Memory Bandwidth for New PowerEdge Servers is Significantly Improved with Ice Lake Architecture
Download PDFMon, 16 Jan 2023 13:44:20 -0000
|Read Time: 0 minutes
Summary
New PowerEdge servers fueled by 3rd Generation Intel® Xeon® Scalable Processors can support sixteen DIMMs per CPU and 3200 MT/s memory speeds. This DfD will compare memory bandwidth readings observed on new PowerEdge servers with Ice Lake CPU architecture against prior-gen PowerEdge servers with Cascade Lake CPU architecture.
Ice Lake CPU Architecture
3rd Generation Intel® Xeon® Scalable Processors, known as Ice Lake processors, are the designated CPU for new Dell EMC Intel PowerEdge servers, like the R650 and R750. Compared to prior-gen 2nd Generation Intel® Xeon® Scalable Processors, Ice Lake architecture will support 33.3% more channels per CPU (an increase from six to eight) and 9.1% higher memory speeds (an increase from 2933 MT/s to 3200 MT/s.)
Performance Data
To quantify the impact of this increase in memory support, two studies were performed. The first study (see Figure 1) measured memory bandwidth determined by the number of DIMMs per CPU populated. The second study (see Figure 2) measured memory bandwidth determined by the number of CPU thread cores. Both STREAM bandwidth benchmarks have Ice Lake populated with eight 3200 MT/s DIMMs per channel, and Cascade Lake populated with six 2933 MT/s DIMMs per channel.
Figure 1 – Ice Lake and Cascade Lake bandwidth comparison by # of DIMMs per CPU
Figure 2 – Ice Lake and Cascade Lake bandwidth comparison by # of CPU core threads
Related Documents
Achieve Real-Time Data Processing with Confluent® Platform and Apache Kafka®
Tue, 17 Jan 2023 07:15:23 -0000
|Read Time: 0 minutes
Summary
Enabling mission critical application, system and connecting data to the entire organization with real-time data flow and process means that the system and software stack must be optimized. In this document Intel and Dell discuss key considerations and sample configurations for PowerEdge server deployments to ensure your Confluent Kafka architecture is robust and takes advantage of the most recent advancements in server technology.
Mission-critical applications need to analyze large amounts of data in real time, but this requires refined tools built on scalable platforms.
Originally developed at LinkedIn by the founders of Confluent, Apache Kafka® is an open-source, high-throughput message broker that fills this need. It quickly decouples, queues, processes, stores and consumes high-volume streams of event data. With Apache Kafka, enterprises can acquire data once and consume it multiple times.
Confluent continues to enhance the Kafka platform with tools like cluster management, additional security, and more connectors. Companies like Square, Bosch and The Home Depot use Confluent’s distribution of Apache Kafka to identify actionable patterns within business datai. Intel created an Apache Kafka data pipeline based on Confluent® Platform for faster security threat detection and response for its Cyber Intelligence Platform (CIP). Data flows to a Kafka message bus and then into the Splunk® platform.
Organizations that are looking for a solution to enable real-time processing of massive data streams should consider Confluent Platform and Apache Kafka running on Dell EMC™ PowerEdge™ servers with high-performing Intel compute, storage and networking technologies.
Key Considerations
- Compute. 3rd Generation Intel® Xeon® Scalable processors ingest and analyze massive quantities of data fast in the decoupling work common to Apache Kafka broker nodes.
- Storage. The Intel SSD P5500 is recommended for storage for all node types. Architected with 96-layer TLC and Intel 3D NAND Technology, it optimizes performance and capacity. The Dell™ PowerEdge RAID Controller (PERC) H755N is recommended for Brokers + Apache ZooKeeper™ nodes. It offers expandable storage capacity to improve performance.
- Networking. Network speed is one of the most important factors in Kafka performance. Intel Ethernet 800 Series network adapters enable scaling from 10 gigabit Ethernet (GbE) to 100 GbE for accelerated packet processing.
Available Configurations
Configurations for the control center node, ksqlDB + Kafka Connect + Schema Registry, and Brokers + Apache ZooKeeper are shown below.
| Control Center Node (One Node Required) | ksqlDB + Apache Kafka® Connect + Schema Registry (Minimum of Two Nodes Required) | Brokers + Apache ZooKeeper™ (Minimum of Three Nodes Required) |
Platform | Dell EMC™ PowerEdge™ R650 or R750 chassis supporting NVM Express® (NVMe®) drives | ||
CPUii | 2 x Intel® Xeon® Silver 4316 processor (20 cores at 2.3 GHz) | 2 x Intel® Xeon® Gold 6330 processor (28 cores at 2.0 GHz) | 2 x Intel® Xeon® Silver 4316 (20 cores at 2.3 GHz)—small throughput clusters 2 x Intel® Xeon® Gold 6338 (32 cores at 2.0 GHz)—medium throughput clusters 2 x Intel® Xeon® Platinum 8368 (38 cores at 2.4 GHz)—high throughput clusters with full encryption enabled |
DRAMiii | 64 GB (4 x 16 GB) | 128 GB (8 x 16 GB) | 128 GB (8 x 16 GB) or more |
Boot device | Dell EMC™ Boot Optimized Server Storage (BOSS)-S2 with 2 x 480 GB Intel® SSD D3-S4510 M.2 Serial ATA (SATA) | ||
Storage controlleriv | None | Dell™ PERC H755N Front NVMe | |
Storagev | 2 x 3.84 TB Intel® SSD P5500 | 4 x 3.84 TB Intel® SSD P5500 | |
Network interface controller (NIC) | Intel® Ethernet Network Adapter E810-XXVDA2 for OCP3 (dual-port 25 Gb) | Intel® E810-XXVDA2 for OCP3 (dual-port 25 Gb) or Intel® E810- CQDA2 PCIe® (dual-port 100 Gb) for high-throughput clusters |
Learn More
Contact your dedicated Dell or Intel account team. 1-877-289+-3355
Download the solution briefs and white papers below:
- Enabling Real-Time Processing of Massive Data Streams
- IT@Intel: Building a Modern, Scalable Cyber Intelligence Platform with Apache Kafka
- IT@Intel: Transforming Intel’s Security Posture with Innovations in Data Intelligence
The information in this publication is provided as is. Dell Inc. makes no representations or warranties of any kind with respect to the information in this publication, and specifically disclaims implied warranties of merchantability or fitness for a particular purpose.
