I recently joined Red Oak Consulting with a background in Enterprise IT. In the world of Enterprise IT, the arguments about whether or not to move to the cloud are all but over and organisations building new Enterprise IT solutions in 2019 are doing so in the cloud. The world of HPC with its higher demands on compute, networking and I/O subsystems is only now starting this journey that Enterprise IT started over a decade ago. So, what does the future look like?

Some observers believe cloud will only be applicable to a subset of the HPC industry, and right now they are right. Cloud can’t operate the high-end workloads effectively, let alone deliver value for money. However, the primary focus of these users is not the technology, but the value delivered by it. There is no intrinsic benefit for these users in operating HPC systems themselves, but it is currently a necessity. As cloud matures, I believe capabilities will increase and benefits will grow, and the economics will improve further. I think all this will happen quite quickly by IT standards.

The HPC move to cloud has already started, albeit very slowly. However, once it gets going, like Enterprise IT, it will be an autocatalytic runaway reaction. More customers in the cloud will increase cloud market size and reduce on-premise market size. The growth will impact the economics of both cloud and on-premise markets and the targeting of new capabilities. In only a few years, the latest technology will be available in the cloud first. A while later, the latest developments may only be available in the cloud. These market changes in turn will accelerate the move to the cloud and that is what creates the runaway effect. This is already evident in Enterprise IT: just compare on-premise and cloud machine learning tools/services.

Like Enterprise IT, customers will be concerned about security. This may worry the HPC world a bit less (but shouldn’t) and there will be those who believe putting our stuff “out there” will be less secure than running it on-premise. However, since your datacentre is already connected to the internet, it is already “out there”. Now the question becomes, who follows better security practices, your organisation or a major cloud provider? The number of accreditations like ISO27001 that the major cloud providers such as Microsoft, Amazon and Google have picked up in the last few years should provide a clue.

Customers will look at cloud costs and think they can do it cheaper, a lot cheaper! How can cloud be so successful if it is more expensive? The reality is, most organisations don’t look at true costs that often. When was the last time your organisation priced up an on-premise server and included a proportion of the office building costs or the cost of procurement itself? Lots of small overheads build up and sit in cost centres across your organisation and are just “absorbed”. Doing a true total cost of ownership analysis, the costs are not only comparable, I believe they will often be favourable. The economics of scale aim to cover the operating costs of the technology, the service and the profits and have enough left over to lower the cost to customer. There maybe a few, a very few organisations who operate their on-premise technology as efficiently as cloud providers. However, I am left wondering, if an organisation can operate more efficiently than Microsoft et al, shouldn’t they be providing that value as a cloud service to others?

If you haven’t seen it already, you should read my colleague’s post on Financial Modelling in HPC here: https://www.redoakconsulting.co.uk/financial-modelling-in-hpc/

After costs, customers will start examining the detailed functionality pointing out the differences with on-premise and that the cloud doesn’t support their way of working. They will emphasise the uniqueness of the need and how the commodity approach of the cloud doesn’t meet their needs. Undoubtedly, all organisations have a special and sometimes totally unique core to their business. However, unless you are Microsoft, Amazon, Google or a handful of others, running datacentres, servers, platforms and networking is not your core business. You probably don’t normally generate your own electricity or write your own operating systems and you shouldn’t run your own cloud.

If the journey continues to progress as it has in Enterprise IT, it is around this point in the journey that your organisation should become aware of the real challenge with cloud. That real challenge is your organisation. The doubts around security, the lack of understanding true costs and believing your organisation’s needs are totally unique are symptoms that your organisation itself isn’t quite ready. Your organisation needs to change.

I am old enough to remember being in a company when email was being introduced. If you wanted an email account, you needed to fill in a form providing a business justification to get an account. The organisation took some time to realise the value of the opportunity.

It is hard to understand something new, especially if it changes the framework you might use to understand it. Most cost models comparing cloud and on-premise use a cost model for on-premise computing because it has worked well for years.

There is a natural and a necessary resistance to change, permitting positive change and prohibiting marginal or negative change. Unfortunately, the judgement on the nature of a change often requires new skills and insight. Judging cloud by the metrics and comparisons used to examine on-premise solutions just doesn’t work. A terabyte of storage for two months costs more than a terabyte hard disk. Of course, the cloud storage is redundant, software configurable and billed as spaced used. The physical hard disk has no connectivity or chassis, requires someone to install it, has no redundancy and costs the same whether it is full or empty.

Moving to the cloud requires different approaches, requiring changes in the organisation’s processes. These changes in processes are, in part, changing the organisation itself. Changing an organisation obviously should be strategically driven and cloud needs to be an integral part of the IT strategy. However, as the phrase attributed to the management consultant Peter Drucker goes, “culture eats strategy for breakfast”. The organisation’s culture needs to be receptive to the change in strategy for it to be effective.

Moving to the cloud represents a change in skills, staffing, budgeting, collaboration, security, technical capabilities, dependencies and most significantly, culture. Change is hard and organisations that move successfully do so by putting enough focus on culture. Who knew the changes in moving to cloud were going to be so pervasive…

There is a deeper point here. A significant proportion of organisations would see their ability to adapt as key to their success or even survival. If they are struggling to adapt to well signposted change such as the move to cloud, how will they react to unannounced changes in their industry?

This is a long-term view but the move to the cloud will be rapid by IT standards. However, it will also be a phased move with smaller, simpler workloads moving first. Complex, mission critical workloads will move only when the capabilities have improved vastly. In all cases, the time to start planning is now. Procurement cycles can be long, but culture change can take even longer. The change is pervasive, and preparations in numerous areas need to be started early to maximise the chance of success. Some may not move to the cloud but I think that will leave them at a disadvantage with a growing capability gap between what is possible on-premise versus cloud-based solutions.

We are not there yet, cloud cannot support the high-end workloads many organisations require. However, the cloud services are changing, and organisation need to change to properly assess the opportunity. Military strategist and United States Air Force Colonel John Boyd talks of an OODA loop. OODA stands for “observe”, “orient”, “decide” and “act” and decomposes the anatomy of the decision-making process that pilots and organisations go through. “Orient” is the step that organisations have done slowly in Enterprise IT and the world of HPC could learn. The organisation itself needs to change at a cultural level to understand opportunity and embrace it. I believe cloud is as inevitable for HPC as it was for Enterprise IT and it will only accelerate.

David Prebble – Managing Consultant