Mainframe - comeback time?

16 April 2018

Not sure if it’s the circles I’m frequenting, or a genuine phenomena, but is mainframe making a comeback? There seems to be a lot of press about it at the moment, as people re-centralise their IT into Data Centres, and the need for massive computation of data is required, it has gone from being the dead duck of the computing world, to potentially the next new retro-trend.

I went to university between 1990 and 1994 and was lucky enough to do a sponsored course that allowed me 3 days a week at IBM and 2 days a week at Uni. In my 3 days a week, I was 3rd line support for IBM’s National Office System (NOSS for short, I can’t remember the last “S”) which ran on its PROFS / VM operating system. It basically provided desktop services through a green screen terminal from a mainframe back end.

At the time, it was revolutionary. Everyone had email and instant messaging, an “A-Disk” for their personal files and shared spaces for team collaboration. We had service machines, which sat and performed functional tasks on a scheduled basis, and performed word processing by coding in HTML, using tags to embolden texts and/ to create unordered lists. Life was good.

This was possibly the best grounding for me in a future in IT. All of the subsequent concepts that are mainstream today were based around the foundations established in that late ‘80s / ‘90s era where I grew up. To a large extent, the 1980’s NOSS mainframe was running a VDI service with core Exchange and Office functionality built in, from a single centralised, virtualised and shared “private cloud” service in a data centre in Portsmouth. Sounds a bit cloudy now.

Then through the 90’s and early 2000’s, people didn’t think it was cool anymore. Big expensive lumps of tin, sat in a data centre, weren’t as much fun as a little lump of tin sat under your desk. GUI’s were obviously better than green screen, but fundamentally performed the same business function, just in a better user interface. Everyone dashed to be client/server, the networks became paramount, and mainframes were the reserve of the big banks who couldn’t afford to change their applications to a client / server architecture.

I think it’s a testament to IBM’s clever scientists in the 1980’s that these monoliths are not just still processing significant portions of the worlds financial systems, but also that they have the potential to rejuvenate to be the future of the computing again. In recent product announcements, IBM continues to invest in the “zSeries”, recently launching a rack mountable unit that will continue to drive high end processing but in a standard enterprise rack. They are targeted at service providers who have a volume of distributed systems and are looking to centralised, or to attempt to compute the vast quantities of complex machine learning and big data analytics from a centralised architecture. What goes around comes around.

Its probably fair to say that the current mainframe footprint in UK industry is fairly small (I’ve been quoted about 200-300 customers, but not sure how accurate that would be?). But if you ignore the statistics for a short while, there is a perfect storm brewing that could suit mainframe:

- Cloud has made centralisation of compute resources cool again.

- Shared and scalable infrastructure is pretty much what mainframe was designed to support, especially on a bureau model.

- The graphical presentation of data can easily be performed by desktop machines now, but the mass crunching of complex data sets still needs enormous processing power which suits the mainframe architecture. The combination of the two is a pretty powerful alliance.

- Network bandwidth are now ubiquitous, which means your processing can be performed “anywhere” and served to you across distance – centralisation works

- Moores Law may start to struggle soon, new (or dusted off old….) architectures may be required to keep advancement in technology working.

Which brings us nicely onto Quantum Computing. People who re-centralise all of their processing into a single architecture are probably best positioned to capitalise on Quantum, should it ever make it out of the lab and into mainstream. A migration from client/server to mainframe services could be a useful stepping stone to leveraging the power of quantum in a decade or two?

Death, taxes and the circularity of IT trends are the only certainties in life. Even if mainframe doesn’t make a resurgence, I for one am enjoying reminiscing about the JCL Batch Jobs of my youth, REXX parsing and building the “NOSS Bomb”, a simple loop which bombarded my Uni colleagues with repeating instant messages to ruin their day. Bring back the mainframe if that’s what it was all about!

We use cookies and other technologies to allow us to remember you, improve our service and display relevant ads to you. To accept cookies, continue browsing, or view our Cookie Policy to find out more, including how you may withdraw your consent.