Accelerate Engineering Workloads
Test of Time: More Engineering, Faster with Avere for Cloud and Scale-out NAS Optimization
Teradyne customers rely on the company’s test equipment to accelerate production of reliable electronic devices. In a similar critical-path dependency, Teradyne design engineers rely on the performance of internal IT infrastructure. Slow down engineering workloads, and you slow down innovation and product delivery. No one likes to be the bottleneck.
As a solutions engineer and the senior IT infrastructure architect at Teradyne, John DeBella owns responsibility for delivering IT resources to meet production demand. “The pace of development and the large-scale infrastructure requirements of a global engineering team can make it challenging to maintain on-demand compute and storage capacity. To meet requirements, we leverage Avere Systems Edge filers that allows us to derive greater performance from the existing storage infrastructure, as well as to take advantage of cloud-based resources to expand the capacity and throughput of our engineering infrastructure as needed. Using Avere we’ve been able to shrink runtimes by one third at an 85 percent savings over the cost of traditional NAS. Avere solutions help ensure that, in spite of unpredictable workload demand, IT infrastructure never becomes a bottleneck to production.”
Challenge: Support Major Engineering Project with Fast Startup and Indeterminate Scope
Teradyne has for decades been a dominant automated test equipment (ATE) supplier, helping customers more rapidly deliver electronic devices that function at peak—first time, every time. With innovation at the core of both its own and client businesses, Teradyne constantly strives to develop ground-breaking engineering solutions for the most challenging test issues.
DeBella says that when Teradyne kicked off a high-priority engineering project—one of the company’s largest-ever and one without precedent in terms of IT infrastructure requirements, the project presented both capacity and budget challenges. Engineering’s best estimates called for 50 percent increase in compute power with storage infrastructure that could serve data to keep pace with workload demand, with just a two month lead time. “At that scale and with a condensed timeline, expanding our existing NAS environment appeared to be cost- and time-prohibitive. The other problem was correctly sizing the infrastructure without any historical requirements data—we had never before supported a project of this scope. The one certainty: IT infrastructure could not be a bottleneck, either during ramp-up or at any point during the project’s anticipated two-year lifecycle.”
Solution: Avere Systems for Engineering Workload Acceleration
Rigorous evaluation—including benchmarks run on Teradyne’s high-performance compute (HPC) cluster using production instrument tests and datasets—demonstrated that an Avere FXT Edge filer solution would deliver required performance. DeBella notes, “Avere met all of our selection criteria, including performance and the ability to run dozens of sophisticated electronic design automation (EDA) applications without any changes to our programs or user file access. We dropped in the Avere as a very cost-effective performance layer between the compute farm and our existing NAS infrastructure—and we had it running in production in time for project kick-off.”
Today an Avere FXT cluster in Teradyne’s North Reading, Massachusetts, headquarters front-ends hundreds of terabytes of NetApp capacity. The Avere cluster provides high-speed I/O to support an HPC cluster running in excess of 200,000 jobs each day and a technical user base of some 400 engineers, as well as enterprise-wide business systems and users. An Avere cluster deployed at a Teradyne Design Center in Agoura Hills, California, provides high-performance file access to a local HPC cluster, as well as low-latency remote access to the North Reading HPC cluster.
Benefits: Scale-out Performance, Lower Cost, Cloud Access
7X Peak Workload
DeBella reports that the Avere kept pace with the engineering workloads driven by the new project. “Deploying the Avere resulted in an immediate reduction in compile and simulation run times—by 33 and 28 percent, respectively. Overall Avere enabled a 3X increase in sustained workload and up to a 7X peak workload. Just months into the project, engineering requested the addition of nodes to the HPC cluster, and we were able to rapidly scale the Avere cluster to serve data at a rate matched to the heavier workload demand. Scale-out was easy and as importantly, non-disruptive. Using Avere helped us successfully meet workload requirements for the life of the project, ensuring that IT was indeed never a bottleneck to production.”
85 Percent Lower Infrastructure Cost
DeBella points out that the new engineering project put pressure on more than IT infrastructure. “We had to tap our entire next-year budget to support this project and secure financial approval at the highest levels of the company. Building out our traditional NAS footprint was not an option, it was simply too expensive. Using Avere we were able to support the new workload at just 15 percent of what we would have had to spend on traditional high-end NAS. Avere savings also allowed us to provide sufficient tier 2 storage capacity for the project—which was well beyond original estimates—in the form of cheap/deep disk.”
Shattered Storage Limitations
The recent addition of an Avere virtual Edge filer (vFXT) further extends the Teradyne infrastructure by enabling access to Amazon Web Services (AWS) Elastic Cloud Compute (EC2) resources. “Traditionally, our only means for increasing compute capacity has been to purchase more physical nodes,” explains DeBella. “In fact, the thinking here has always been that, given the size of our datasets (tens of GBs/file) and the performance requirements of engineering workloads, we could never run compute anywhere but on premises. By mitigating the problems of network latency and data migration, Avere has changed our thinking, proving the viability of cloud infrastructure for Teradyne.”
The Avere vFXT makes it possible to accelerate NFS workloads using both on-premises HPC as well Amazon EC2 capacity. Teradyne can burst large simulations or regressions to the cloud—rather than investing in HPC nodes that will only be used for a few months—or run lower-priority jobs on Amazon EC2 and reserve high-performance on-prem compute for mission-critical workloads.
“Another benefit is the ability to utilize the storage cloud,” adds DeBella. “Our data growth averages 20-30 percent annually, and we effectively keep data forever for re-spins, support, and other requirements. The cloud holds great potential for more affordable and manageable archive, as well as backups using Avere cloud snapshots. Avere really is the enabler for all of these possibilities. However we choose to leverage on-prem and cloud resources, we don’t have to worry about moving or synchronizing data or modifying workflows—all of our tools and access work identically.
“For the business, the impact is significant—by delivering on-demand, never-a-bottleneck IT infrastructure, we’re directly contributing to faster product introductions, competitive innovation, higher value-add to our customers, and, perhaps most importantly, the flexibility to follow opportunities.”
Teradyne is a leading supplier of automation equipment for test and industrial applications. Teradyne Automated Test Equipment (ATE) is used to test semiconductors, wireless products, data storage, and complex electronic systems that serve consumer, communications, industrial, and government customers. Teradyne Industrial Automation products include Collaborative Robots used by global manufacturing and light-industrial customers to improve quality and increase manufacturing efficiency. Founded in 1960, the company is headquartered in North Reading, Massachusetts, and employs some 4,200 people in 70 locations throughout Asia, Europe, and North America. (www.teradyne.com)