Accelerating seismic performance and reducing costs

Seismic processing algorithms are becoming more and more complex. In order to take full advantage of today’s high performance processors, steps must be taken to remove the delays in getting the seismic data to the processing core. Design innovation and the incorporation of the latest advances in cluster technology has enabled Sun Microsystems to develop a solution that delivers the best of all worlds — performance, low cost, density, and environmental efficiency.

Using Hybrid Storage Pools, Sun’s ZFS optimizes performance

“Sun’s Seismic Data Processing solution allows you to get more value from your investment, increase productivity, reduce project cycle times and lower the risk”

Traditionally, the emphasis with high-performance computing (HPC) applications has been on CPU performance. Today, however, the emphasis is shifting. In a recent IDC survey of medium to high-end HPC users at the IDC HPC User Forum, not only were more than 90% of respondents concerned about I/O in general, but they also expected further I/O constraints in the next three years. [1] Seismic processing is no different.

Seismic processing algorithms are becoming more and more complex and can definitely take advantage of today’s multi-core multi-socket HPC nodes, but these nodes are becoming more and more constrained and less efficient (more idle cpu time) by limited access to the data to be processed.

Oil and Gas companies need a High Performance Computing solution that can scale with the rapidly increasing volumes of data to be processed and stored yet take up less space, use less power and cost less to cool and manage. Understanding that the biggest challenge in Seismic Data Processing is the large volumes of data needing to be processed and analyzed, Sun’s Seismic Data Processing solution allows you to get more value from your investment, increase productivity, reduce project cycle times and lower the risk. There is now a solution that maximizes throughput performance, saves space with dense storage and requires less cooling — Sun Seismic Data Processing Solution.

Design innovation and the incorporation of the latest advances in cluster technology has enabled Sun to develop a solution that delivers the best of all worlds — performance, low cost, density, and environmental efficiency. Sun’s Compute Cluster technology is the best in class and provides the dual socket, multi core, racked or blade systems with significantly faster I/O and the dense storage solutions you need to perform. One of the keys to improving seismic processing performance on HPC systems is improving storage performance, avoiding CPU idle time and improving the efficiency of the CPU core. This has the dramatic effect of improving the ‘time to result’ for no increase in CPU clock rate.

Oil & Gas companies are facing the following key challenges with their seismic storage environments:

o I/O performance bottlenecks
Causing CPU idle time while waiting for storage systems to complete I/O. Keeping high volumes of data flowing to and from the CPU is key to fast time to solution and requires low latency access to storage. Even more challenging is the fact that performance and cost are intimately connected. What companies are really looking for are solutions that provide the most I/O throughput per dollar.

o Explosive growth of data
New seismic algorithms and higher fidelity data and analysis requirements are driving creation of more data per site, making scalable I/O processing an important requirement for today’s seismic processing solutions. In addition to processing greater volumes of data, today’s seismic solutions often must maintain data for longer periods of time and the data must be shared throughout a global community of users.

o Unsustainable costs
A new more cost-effective approach to storage is needed where general-purpose systems and open-source software can be used to deliver the necessary performance while keeping storage costs within budget.

The need to manage growing volumes of data over time is also driving seismic storage toward multi-tier storage solutions. As data becomes less active, it should be archived to a lower cost media such as slower disks and then to tape where it consumes less power and cooling resources.

The trends of faster and faster processors and increasing needs for storage bandwidth, to feed these processors, and for storage capacity are expected to continue. Oil & Gas companies must find new and more cost-effective ways to get additional performance out of their seismic storage environments while simplifying deployment and management of the storage infrastructure.

Recent advances in the production of flash technology have made solid-state drives (SSDs) and flash array products much more cost-effective, enabling a new approach to storage architectures. While SSDs are more expensive than mechanical drives, they are non-volatile and significantly cheaper than DRAM. They also offer much higher performance and greater power efficiency than hard disk drives, making them a strong alternative that provides a means to rebalance system and storage I/O performance.

Sun’s Open Storage solutions utilise open-source operating systems and open-source management tools and applications to enable rapid innovation and help reduce costs, the same way that the shift from proprietary, custom-built, massively parallel processing systems to horizontal clusters of low cost servers has dramatically reduced the cost structure of HPC deployments.

Lustre file system offers near linear performance scaling

Sun Storage 7000 Unified Storage Systems take advantage of flash technology and open storage techniques to deliver high performance and up to 75% lower cost than traditional storage solutions.

Sun has also fully integrated solid-state drives (SSDs) based on flash technology into its new Sun Blade, Sun x64, and Sun CoolThreads servers, enabling up to 65x better response time than traditional hard disk drive solutions. This dramatically improves CPU efficiency (reducing idle time waiting for I/O), and improving reliability (no spinning disk) and reducing energy consumption and heat production (no disk).

Using Sun’s Lustre™ file system together with Sun’s storage to build high performance compute systems allows you to handle the rapidly growing and difficult to predict data volumes which have resulted from new and advanced seismic data collection techniques.

Lustre provides a way to easily scale the number of cluster nodes on a single file system, improving performance and reducing cost. Sun’s Seismic Data Processing solution prevents bottlenecks and has all the data processing power and data management efficiency to meet today’s seismic requirements and scale to meet future seismic requirements.

Sun’s solution for Seismic Data Processing is a cluster of low cost computing systems using parallel processing. Sun’s dual node, quad socket, multi-core blade servers with flash memory based system ‘drives’ couples with dense open storage solutions, and fast interconnects to give you the best in class High Performance Compute Cluster solution in the industry. Increase your productivity by reducing project life cycle by up to 40% utilizing your current cost, space and power budget. Performance, cost, density and environmental efficiency are all benefits of Sun’s Seismic Data Processing Solution.

Learn more about what Sun’s Seismic Data Processing Solution can do for your organization today. Visit us at www.sun.com/solutions/hpc/industries.jsp.

Note:
[1] HPC Storage and Data Management: User/Vendor Perspectives and Survey Results,” IDC, Doc # 206967, June 2007.