As the manufacturing operations mature and support a broad range of services, workflows, and user base, it becomes necessary to up-scale the environments. Similarly, old applications may retire, plants and operations may get consolidated, and the systems used for production may move into a lower tier to support QA and reporting, all of which may require downscaling of those environments. This section covers the considerations for infrastructure updates and data or system migration aspects. Specific tasks may vary based on the environments, but it helps to follow the general guidance provided in this section.

The system size and resources required for a given ISV deployment vary based on the expected load, required performance, and configuration options selected.

The following are parameters to consider when making sizing decisions:

• The number of connected sensors and equipment (referred to as devices)
• The number of tags collected, the data types of the tags, and frequency of data collection
• The number of images, frequency of image captures and resolution, data streams, type of applications, and dashboards
• The number of factories or facilities
• The requirements for local data processing, AI, ML and Deep Learning, and GPU requirements for high-performance configurations
• The applications running real-time and historical analytics using telemetry and machine vision data and generating the recommendations to improve manufacturing operations

Deploying the DVD for Manufacturing Edge solution on VMware comes with the significant benefit that—if your load or performance requirements should change in the future—CPU, storage, and memory configuration changes can be made easily in the VM’s settings. Some of these changes can even be made on the fly, with no production downtime.

The VxRail system monitoring dashboard provides detailed insights into compute, network, and storage consumption at the VM level. PowerCLI, which is tightly integrated with Windows PowerShell, can be used to collect such monitoring data at various granularity and frequency. CSV files can be generated from the PowerCLI-collected dataset, and it can be used for further analytics and charting for identifying bottlenecks and tuning the environment. Application and VM-level performance data can also be collected, using tools like Windows Performance Monitor (Perfmon), IOstat, and others, to compare VxRail-reported data with application-level performance for further application-level guidance and tuning.

Evaluation of the performance of the system and future operational needs may require proactively updating the environments to meet the changing demands. These are the key points for consideration for infrastructure changes. The points below are based on the service level priorities of the target environments, and not all of the points have the same impact for all environments.

• Consider the service level changes needed with the new scale and plan the environment as per the guidance from earlier section.
• Based on the expected data growth rate, plan for some extra capacity.
• Consider the data services requirements. Some of the factors include writes per second and reads per second rates, number of connected users, size, and type of the data set to estimate the throughput requirements, desired latency based on the service levels of the environment, and data retention period.
• Images pose significant performance challenges based on the resolution, size, training models built to analyze them, how frequently those models are updated, and deep learning methodologies applied to process them. Sizing the machine vision data systems should consider these factors, along with performance and processing requirements, coupled with availability of GPU accelerators.
• Add enough data and log devices and increase their count for scaling up for performance as well as concurrency.
• Use separate VMware paravirtualized SCSI controller for each device to minimize contention for IO operations.
• Databases coalesce writes and many sequential reads and log writes are generally 256 KB or larger. When formatting the devices for the file system use, apply proper allocation unit sizes for higher performance. Allocation unit size of 1 MB and higher are available on Linux and Windows Server 2019 and higher versions. An allocation unit of 1 MB minimizes the storage requests, offers better alignment with underlying disk technology, and offers higher performance.
• Analytics tools require heavy reads and may span various database tables making them very memory and I/O intensive. Such operations also feed the results to other tiered services to augment the decisions or provide additional capabilities. Consider the type of analytics operations to run on the data and the locality of that on edge, core, and cloud with the data exchange and security requirements.
• Identify the data protection needs, high availability, and mechanisms to recover the environments in the event of failures and corruption.
• Some components of ISV applications may greatly benefit from using GPU accelerators along with existing CPUs. As GPUs are deployed in pass-through mode and are not recommended for sharing across various applications, understanding application profile and needs for GPUs can help plan the system sizing. If GPUs are not needed in early stages of the deployment, consider adding them as the performance demand grows.
• This DVD also works across functional edge, far edge, and near edge, and several ISV applications are purpose-built for specific edge deployments. Dell Edge Gateways, PowerEdge servers, and VxRail systems are among other products and services that offer a broad range of options to meet the needs of these environments. Consider ISV application-specific scaling and sizing to optimize the performance at all of these layers for the best return on investment and efficiency.
• This DVD also covers multiple ISV applications and allows cross-ISV application deployment to address a variety of use cases. Scale and sizing should also consider how multiple ISV applications interact with each other, what use cases they are addressing, and the components of these ISV applications. For example, small sizing for one ISV application may be different from small sizing for another ISV application, or when multiple ISV applications are combined.
• This DVD takes into account ISV-specific scale and sizing guidance, but the flexibility that Dell infrastructure products and services offer allows updating the scale and sizing as the application demands change.

• Use a pre-tested clone from the source environment to seed the new target environment. Once that is complete, all industrial connection and network configurations can be updated to confirm the data flow onto the target environment. These are the steps to adhere to new scale requirements:
  1. Configure the appropriate directory structure and files as per the desired scale. New data files and log files can be added as necessary to support data growth requirements at the application, databases, and support file levels.
  2. Introduce these new files to the application ecosystem using application mechanisms.
• Make sure that you have fully tested the backup or clone available from the source environment, which can be used to revert the changes, if necessary.
• Migration may also involve moving to a different hardware type, different data services, or consolidation of multiple systems into larger systems, or splitting the applications across multiple systems. Ensure that the intricacies of the applications are properly considered when planning for such migrations.