(Tco 8) When Reading Data From a Sequential File, the Data Will Be
The era of digital business organization transformation has been in full thrust in 2019, equally emerging applications and information middle infrastructure refreshes accept been top of heed for digital stakeholders.
In this yr-finish recap, I summarize digital use cases, vertical applications, and cloud architecture options for your perusal, matched with the key Western Digital product and technological announcements we fabricated throughout the year.
The stakes are high! Infrastructure architects that fail to empathize and evaluate the latest in storage options risk price overruns, needless OPEX overhead, risk losing out on the next greatest innovation.
Data eye storage options are evolving to proceed up with data center demands of the digital era – emerging SSD form factors, SSD write endurance levels, chapters point increases, host-side FTL optimizations, HDD gains (areal density, SMR) – to address performance at scale at better infrastructure TCO.
Western Digital's 2019 new products and technology explainer videos, data sheets, news, and other resources are all here in this compendium web log for the following topics.
- AI/ML
- Analytics
- Media & Entertainment, CDNs
- In-Retentiveness Applications
- Cloud Architectures
- HPC
- Industrial IoT
- Databases
AI/ML
Artificial Intelligence and Machine Learning workloads span data ingest, information training, preparation clusters, inference, and data archive. Use instance deployments across vertical applications typically crave high reliability and depression-latency functioning for accuracy or to avert false positives.
Aligning AI/ML workloads to the advisable SSD or HDD options tin help assure successful production rollouts and go on infrastructure costs in line with upkeep expectations.
| AI/ML Pipeline Phase | Storage I/O Patterns |
|---|---|
| Ingest | High-Bandwidth Throughput |
| Data Preparation | High-Bandwidth Throughput, varying file sizes |
| Training Cluster | Low-latency, High-Bandwidth, pocket-sized random I/O |
| Deployment/Inference | Very Low-Latency, mixed read/write |
| Archive Data | High-Bandwidth Throughput, sequential write streams |
2019 Western Digital Production and Technological announcements to consider for certain pipeline stages of AI/ML:
Ultrastar® NVMe™ SSD form factor EDSFF E1.Fifty
Achieve greater storage density – up to 30.72TB NVMe SSD mixed I/O workloads, capacity optimization for server/rack density.
Ultrastar 18TB CMR HDD
Annal AI Data with 18TB CMR HDD (Conventional Magnetic Recording).
Zoned Storage Overview Diagram
Storage Orchestration for serializing AI data at a better TCO than conventional SSDs and HDDs.
Analytics
Analytics is acme of mind for digital stakeholders to provide the insight for critical business conclusion making. Whether it'due south existent-time, historical, or augmented to transactions, e.g., logistics, supply chain, or due east-commerce, we take the new products to enable scalable data mining.
Ultrastar DC SN640 NVMe
Value storage solution for mixed I/O workloads for analytics. Consider for the replacement of SATA SSDs.
Ultrastar DC SS540 SAS SSD
SAS SSD for traditional information warehouses on AFAs or server-based direct-attached storage.
Ultrastar 18TB CMR HDD
Batch Processing Analytics using 18TB CMR HDD (Conventional Magnetic Recording).
Cloud-based Media & Entertainment workflows and digital content creation pipelines enable the global content creation process for video, amusement, gaming, and other monetization of rich media.
CDNs (Content Delivery Networks) at the edge are critical for the delivery of streaming video, audio, app updates, VR/online gaming downloads. SSD depression latency for better end-experiences, storage cost optimizations are important for the deployment of read-intensive infrastructure for CDN border delivery/Points of Presence (PoPs).
Ultrastar DC SN340 NVMe
Purpose-congenital SSD for the read-intensive requirements of CDN "Write ane, read many" workloads .
Consider for the replacement for SATA SSDs.
Ultrastar DC SN640 NVMe
Mixed I/O workloads for cloud-based media workflows and content creation pipelines.
Consider for the replacement for SATA SSDs.
Retentivity Expansion solution
CDN caching – create large virtual memory pools by leveraging NVMe SSDs for scalable retentiveness for 'hot' files for CDN delivery.
Now supported on AMD® EPYC™ systems.
Ultrastar 18TB CMR HDD
Monetize long-tail content aligned with the areal density advantage of the 18TB capacity signal.
In-Retention Applications
In-Retentiveness computing delivers deliver faster performance for DRAM-based configurations.
Western Digital announced its partnership with AMD® for EPYC™- based, greater scalability and deeper insight for memory-centric applications at a improve TCO than 100% server DRAM/DIMMs using memory extension.
Redis, DevOps, databases, and key-value stores are prime number candidates for the Retentivity Extension solution.
