IT trends for 2015: Power Systems, security and flash-as-memory

These three technologies will continue to evolve in 2015.

man walking calendar dates appointments
Amana Images

Unlike some of the larger IT research and analysis firms that place analysts into confined “specialty” areas, we at Clabby Analytics are free to research whatever topics we choose across the broad spectrum of IT systems and software. A few years ago, for instance, I focused heavily on mainframe technology and converged systems; this past year, on security, Flash and Power Systems. Here is a look back at a few technology trends from 2014 that will make an even bigger impact in 2015.

Security

2014 was the year that many enterprises have started to really concentrate on security – as dozens of large companies have experienced major security intrusions (including Home Depot, Target, Apple – and most recently Sony). Billions of dollars in litigation are now at risk; enterprises see the writing on the wall and are becoming very active about reducing their security exposure.

Early in the year I focused on risk and fraud as they relate to enterprise security. Unfortunately, the “bad guys” are winning when it comes to fraud (with individuals and organized crime defrauding people and governments of almost $3.5 trillion in revenue and savings each year). U.S. Army General Keith B. Alexander -- the director of the National Security Agency (NSA) and chief at the Central Security Service (CSS) -- calls cybercrime: “the greatest transfer of wealth in history.” 

What’s being done to reduce fraud? Plenty. For instance, IBM has a broad, highly integrated stack of security, infrastructure and analytics products designed to detect, identify and deal with fraud. 

In the realm of advanced data security, organizations are taking action to protect data and personal identifiable information. Some of these activities include data governance, encryption key management, data masking/redaction, data activity monitoring, data loss prevention, new encryption activities, database access control and database auditing. IBM has created an entirely integrated stack to control and manage data security, although enterprises can also purchase several point products and eventually achieve the same result.

The lesson learned in security this year should be firewalls, anti-virus and encryption are no longer enough to protect against fraud and intrusion. You need to take steps to watch your users more closely; to better protect your data (using some of the techniques described above; and you need to harden your systems, networks and associated infrastructure.

Flash-as-Memory

As 2014 wound down, Flash (a non-volatile computer storage medium) was being used increasingly as a means to augment main memory. IBM’s Coherent Accelerator Processor Interface is a great example of a fast Flash-to-CPU superhighway. Microsoft's SMB Direct (that combines Microsoft file transfer facilities with Violin Memory Flash arrays and Mellanox switching) is another excellent example. [Disclosure: Microsoft is a client of Clabby Analytics.] 

New systems configurations are coming to market that have created bus superhighways (fast paths from Flash to the CPU) – enabling enterprises to build low-cost, very large “memory” (Flash acts like slow memory) systems that cost 1/3 as much as traditional primary memory DIMM-based designs.  

As an example of the type of systems (very large memory/low cost) that I’m talking about, consider IBM’s Power System-based Data System for NoSQL – a configuration that can place 40TB of Flash within a single rack. This Flash acts like extended memory (albeit slower memory) – and helps users get results from large databases extremely quickly. More and more machines like this will be coming to market in the near future to better and more cost effectively serve Oracle/IBM/SAP in-memory databases.

My philosophy on microprocessors/servers

Before I discuss the final trend I think will be big in 2015, it is important to understand my philosophy regarding microprocessors and servers.  On microprocessors, no single microprocessor does every job optimally, so it is important to choose a microprocessor that can execute a given workload most effectively. 

There are very significant differences between x86, POWER, and z processors in terms of how they handle parallel, serial and compute-intensive workloads.  Plus there is a new generation of system designs that are using general purpose microprocessors plus job specific microprocessors to execute workloads ever faster (these include graphical processing units [GPUs], Advanced Risc Machines (ARM), and field programmable gate arrays [FPGAs]). 

As for systems designs, some systems are designed for heavy-input/output, others are designed for parallel processing and still others strip out management layers and focus primarily on performance.  

Given my basic philosophy that one single processor/server doesn’t handle all workloads optimally, I cover IBM z System mainframes and IBM Power Systems – as well as x86 processors/servers – pretty closely. Here is the conclusion I have reached.

Power Systems

IBM’s Power Systems is one of the most interesting stories in the industry right now. About a year ago Power Systems revenue started to decline (very significantly) as users sought less expensive Linux-based solutions. A research analyst friend of mine and I told IBM that there would be a big market swing toward Linux when IBM announced its POWER7-based systems in 2010, but IBM was too busy scooping up a lot of the Unix migration business as Unix users bailed off HP Itanium-based servers and Sun systems. In short, IBM got caught flat-footed as the market moved strongly toward Linux-based soultions. [Disclosure: IBM is a client of Clabby Analytics.] 

Luckily, IBM has taken corrective actions to get Power Systems back on the right track. Those include:

  • Replacing top management (a strong move. Doug Balog moved over from the mainframe business and knows exactly how to revive a troubled product line – he was heavily involved in the mainframe Linux movement which has been highly successful).
  • Rolled out the low-end Power Systems first – with special Linux-based Power Systems at attractive prices with better performance than equivalently priced Intel servers.
  • Introduced an Integrated Facility for Linux that runs on Power Systems. This mix of software makes it possible to configure large numbers of Linux servers on Power Systems.
  • Added little endian support to the Power Systems environment. “Endians” refer to byte order – and Power Systems can now read bits in both ascending and descending order. This is important because most Linux applications are in the little endian byte order – so having the ability to read little endian opens-up the great big world of Linux applications to Power Systems. Almost all Linux applications can now be easily run on Power Systems with a simple recompile. And people buy applications first.
  • Power Systems are the basic system design for IBM’s Watson cognitive computing environment. Watson has generated a lot of discussion about the future of computing and is poised to start gaining some decent marketshare. If it does, it will drag a lot of Power Systems revenue along with it.
  • POWER8 was introduced this year – it is an extremely powerful microprocessor capable of executing 8 threads per cycle (as compared to Xeon’s 2 threads). As such, a POWER8 server can do a lot more work than a system based one and an equivalent number of Intel x86 processors – saving customers money in software licensing, hardware and energy.

In short, given huge investments in Watson and in making Power Systems into powerful Linux machines, Power Systems are well positioned to recover in 2015.

Summary Observations

There is one final trend in systems design that I expect will have a bigger impact in 2015. This is the emergence of “accelerated systems” designs where multiple types of microprocessors are used within a system to accelerate performance. Field programmable gate arrays (FPGAs) are now being used to accelerate communications; graphical processing units (GPUs) are now being used to speed parallel computing tasks – while general purpose processors have added new big data processing facilities. 2015 is shaping up to be an exciting year in IT.

Copyright © 2015 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon