Assets

Competition in the era of autonomy

The automotive industry is in a highly competitive transitional period where success is not about winning, it’s about survival. Once an industry of pure hardware and adrenaline, automotive design is increasingly dependent upon, and differentiated by, software. This is especially true for Advanced Driver Assistance Systems (ADAS) development, which is introducing disruptive requirements on engineering IT Infrastructure – particularly storage, where even entry level capacities are measured in petabytes. 

SAE International, formerly known as the Society of Automotive Engineers, has defined different levels of automation. Most modern cars today have features that are at level 2-3. Today’s SAE level 3 ADAS projects have already outstripped legacy storage solutions, and with level 4 and 5 projects around the corner, the need for storage solutions optimized for high performance, high concurrency, and massive scalability has never been greater.

The value of data

As ADAS solutions evolve from simple collision-avoidance systems to fully autonomous vehicles, these systems will combine complex sensing, processing, and algorithmic technologies. This vehicle-generated data is a critical component to improving ADAS systems, feeding into integrated test and development cycles (or development tool chains) for these systems. 

In addition to vehicular data, ADAS Test and Development systems in the next three to five years will also rely on inputs from infrastructure to support the growing scale of data movement, computing, and storing required between the vehicles, across the edge, through the cloud, and within on-premises environments. Such data will support ongoing modifications to current simulation, SiL (Software in-the-Loop), and HiL (Hardware-in-the-Loop) testing to improve the reliability of services after deployment.

The following figure illustrates the typical ADAS development life cycle for automotive OEMs and suppliers leveraging the Dell EMC PowerScale scale-out NAS as the central data lake with our Data Management Systems (DMS) for ADAS:

Scaling and evolving ADAS systems will require a seamless data management process and IT infrastructure that is flexible enough to handle challenges such as:

• Future-proofing ADAS simulation and architecture, to adapt to changes in vehicle sensors and other environmental data points.
• Managing data storage to comply with regulatory and privacy requirements, while addressing performance, security, and accessibility needs. 
• Analyzing massive volumes of unstructured data sets, to support analytical modelling and querying of ADAS data. This requires costly and time-consuming data preparation steps, such as labeling data for analysis.
• Most of the sensor data is required to be used for quick data restoration for decades, so it has to be added to long term archives. 

Ideally architected for ADAS development and certification, Dell EMC PowerScale provides the scalability, performance, parallelism, and easy to use management tools to help OEMs and Tier-1 suppliers accelerate their ADAS projects. PowerScale supports simultaneous ingest from thousands of concurrent streams from around the globe, provides simultaneous access for Model-in-the-loop (MIl), Hardware in the loop (HiL), Software in the loop (SiL) testing, Deep Learning/AI, and includes archive options to meet regulatory resimulation SLAs.

Accelerate and scale your ADAS/AD development success

The data management and computational demands underpinning the ADAS/AD (autonomous driving) test and dev environment are substantial and require solutions that can scale to accommodate complex exponentially growing ADAS/AD data sets. Essential to helping ADAS/AD development teams unlock the data and create value is a high performance and high-capacity platform that can provide the following:

• A consistent, high throughput solution to ingest data from test vehicles while simultaneously delivering the test data into hundreds to thousands of concurrent streams to MiL/SiL/HiL servers, test stands, and even deep learning training. It must also scale performance near-linearly, so performance isn’t degraded as capacity is added—critical for ADAS development where sensor data ingest rates of 2 PB+ per week are becoming common.
• A high performance and predictable storage solution that will scale and manage ADAS/AD data sets and workloads as they grow centrally and regionally. Essential elements of the platform include an expandable single namespace, eliminating data silos by consolidating all globally collected ADAS/AD data; automated plug and play hardware detection and expansion that won’t disrupt ongoing projects; automated policy-based tiering to reduce file server sprawl and performance bottlenecks; and file-object orchestration and encryption that will allow data movement between high performance network-attached storage and lower-cost private and public cloud options.
• Distributed deep learning frameworks are core to unlocking data capital and foundational to ADAS and AD development. Because deep learning models are very complex and large, developers can benefit from using a deep learning framework — an interface, library, or tool that allows them to leverage deep learning easily and quickly, without requiring in-depth understanding of the underlying algorithms. These frameworks provide a clear and concise way for defining models using a collection of pre-built and pre-optimized components. Essential characteristics of well-designed deep learning frameworks, such as Tensorflow, Keras, PyTorch, and Caffe, including optimization for GPU performance, easy to understand code, extensive community support, process parallelization to reduce computation cycles, and automatically computed gradients.
• An optimized and scalable accelerator-based platform that has the capacity and ability to run AI in place as well as deep learning (training) and MiL/HiL/SiL workloads. Engineers and data scientists continuously manage massive data sets and compute-intensive workloads to run their ADAS/AD test and dev operations across departments and around the globe. A large capacity and distributed GPU-based compute and storage infrastructure gives development teams the ability to rapidly build, train, and deploy test cases and AI models, predictive analytics, simulations, and re-simulations.

