2020 in Review | Reflections and Predictions

After an undeniably turbulent fifth year in business, our CEO, Simon takes a look back at the predictions we made at the start of 2020, and looks ahead to what’s next for YellowDog.

Our Fifth Year as YellowDog

It seems hard to believe that nearly a year has gone by since we celebrated our fifth birthday back in January. It has been a year filled with challenges, across all aspects of life. From businesses having to adapt to remote working, to parents having to juggle childcare and their careers, and to everyone battling the trials associated with restricted socialisation. However, it hasn’t all been doom and gloom. We’ve had more time to spend with our families, greater opportunity to explore different hobbies and, with less commuting, more downtime to unwind from the busyness of a working week.

As the end of a year commonly marks a time of retrospection, I thought it would be interesting to revisit the predictions we made at the start of 2020, and to compare them to our predictions now.

Cloud Infrastructure: Then and Now

In 2018, Gartner forecast that worldwide public cloud service revenue would reach $240 billion by 2020. Two years later, their forecast showed an increased revenue figure of $258 billion. The pandemic has accelerated cloud adoption beyond what analysts initially predicted, as companies were forced to look at how they could remain agile in volatile market conditions.

However, even though cloud adoption has seen rapid growth, there is still some way to go until cloud is used to its maximum efficiency. For example, we haven’t yet seen evidence of cloud and on-premise infrastructure falling under the same umbrella, as synonymous sources of computing power that can be dynamically allocated depending on a workload’s specific needs. Businesses still rely on relatively manual processes to provision their resources, and therefore suffer the opportunity cost of not using the best source of compute, whether that is on-premise, or in the cloud.

I believe as companies continue to look for ways to further advance their operations, intelligently provisioning compute on a workload-by-workload basis, will become more important.

One customer that benefited in particular from our HPC solution was OMass Therapeutics. To assist in accelerating their drug discovery process, we deployed a 162,120 core supercomputer in ten minutes using 2,000 Spot instances, spread across four geographically dispersed regions, to virtually screen 447 million compounds in less than a day. With other technologies the provisioning time alone would have been 13 hours, before the work even commenced.

This considerable reduction in time taken to perform this number of molecular docking analyses, will help to significantly accelerate the time it takes OMass to identify lead compounds. It will help diminish the risk of error at an early stage in the drug discovery process, which will hopefully bring life-changing benefits to those in need sooner.

Given the significance of our work within Life Sciences, I predict our technology will continue to open up the concept of HPC to many different industries.

As mentioned above, when the first lockdown was implemented in the UK, cloud usage rocketed as businesses rushed to maintain normalcy in uncertain conditions. However, I predict that next year cloud spending will be less manic. With the initial panic over, businesses will consider how they can best make use of the capacity they already have on-premise and in the cloud, before they look to increasing the number of servers they have at their disposal.

Data Centre Virtualisation

I predict data centre virtualisation will move entirely to Linux. Today, the lion’s share of virtualisation is based on Linux Hypervisors, with Microsoft Azure being the odd-one-out with the Windows-based Hyper-V.

In September, Microsoft submitted a series of patches to “create a complete virtualisation stack with Linux and Microsoft Hypervisor”. This is likely to be successful and I would expect Azure to start switching to Linux-based virtualisation by the end of 2021.

Arm-based Servers

I think we can expect to see a lot more Arm-based servers being run by the Hyperscalers. Back in 2018, VMware announced ESXi-on Arm. Amazon Web Services has its own Arm-based CPU, the Graviton2 and Oracle Cloud Infrastructure will also launch 160 core Arm-powered servers in 2021.

I expect to hear a lot more about Arm-based workloads in the cloud next year.

Overcoming Obstacles

While we may have started this year with a very different 2020 in mind, I’m grateful that the team have remained healthy and well and that we have managed to overcome and adapt to the obstacles we’ve encountered.

I sincerely hope everyone has a happy and safe New Year. Until 2021…