Spooky HPC Cloud Computing Stats Just in Time for Halloween

By Andy Morris, IBM Cognitive Infrastructure

October 28, 2019

[Attend the IBM LSF & HPC User Group Meeting at SC19 in Denver on November 19!]

Taking the fright out of Hybrid Cloud computing

Spending on cloud computing is growing at a frightening pace. According to 2019 research from IDC, worldwide spending on public cloud services and infrastructure is expected to reach $370 billion by 2022, a five-year CAGR of 22.5%.1 While cloud sounds like a boo-tiful thing, it’s been slower to catch on in HPC circles. Hyperion Research estimates that while 70% of HPC sites run jobs in public cloud, these jobs comprise just 10% of all workloads.2 In other words, although adoption of cloud is broad, it is not very deep. Some think that HPC users are just “fraidy-catscompared to their corporate brethren, but HPC users have good reasons to be cautious.

Cloudy things that go “bump in the night”

Cost – while some mortals view cloud as less costly than local infrastructure, this is true only for occasional use and short-duration workloads. Most HPC workloads consume IaaS services, requiring that users provision advanced witchcraft and spells to manage complex HPC software stacks, significantly limiting potential savings from the cloud. In some cases, cloud computing can be up to 10x the cost of operating on-premises.3 Similar spooky findings from Gartner estimate that 70% of cloud users will overshoot IaaS budgets through 2020.4 Cloud users can find themselves with monstrous cloud service bills that will make their heads spin, while frightening away their limited OPEX budgets.

Daemons (sic) and Zombies are major contributors to cost overruns. Anyone who’s watched a horror flick knows that daemons are notoriously hard to kill. Even powerful charms such as “kill -9” can leave background services running on busy clusters, prevent containers, and cloud instances from shutting down.5 Zombie compute instances and orphaned services are major cost concerns. According to the 2019 RightScale State of the Cloud report, approximately 35% of cloud spending is wasted, and zombie instances and applications that over-subscribe resources are major causes.6

Lock-in – Shifting services to the cloud without an exit strategy is a risky proposition. Like a deal with the devil, users trade short-term convenience for a witch’s brew of proprietary APIs complicated pricing schemes that can leave them locked-in to a single cloud provider. This cauldron of complexity only grows as users add services, deal with challenges such as data security, data movement, egress costs, and inter-instance latency and storage bandwidth concerns that can crush application performance. If users aren’t careful, they can find themselves lost – doomed to wander the clouds for eternity, with inadequate funding, and no earthly compute infrastructure to return to.

[Additional reading: HPC in The Cloud: Avoid These Common Pitfalls]

 

Vile deeds in the cloud – Cloud security is another big issue, especially for sites handling sensitive data. While most users worry about external threats, sometimes the call is coming from inside the house.7 The same tools that quickly deploy cloud-based clusters can rapidly tear them down with catastrophic consequences. Such was the case with the well-publicized Code Spaces hack where an individual gained access to cloud credentials destroying an entire business, removing cloud instances, snapshots, and even data backups in other cloud regions.8 Cloud providers are responsible for the security “of the cloud,” but users are responsible for security “in the cloud.” To avoid being the victim of nefarious deeds, users need to carefully manage credentials and be wary when putting previous data in the cloud.

Before entering the cloud realm

Opening the door to the cloud “ever after” (a ghostly netherworld of data centers outside your control) carries risks. Depending on the cloud provider you plan to summon, you should be prepared. Most IaaS clouds are benign, and for these, simple protective wards such as cloud bursting software may suffice. Clouds that extend higher up the application stack (MSPs, PaaS, and SaaS providers as examples) can be more malevolent leaving you vulnerable to lock-in and demonic upside cost surprises.

For most HPC users, hybrid cloud is a solid strategy. Data is the soul of your organization, and a hybrid cloud lets users control where compute and data reside, easily replicating or transferring data to the cloud when it safe, time, and cost-efficient to do so.

[Also see: Preparing for an HPC Multi-Cloud Future]

Tools that can help protect the soul of your HPC applications from dangers in the cloud include on-premise infrastructure, workload managers, hybrid data management solutions, and multicloud-friendly bursting solutions. Cloud services based on open-source, container-friendly software frameworks such as Kubernetes can take the fear out of cloud computing and ensure that you don’t fall victim to the cloud ever after.

Taking the fear out of Hybrid Cloud Computing

IBM builds the world’s most powerful supercomputers, optimized for HPC, AI and data analytics.9 IBM Power Systems and hybrid cloud-friendly software solutions such as IBM Spectrum Computing and IBM Spectrum Scale can help you operate safely and economically across your choice of clouds and data access methods. Multicloud solutions such as IBM Cloud Private offer Casper-like friendliness, harnessing Kubernetes and enabling cloud-spanning applications that avoid the evils of proprietary APIs.

Ghostly greetings and happy hauntings to our friends at HPCwire and our IBM Spectrum LSF Community members. If you are not yet a member, you can meet other HPC practitioners and learn new spells to help navigate the cloud at the IBM Spectrum LSF User Community.

 


  1. Worldwide Public Cloud Services Spending Forecast to Reach $210 Billion This Year, According to IDC https://www.idc.com/getdoc.jsp?containerId=prUS44891519
  2. Hyperion Research Update 2019 – https://hyperionresearch.com/wp-content/uploads/2019/06/Hyperion-Research-ISC19-Breakfast-Briefing-Presentation-June-2019.pdf
  3. Hyperion Research Update 2019 – https://hyperionresearch.com/wp-content/uploads/2019/06/Hyperion-Research-ISC19-Breakfast-Briefing-Presentation-June-2019.pdf
  4. How to Identify Solutions for Managing Costs in Public Cloud IaaS – https://www.gartner.com/en/documents/3847666/how-to-identify-solutions-for-managing-costs-in-public-c0
  5. A Daemon refers to a background process that runs continuously on modern operating systems – https://en.wikipedia.org/wiki/Daemon_(computing)kill -9 is a common command on Unix-like operating meant to kill a running process.
  6. RightScale 2019 State of the Cloud Report from Flexera – https://info.flexera.com/SLO-CM-WP-State-of-the-Cloud-2019
  7. When a Stranger Calls, the horror film – https://en.wikipedia.org/wiki/When_a_Stranger_Calls_(1979_film)
  8. Murder in the Amazon Cloud – https://www.infoworld.com/article/2608076/murder-in-the-amazon-cloud.html
  9. IBM Summit – the most powerful supercomputer in the world https://en.wikipedia.org/wiki/Summit_(supercomputer)
Shares

Taking the fright out of Hybrid Cloud computing

Spending on cloud computing is growing at a frightening pace. Read more…

" share_counter=""]
Return to Solution Channel Homepage
HPCwire