Experience of Seismotech Company: Yandex.Cloud for variable computing needs
How to process 50 TB of new seismic data promptly
OBJECTIVE
Petroleum databasin
Variable need for computing resources (number of servers, server characteristics: CPU number, RAM, disc space) is a special feature of seismic data processing. Stages of user interactive operations with low demand in these resources change for the computing stages which conduction in reasonable time is impossible without involvement of considerable resources. For instance, pre-stack depth migration of the offshore seismic data over the area of 700 square km was completed on 70 servers (12 cores, RAM 48 GB) in 30 days. Data volume in this project was 50 TB.
Moreover, stages of interactive work and computations are often simultaneous rather than consecutive. So geophysicist can start a long-running computing task and turn his attention to another tasks in the project. The important advantage is in the fact that availability of extensive resources reduces the useless time when the geophysicist is watching the running progress bar on the screen and unable to continue working.
Seismotech Company chose Compute Cloud and Object Storage services to attain two goals. The first is to optimize the cost of renting computing equipment. The second is to minimize the time for computation during the course of processing. It is not uncommon in petroleum industry to see that the project implementation period is even more important for customers than its cost. Availability of essentially unlimited computing resources without booking them in advance and pre-order is a competitive advantage.
Additional value for business is the fact that Yandex.Cloud data centres are located in Russian Federation, notes Dmitry Mosyakov, General Director of Seismotech, Ltd., as almost all Russian oil and gas companies disallow transferring seismic data abroad".
SERVICES
Compute Cloud
Object Storage
SOLUTION AND RESULTS
Research results
The Company's software is the offspring of Russian school of physical and mathematical thought, and the team has been developing it for over 25 years. Seismotech team consists of the Prime software developers and geophysicists who use the Prime to process seismic data. So there are enough specialists here, and there was no need to involve contractors for the project implementation. The qualified technical support of Yandex.Cloud was enough to solve the issues emerging in the course of the development.
"This is the first time for our company to use virtual resources", said Dmitry Mosyakov, Director General of Seismotech, Ltd., "That's why we had many different questions at the outset, and the most important of them related to reliability of data protection and computation productivity. With the experience gained, the calls to technical support became more often related to solving specific situations and problems, when the speed and quality of technical support response is critical. It should be noted that Yandex.Cloud is actively developing, so we have to adapt to the changes taking place, and the information "from the first hand" allows doing it as quickly as possible".
In the technical preview stage and during the first months of use, the developers adapted the software for operation in the Cloud.
This means that the required resources are automatically created for each computation job. In terms of capacity and characteristics (number of cores, RAM, and disk storage), these resources correspond to the specifics of each job. When calculations are finished, the resources are automatically deleted.
It took three to four months to get ready to move. During this period, the required infrastructure was deployed in the Cloud (clustered storage system, server for interactive remote operations, database server). Prime software was also obtaining the cloud API, that is, Prime "learned" to create virtual clusters for computing. Since the infrastructure and software were ready, we started loading all our new seismic data processing projects into the Cloud.
The company's staff is located in Moscow, but employees are able to connect to the Cloud resources from any location. Company's specialists involved in data processing now have the opportunity to access the necessary amount of computing resources on demand and spend minimal time waiting for the computation to be finished.
"Our company has been leasing cluster equipment for the last few years," said Dmitry Mosyakov, Director General of Seismotech, Ltd. "We expect that the transition to the Cloud platform with a per-second billing will reduce these costs by at least 30%".
Find how Yandex.Cloud tells about it