Memory Expansion solution
Run in-memory applications in virtual retentiveness pools by leveraging NVMe SSDs for scalable memory for 'hot' files for CDN commitment.
At present supported on AMD® EPYC™ systems.
Cloud Architectures
Deject architectures continue to proliferate at hyperscalers, private cloud, multi-cloud/hybrid cloud, and for edge data centers.
Driven by TCO at scale and performance, this year's announcements enable deject architects to accost latency, scale, and an overall TCO (based on $/IOPS, $/throughput, etc.), density, power efficiency, and the right amount of endurance for specific workloads.
Zoned Storage Overview Diagram
Storage Orchestration for serializing AI data at a amend TCO than conventional SSDs and HDDs.
Ultrastar® NVMe™ SSD form cistron EDSFF E1.L
Gain the storage density reward on this new SSD form factor. Upward to thirty.72TB NVMe SSD mixed I/O workloads, capacity optimization for server/rack density.
Ultrastar DC SN640 NVMe
Containers and Hyper-converged Infrastructure for scale and read/write functioning using Ultrastar DC SN640.
Ultrastar 20TB SMR HDD
Shingled magnetic recording (SMR) HDD for the data center infrastructure TCO advantage.
Ultrastar 18TB CMR HDD
Object storage for large-scale repositories. File System Gateways: NFS and SMB.
Ultrastar DC SN340 NVMe
"Information center at the Edge" for data reads for apply cases that involve read-intensive performance and reliability.
Server Boot Drive
NVMe-based OS kicking drive.
Boot performance and reliability is disquisitional for large-node compute clusters. Consider as a replacement to SATA SSDs
HPC
On-bounds and Cloud-based HPC (High Performance Computing) workloads for scientific calculating, government labs, and academia continue to button the limits of operation of calibration to uncover discoveries in biosciences, weather/climate written report, applied science, and other disciplines.
Whether for infrastructure refresh or new cloud-based HPC architectures, this year'south announcements volition be of high interest to HPC stakeholders for OPEX and CAPEX efficiency.
Ultrastar DC SN640 NVMe
NVMe-based Bone kick drive.
Boot performance and reliability is critical for large-node compute clusters. Consider as a replacement to SATA SSDs.
Memory Expansion
In-retentivity workloads for DRAM-centric calculating needs to scale, but the costs of DRAM and the ratio of DIMM slots to CPU cores should not limit HPC architects from scaling in-memory computing.
Now supported on AMD® EPYC™ systems.
Ultrastar 18TB CMR HDD
Bulk storage, batch processing, NFS file systems, and archive using CMR HDDs at 18TB capacity points.
Zoned Storage Overview Diagram
Storage Orchestration for serialized HPC information at a better TCO than conventional SSDs and HDDs.
Industrial IoT
Industrial asset monitoring, operational monitoring, and PLM digital twins are key to IIoT platforms, applications, and OT apply cases.
OT and Information technology groups that are responsible for scalable IoT platforms are going to want to evaluate storage options and engineering to enable reliable systems at the best possible TCO.
Zoned Storage Overview Diagram
Storage Orchestration for serialized IIoT data at a improve TCO than conventional SSDs and HDDs.
Ultrastar® NVMe™ SSD course factor EDSFF E1.Fifty
Upwards to 30.72TB NVMe SSD mixed I/O workloads, capacity optimization for server/rack density.
Ultrastar DC SN640 NVMe
NVMe SSD choice for read/write performance for IIoT platforms and containers.
Memory Expansion
In-Memory: Streaming Data and PLM Digital Twins – achieve a better retentiveness configuration TCO by leveraging Ultrastar DC ME200 Retentiveness Extension Drives. This solution creates virtual retentiveness pools by using NVMe SSDs.
At present supported on AMD® EPYC™ systems
Databases
Database and Analytics stakeholders who are developing cloud-architectures to run mission-critical databases and analytics will desire to consider the latest SSD options from Western Digital to calibration latency-sensitive transactional databases.
Provisioned IOPS at deject services continue to abound to enable the proliferation of cloud-based databases and applications.
Ultrastar DC SN640 NVMe
Scale big compute clusters to run NoSQL databases.
Consider for the replacement for SATA SSDs with the performance advantages of NVMe SSDs.
Ultrastar DC SS540 SAS SSD
Refresh at a better TCO for databases running on SAS SSD for traditional data warehouses; legacy AFAs or server-based directly-fastened storage running databases on virtualization.
We look forward to working with you in 2020!
Source: https://blog.westerndigital.com/2019-data-center-year-in-review/
0 Response to "(Tco 8) When Reading Data From a Sequential File, the Data Will Be"
Post a Comment