Download the PDF of this Case Study.

THE CUSTOMER: Incorporated in 2002, Language Weaver develops software that automates the translation of languages. While traditional machine translation systems apply thousands of written grammatical rules to translate documents, Language Weaver uses a new, statistical approach that increases the probability of an accurate translation via computer and delivers more natural-sounding translations. Language Weaver’s statistical machine translation software (SMTS) “learns” language translations by identifying patterns and relationships within previously translated or parallel texts of any two languages. Trained on a collection of parallel texts in electronic, scanned paper, or other digital form, Language Weaver uses statistical algorithms to align the text and assess the correlation of words and phrases. Language Weaver learns the translation patterns for every word and phrase in the training data. It can then use this knowledge to translate new, previously unseen text of the same language pairs. To translate a new document from a source to the target language, the software references this translation knowledge base, generates several translation possibilities, rates them statistically, and chooses the best option based upon the context of the document.

Language Weaver’s pioneering technology is the result of several years of invention and development at the University of Southern California’s Information Sciences Institute (USC/ISI) by Language Weaver’s founders and their students. The organization currently offers 18 language translation software modules.

THE CHALLENGE - FILE SERVICE RECOVERY TIME AND FILE SERVICE FOR MAC AND WINDOWS ENVIRONMENT: Within this data intense computing environment, Language Weaver was running into difficulties in that its infrastructure—which Wong describes as a direct-attached “grow your own” environment was not well suited to keep up with the expansion. The servers were growing and the number of cross mounts was increasing. As a result, the organization was seeing a high number of idle mounts, which was causing problems with NFS file service. Beyond that, with a storage growth rate of tens of terabytes per month, management issues for disparate islands of storage were becoming more time consuming and problematic.

THE SOLUTION - UNIFIED STORAGE THAT EASILY SCALES: Language Weaver was in a unique situation. While for many organizations, performance is the most significant area of importance, for Language Weaver capacity impacts performance ability. “We’re a highly automated shop with all our data being computer accessed data. Our customers don’t hit our system,” said Wong, noting that customers are best served if Language Weaver has ample storage to collect data and run experiments and ultimately produce better language pairs.
“Our servers are accessing the data in read mode or write mode 24 x 7 x 365,” he added. “Speed is important, but capacity comes first.”

To meet its unique requirements, Language Weaver turned to Datalink. “Datalink had an established reputation with another company that recommended its services to us,” Wong described. After reviewing Language Weaver’s environment, Datalink recommended that the organization consolidate its storage with Network Appliance unified storage. Ultimately, Datalink implemented a NetApp® NearStore® disk-based system at Language Weaver. The organization utilizes the CIFS protocol for connectivity for its Windows applications and the NFS protocol on the Solaris and Linux side. They also utilize Fibre Channel protocol for some of the back-end servers for Linux. “It really didn’t make sense for us to invest in a high availability, higher speed main storage system because we needed more storage for the dollar. For us, this was the right solution for the maximum amount of data that we want to crunch,” said Daniel Marcu, chief technology and operating officer. “We operate in an extreme environment,” he added, noting that the organization is successfully pushing the design and execution of the technology to its limit.

THE BENEFITS - MANAGEMENT, CAPACITY AND SCALABILITY: With the new solution, ease of management is a major benefit, Marcu said. “We no longer have to manage the file system servers anymore—it’s basically become management of one NetApp system. In that sense, we’ve eliminated a lot of potential points of failure,” said Marcu. “We looked at other technologies but the Datalink/NetApp solution was the obvious choice,” Wong said. This has been a new relationship for all parties involved and so far it has been smooth and flawless, he said. “As far as implementation, it has fit seamlessly into the spot it was meant to fill,” elaborated Wong. “In the execution of our IT plan, Datalink has proven to be a valuable expert in supporting our storage needs.”

“In the end, this solution helps us speed our time to market for new language pairs and improve the quality of our existing language pairs. That’s the direct impact to the customer,” Marcu said.

LOCATION: Marina del Rey, CA

SOLUTION: NetApp R200 disk-based storage system

DATALINK PROFESSIONAL SERVICES: Analysis, Design, Implementation, Support

INDUSTRY: Automated Language Translation Software