehal Antani,
Mar 2, 2013, 5:49 AM
Snehal Antani,
Oct 13, 2011, 8:21 AM
Snehal Antani,
Apr 26, 2011, 11:59 AM
Snehal Antani,
Mar 18, 2011, 1:38 PM
Snehal Antani,
Mar 18, 2011, 1:32 PM
Snehal Antani,
Oct 24, 2011, 8:27 AM
Snehal Antani,
Oct 18, 2011, 1:06 PM
Snehal Antani,
Jul 21, 2012, 10:18 AM
Snehal Antani,
Sep 13, 2011, 9:32 AM
Snehal Antani,
Sep 6, 2011, 8:15 AM