For more and more companies, the connection to customers through mobile apps is mission critical. In Australia alone, mobile penetration in 2015 reached 137 per cent, according to telecom research firm BuddeComm – so there are now more mobile phones than people.
As users bank, shop, work and manage their lives on mobile, they touch each provider’s system dozens more times per day than they ever did through websites and laptops, creating a deluge of transactions and requests.
In 2004, each mobile user generated less than a single mobile transaction per day. By 2014, that number had hit 37 transactions per day, and continues to spike.1
At the same time, users expect instant response times, 100 per cent uptime, and completely reliable transactions, or they may switch to a competitor with the touch of a screen. A security breach that results in stolen customer data can devastate trust and loyalty and send a business into crisis mode, as we’ve seen in many news stories of late.
The result is that many CIOs can feel a new migraine coming on. Companies increasingly want to offer products and services built on open source software.
Yet in this era of smart phones, constant connectivity, high-speed transactions, and sophisticated hackers, open source apps running on a typical expanse of servers could lead to a cataclysm.
Until recently, mission-critical apps were likely built on more traditional, costly and inflexible software from major providers. That’s been changing.
Modern IT professionals want the versatility, simplicity and lower cost of constructing apps in open source – MariaDB, MongoDB, and Ruby on Rails – running in containers on Linux.
And that leads to the headache. To run open-source software and meet the rocketing demand of the mobile era, the only solution has been to add thousands of servers.
Database partitioning, or “sharding,” is a common technique for scaling out a database that has become too large to fit within a single server. However, sharding is complex in practice, and carries risks such as higher latency and a lower level of data consistency.
The complexity of a server farm can also shackle innovation. While open source lets IT professionals build apps quickly, deploying the server infrastructure to make it work can eat up valuable time and resources.
Moving all workloads to the cloud was the mantra of three years ago. But many organisations that adopted a largely off-premise model have been reconsidering. Costs can balloon and performance issues can result in a poor user experience.
Some enterprises are coming around to a new solution – one that might’ve seemed counter-trend just a couple of years ago. They’re running Linux and open-source applications on a single system, the kind of platform designed for high-speed, secure, and the ‘always on’ transactional demands of a financial institution or retailer.
One of these systems is IBM LinuxONETM, which can scale to handle tens of billions of interactions daily with sub-second response times. This system can process 54 per cent more data than distributed systems for model building, according to tests using Spark-Perf benchmark suite.
It can also help simplify the creation of innovative intelligent apps when using the Apache Spark big data processing engine; and lets companies tap into Node.js to develop high performance web and mobile apps without purchasing additional hardware.
Running open-source applications on a single high-transaction system can also help solve the "sharding" problem, freeing up resources and helping to make it easier to change workloads and add features.
A single system can help speed up compression and encryption. Imagine compressing Spark RDDs (resilient distributed datasets) or Docker containers at high speed with little impact on CPU consumption.
This can help free up CPU cycles that can be used to perform more analytics. Encrypted transactions can run faster. The computer can do more work while staying flexible and agile.
Discovering business intelligence requires the use of advanced analytics, such as Apache Hadoop and Apache Spark. Such capabilities are commonly deployed on distributed commodity hardware. However, when the applications run on a single high-transaction system, the software can process more data and come up with better insights.
It is estimated that only 30 per cent of the value and effort that goes into creating a mobile app is visible on the front-end app itself. Around 70 per cent of the value is provided by the software and systems supporting mobile workloads.2
Ultimately, using open source software on systems designed to handle high transaction loads can help take the pressure off CIOs and could translate into a competitive advantage for the business.
Wesley McDonald is General Manager, IBM® Systems Hardware Australia & New Zealand
1. New Metrics and Insights for a Mobile World, pg 4, Dr. Howard A. Rubin, CEO and Founder, Rubin Worldwide, Professor Emeritus City University of New York, 2015
2. Based on IBM internal statistics - http://www2.themsphub.com/rs/creationagencyibm/images/MobileFirst_Platform_Overview.pdf