I want to share an insideHPC slidecast interview about an Enterprise-speed Cloud run we did for HGST (a Western Digital company). But first, let’s take a quick look at the business benefits of Cloud cluster computing. Because this particular run was record-setting and included a Fortune 500 manufacturing company, we’ve received a lot of attention & press coverage about it (The Register, ZDNet, EnterpriseTech, HPCwire, etc.). A lot of the focus has been paid to how Cycle Computing was able to enable a 70,000+ core run so quickly. But behind the scenes, it’s maybe even more important to understand what the business benefits are that are driving Cloud cluster computing adoption.
The Business Challenge
The business challenge for HGST was common to other manufacturing companies:
How do we accelerate common business functions, to help reduce design cycle times?
Solution & Benefit:
For HGST, speed & agility for its research and development efforts are of paramount importance. When Cycle Computing found out about the run on a Wednesday, it was able to accomplish all of the set-up, and complete the entire Cloud computing run by the weekend. This particular workload traditionally has taken about one month to complete in the past running on internal clusters. By leveraging Cloud computing, the workload ran from start to finish in about eight hours. And the total cost of this run was only $5,500.
- Manufacturing workload: Design Optimization of >1 Million designs
- Project Agility: From project idea, to completion – about four days
- Time Savings: From 1 Month to 8 hours
- Value: Only cost $5,500
There are a lot of exciting things about this Fortune 500 Cloud computing run, including:
- Leveraged three AWS regions
- 729 teraFLOPS
- 70,908 cores (peak)
- Deployed 50,000 cores in 23 minutes
More details can be found in this blog post: HGST buys 70,000-core Cloud HPC Cluster, breaks record, returns it 8 hours later.
A slidecast interview from insideHPC also provides even great insight and details behind this record-setting Enterprise Cloud computing run. In the slidecast, insideHPC’s Rich Brueckner interviews Cycle Computing CEO Jason Stowe about what this run means for Cloud and high performance computing (HPC).
Below are some highlights from the interview:
What this means for Cloud HPC
JASON STOWE: “… the removal of constraints on the kinds of questions you can ask by making [Cloud cluster computing] really efficient and really cost-effective – for someone to grab a large amount of capacity and ask questions that normally they wouldn’t be able to ask … is a huge change for the industry as a whole.”
Comparing Old ‘Own Your Infrastructure’ Model to Cloud Cluster Computing
RICH BRUECKNER: “… you think about the company the size of Western Digital, and certainly if they had the will, they could pay $30 million for a machine of this size … and probably another $10 million for the building to power and cool it … but [just comparing] the $40 million outlay versus $5,500 … “
JASON STOWE: “Yeah, it’s kinda hard to argue …”
Go Invent, Go Discover
JASON STOWE: “… don’t think about capacity, think about the science. Think about engineering. Go Invent, and discover. Do better work, and we’ll build the infrastructure to answer that question. Whatever that question is, we can … achieve the scale required … to be able to answer it.”