Digital transformation is not a buzzword. IT has moved from the back office to the front office in nearly every aspect of business operations, driven by what IDC calls the 3rd Platform of compute with mobile, social business, cloud, and big data analytics as the pillars. In this new environment, business leaders are facing the challenge of lifting their organization to new levels of competitive capability, that of digital transformation — leveraging digital technologies together with organizational, operational, and business model innovation to develop new growth strategies. One such challenge is helping the business efficiently reap value from big data and avoid being taken out by a competitor or disruptor that figures out new opportunities from big data analytics before the business does.
From an IT perspective, there is a fairly straightforward sequence of applications that businesses can adopt over time that will help put direction into this journey. IDC outlines this sequence to explain the maturity phases of this journey. This sequence starts in familiar territory such as web and cloud; then continues toward newer terrain, like open source and next-gen apps; and finally it enters the new world of advanced data analytics, predictive analytics, and cognitive analytics. This, in practical terms, is what it means from an IT perspective to digitally transform.
IDC also believes that not including POWER8-based servers in a comparative evaluation of one- or two-socket offerings is detrimental to the quality and objectivity of that evaluation and may ultimately prevent a business from selecting the most suitable and cost-efficient infrastructure for its digital transformation journey. This white paper takes a closer look at IBM’s OpenPOWER LC servers built with the POWER8 processor and with such acceleration technologies as NVIDIA Tesla P100 GPUs, NVIDIA NVLink, and CAPI.
The Digital Transformation Era
Digital transformation is not just a buzzword but the approach by which enterprises drive changes in their business models and ecosystems by leveraging digital competencies.
Resisters (14.2%) make up the rear guard, and they provide weak customer experiences and have a defensive posture toward digital. The next category is the Digital Explorers (31.8%) that offer digitally enabled products, services, and experiences albeit inconsistently and not well integrated. The third group are the Digital Players (32.4%) that provide consistent but not truly innovative products, services, and experiences.
The fourth segment are the Digital Transformers (13.6%) that are leaders in their markets, ©2017 IDC #US42253217 3 providing innovative products, services, and experiences. And at the front lines are the Digital Disruptors (9%), who are remaking existing markets and creating new ones to their own advantage. As the data discussed previously indicates, more than 50% of business now fall in the Player, Transformer, or Disruptor categories. The other half are at risk of losing their competitive edge.
A Road Map for Digital Transformation on Volume Servers
Digital transformation requires that a business redefines how it generates revenue and monetizes products and services, with a strong focus on contextualized and personalized customer experience and business efficiency. It also means centering product and services innovation on customer experience and changing the way the company works to be as agile and dynamic as possible — first by gaining insights and responding to the market with innovative products; then by developing products based on the ability to predict market behavior; and ultimately by creating products that completely reshape the market, possibly in disruptive ways.
Volume Servers for Stateless Applications
One- or two-socket infrastructure is very well suited for stateless applications such as VDI and many web applications. “State” is the ability of an application to retain such personalized information as to who a user is or what button the user just clicked. Stateful applications need a way to keep that information either in a client, the web tier, a database, or in cache. Stateless applications do not require such data. Stateless applications are therefore suitable for scaling out as they do not present the challenge of sharing the user’s state between the servers. This makes it easy to spin up more servers as needed for such applications.
Volume Servers for Web Applications
Traditionally, the biggest concern with web applications has been response time and scalability. End users are demanding instant response times, and for web applications with high traffic volume, this presents a challenge. Scalability helps with improving response time by either adding more servers (one- or two-socket servers) or by adding more — or more powerful — CPUs in a single-node server (scale up). Both approaches are successfully used today.
Volume Servers for Next-Gen Applications
IDC predicts that by 2018, enterprises pursuing digital transformation strategies will more than double software development capabilities and that two-third of their coders will focus on strategic digital transformation applications and services. Next-generation applications, a key component of the digital transformation effort, are distinctly different from traditional applications. They use different programming languages and are designed differently. Next-gen application developers require extremely flexible compute capacity, scalability, infrastructure redundancy, performance, storage capacity, bandwidth, and uptime as well as low-cost per unit of compute, storage, and bandwidth. Some typical characteristics of next-gen apps are:
- They are designed using what are called “microservices,” meaning that instead of a monolithic application they consist of a collection of small services that are each responsible for a distinct process, can be deployed independently, and can hook into each other via APIs. A major benefit with regard to scaling an application built with microservices is that a developer only needs to scale the microservices that require scaling, while those that do not can remain unaffected. Monolithic applications need to scale in their entirety.
- They are often stateless and designed with the expectation that the infrastructure they run on is not guaranteed to be resilient. By being stateless, they can simply move to another node in a one- or two-socket environment in case of a failure without the loss of state. If the application itself malfunctions, it simply gets replaced, not fixed.
- They scale based on demand. When demand increases, more instances are spun up automatically. And they are refreshed much more frequently than traditional applications, sometimes several times a day.
The Final Step to Cognitive
Volume Servers for Cognitive Applications (Machine Learning)
IDC defines cognitive systems as a technology that uses deep natural language processing and machine learning, resulting in understanding as well as the ability to answer questions and provide guidance. The system hypothesizes and formulates possible answers based on available evidence, can be trained through the ingestion of vast amounts of content, and automatically adapts and learns from its mistakes and failures.
Required Parameters for Infrastructure on Volume Servers
For IT to traverse the one- or two-socket server road map effectively, all the way from stateless web applications to cognitive and predictive workloads, the server hardware that the infrastructure is designed with needs to meet a variety of requirements. The road map illustrates that as these workloads evolve, they cannot be successfully deployed on low-end commodity hardware. The demands on the infrastructure will be too great.
The Current Market for Linux-Based Volume Servers
The current market for one- and two-socket servers running Linux, is growing rapidly. IDC forecasts that from 2016 to 2020, the number of shipments of such systems will increase with a 7.5% CAGR. In 2020, the Linux share of the one- and two-socket server market will have grown to 42.6% of the total market in terms of vendor revenue.
A question that is on many people’s minds is, whether IBM would actually provide comparable HA features on its one- or two-socket server line as on the enterprise-class scale-up systems, with which it has built its reputation for reliability. What differentiates POWER8 with regard to HA resides in its architecture and has been carried over from its enterprise-class systems to its one- or two-socket line.