HOWMAN SEISMIC SERVICES

Howman Seismic Services is an emerging seismic processing business. Formed in 2016, they have continuously grown, developing new software tools and algorithms. Their business is processing seismic data from raw field data, and producing final migrated images of a subsurface. These include, but are not limited to, near-surface model building, velocity analysis, residual statistics, noise removal, and migration. In simpler terms, Howman performs seismic data processing, which includes applying specific signal processing algorithms to large quantities of data. This data is often used by large oil companies in the exploration of oil, who rely on Howman to compute power.

Challenges

Howman approached DiGiCOR wanting to both improve compute power and resolve workload problems. They expressed that they were expecting an increase in workload and wanted to receive expert advice on what systems they could purchase within their budget. The goal was to improve HPC capacity.

In addition to facing computing power and workload problems, Howman was also facing limitations caused by their current systems. These limitations included taking eight hours to turn around jobs. Howman highlighted that if the option was there to turn around jobs in two hours instead of eight, then they would want to do that. The reason is, that they would be able to proceed with other workloads quicker because of faster job turnover.

Howman’s intensive algorithms run in parallel on all available threads, with a requirement for a high core count. A high core count was also necessary because it was preferable for Howman to have data flowing to the fewest servers, as this would mean that they could allocate a higher core count to one server, specifically to deal with data being transferred. Ultimately, the goal was to use these systems to run HPC floating-point calculations.

Their other requirements included a cost-effective solution. Lastly, Howman wanted to improve their green behaviour by reducing the number of servers in use. This created a challenge for DiGiCOR as we needed to provide a solution that would increase Howman’s current performance, while also cutting down the number of systems.

How We Helped

DiGiCOR provided Howman with three 1U ASUS-based AMD EPYC 64-core servers.

ASUS was recommended for four reasons. Firstly, Howman wanted a reliable server. Secondly, Howman needed a system that would be compatible with the operating systems they needed to run, which included Linux. Thirdly, given global logistic demands, ASUS was able to fulfil Hodman’s request given the client’s tight development schedule. Lastly, ASUS was recommended for its scalability. Howman at the time of enquiry wanted to upgrade their nodes in a stepwise manner.

AMD EPYC 64 core processors were recommended to allow Howman to confidently transition to AMD EPYC, DiGiCOR provided Howman with a proof-of-concept server. This enabled the client to run tests and analyse performance before committing to an order. After testing and evaluating the systems, Howman was satisfied with the results and decided to proceed with their order.

Results

The solution provided enabled Howman to match capacity in terms of compute power and workload, therefore allowing them to process data and algorithms on HPC clusters. In addition, with the high 64-core count delivered by AMD EPYC, Howman was able to see 5x improvements in performance.

This solution enabled Howman to achieve their objectives. Firstly, they were able to turn jobs around faster. This meant that jobs that ran in eight hours could now run in two. Having single machines that could run very fast jobs allowed Howman to do small test examples of bigger jobs and experience a fast turnaround. This also allowed the client to run bigger jobs on the HPC cluster once testing was refined. Secondly, due to ASUS’s flexible fan control to reduce noise levels and maximise power consumption, Howman experienced zero overheating. Howman in relation to this stated that not only were they able to adopt greener behaviour because they required fewer servers than previously, but there was also less power being consumed for cooling.

Thirdly, the ASUS server was easy to use and did not require maintenance, which was praised by Howman. They stated that once they put the servers into place, the servers could handle the workload and they did not have to worry about them. Fourthly, being a compact solution, ASUS gives Howman the flexibility to add more servers to the HPC cluster in the future, without constricting space. Furthermore, the solution was scalable to accommodate growth and enable Howman to add more systems in the future as intended. In addition, ASUS was able to fulfil the client’s tight schedule, which was once again praised by Howman who stated, “Prompt delivery” as being one of the benefits of this solution.