Top reasons to choose Dell EMC PowerScale for ADAS/AD

Small footprint, big performance for the edge 

PowerScale F200 and F600 are new small-scale all flash nodes offering high throughput for small deployment scenarios such as on-prem data caching, required when streaming data from public cloud for Hardware-in-the-Loop (HiL) testing, or regional sensor data ingest stations. These low-cost nodes can be added to existing PowerScale/Isilon clusters - making it simple to expand with high performance.

Massive scalability for the data center

AD/ADAS datasets are growing exponentially, with requirements ranging from petabytes to exabytes of data. Dell EMC PowerScale scales as your needs grow so you can invest in infrastructure that fits your current ADAS storage requirements without overbuying performance or capacity. Scalable to tens of petabytes (PB) in a single cluster, PowerScale offers truly scalable performance and an ever-expanding single namespace that eliminates data silos by consolidating all globally collected ADAS/AD data. Tools like CloudPools take this scalability into the exabyte (EB) range, allowing data to be moved between the high-performance NAS and multiple lower-cost storage options like Dell EMC ECS object storage.

Throughput to accelerate ADAS/AD time-to-market 

PowerScale delivers the consistent, high throughput required to concurrently deliver test data into hundreds to thousands of MiL/SiL/HiL servers, test stands, and Deep Learning networks simultaneously. Multiple node types can be deployed within a single cluster, so you can deploy the storage infrastructure that meets your exact needs from high performance all-flash for AI to low-cost SATA for long term archiving. PowerScale also scales performance linearly as additional capacity is added to the cluster – critical for ADAS development where sensor data ingest rates of 2PB+ per week are becoming common.

Maintain sensor data compliance 

Most ADAS projects face strict requirements for data compliance and retention, including data privacy, physical media security, and even service level agreements (SLAs) that dictate retention of petabytes to exabytes of data for decades, with as little as a few days’ notice for full data restoration. Policy-based SmartPools and CloudPools alleviate SLA challenges by automatically tiering data to lower cost storage for long-term retention, and to higher-performance storage for revalidation. Keeping sensor test and verification data within easy reach avoids the “mad-dash” to restore large data sets from archive in the case of a defect, safety recall, or audit. The necessary data remains directly accessible within the PowerScale and ECS storage infrastructure. To protect sensitive sensor data, CloudPools fully encrypts data before offloading it to the target, which can include your own on-premises Dell EMC ECS object storage and third-party providers.

Debug designs faster 

The PowerScale OneFS operating system includes native multi-protocol support so workflows can quickly access data stored on a single cluster, eliminating the need for additional data movement. OneFS offers simultaneous access to all PowerScale nodes for a mix of AD/ADAS workloads from data ingest, MiL, SiL, and HiL testing, to Deep Learning using TensorFlow. OneFS also supports data enrichment with access to on-line databases for weather, GPS location queries, road surface types, and so on. In-place analytics for sensor data and simulation results eliminates the time and expense of moving large data sets between file and other storage solutions typically required. Multi-protocol support includes NFS, SMB, HDFS, SWIFT, HTTP, REST, and others. OneFS also supports S3, an essential protocol for cloud native applications. PowerScale easily integrates with the Dell EMC streaming data platform, offering insights on real-time and historical sensor data.

Dell complete solutions: a complete autonomous driving data lake reference architecture

Our new Dell Autonomous Drive ecosystem supports the most important steps in the ADAS/AD data process. Developed in conjunction with leading industry and technology partners, Dell Autonomous Drive combines Dell Technologies and partner infrastructure, software, and services to offer a complete end-to-end toolchain.

For more information:

• See the blog What DevOps for AD/ADAS Looks Like.
• See our white paper Dell Technologies Automotive Solutions.
• For regularly updated automotive industry events and content, please bookmark https://Dell.to/automotive-events.
• Stay tuned for future blogs!

Author: Frances Weiyi Hu  LinkedIn