Use, copying, and distribution of any software described in this publication requires an applicable software license.
Copyright © 2021 Dell Inc. or its subsidiaries. All Rights Reserved. Dell, EMC, PowerEdge and other trademarks are trademarks of Dell Inc. or its subsidiaries. Other trademarks may be the property of their respective owners.
Dell Inc. believes the information in this document is accurate as of its publication date. The information is subject to change without notice.
i Confluent. “Set Your Data in Motion.” 2021. www.confluent.io/.
ii Small throughput: less than 10 gigabits per second (Gbps), medium throughput: less than 25 Gbps, high throughput: more than 25 Gbps
iii Brokers and Apache ZooKeeper™: More memory might be required to accommodate traffic bursts.
iv Brokers and Apache ZooKeeper™: An NVMe® RAID controller is optional for small- and medium-throughput clusters.
v Brokers and Apache ZooKeeper™: Add more drives or add higher capacity drives as needed for higher throughput, extended data-retention periods or desired (optional) RAID configurations.
Protect Sensitive Data with a Confidential Computing Solution
Tue, 17 Jan 2023 06:08:27 -0000
|Read Time: 0 minutes
Summary
There are multiple considerations to take into account when deploying artificial intelligence and machine learning environments. This paper serves as a discussion and suggestion as to the possible hardware configurations to achieve a server infrastructure deployment that is secure and can grow with your increased need.
Enterprises in most industries are applying artificial intelligence/machine learning (AI/ML) to data. However, data privacy and sensitivity issues are preventing the use of AI/ML in health and financial sectors. This data cannot be shared, and it is limited to on- premises usage. Although this data must be protected from exposure to unauthorized parties, it is a valuable resource that could lead to groundbreaking discoveries and innovation in areas such as pandemic response, anti-money-laundering tactics and human trafficking
Confidential computing offers a way to expand the utility of such data while also keeping sensitive details sequestered and private. Dell EMC™ PowerEdge™ servers, built on 3rd Generation Intel® Xeon® Scalable processors, are available for the first time with confidential computing. A key feature is Intel Software Guard Extensions (Intel SGX), which provides an extra layer of hardware-based encryption in memory that helps protect data while it is being accessed. With Intel SGX, organizations can access and use multiple expansive datasets for AI applications, leading to greater insights. Intel SGX also helps ensure the integrity of the AI app to protect against intrusion, and it provides increased integrity to the platform whilst helping satisfy sovereignty requirements.
Key Considerations
- Decrease attack surfaces. Dell EMC PowerEdge servers based on Intel technologies expose fewer attack surfaces to hackers by making use of Intel SGX. The entire solution is built on an architecture that provides multiple layers of hardware- and software- based security. Security settings can be tailored to sequester only the most sensitive, privacy-protected data on Intel SGX to optimize performance.
- Accelerate inferencing. Intel performance optimizations that are built into Dell EMC PowerEdge servers can help speed AI inferencing. These optimizations include Intel and open-source software and hardware technologies, such as Intel Deep Learning Boost (Intel DL Boost) with Vector Neural Network Instructions (VNNI) for AI acceleration, Intel oneAPI Deep Neural Network Library (Intel oneDNN), the OpenVINO™ toolkit and optimized versions of TensorFlow™ and PyTorch®.
- Secure ecosystem. By creating a confidential computing environment with Intel SGX, ecosystem partners can improve the security of data collaboration between organizations that might have a need to keep data confidential and to protect intellectual property, even if a lack of trust exists between the parties. Secure ecosystem partners enable code and system integrity with Intel SGX for a growing list of AI solutions.
Available Configurations
| Base Configuration | Plus Configuration (More Memory for Larger Workloads) |
Platform | Dell EMC™ PowerEdge™ R650 servers, supporting 10 NVM Express® (NVMe®) drives (direct connection with no Dell™ PowerEdge RAID Controller [PERC]), 1 RU | |
CPU | 2 x Intel® Xeon® Gold 6348 processor (28 cores at 2.6 GHz) with 64 GB/CPU Intel® SGX enclave capacity | 2 x Intel® Xeon® Platinum 8368 processor (38 cores at 2.4 GHz) with 512 GB/CPU Intel® SGX enclave capacity |
DRAM | 256 GB (16 x 16 GB DDR4-3200) | 512 GB (16 x 32 GB DDR4-3200) |
Boot device | Dell EMC™ Boot Optimized Server Storage (BOSS)-S2 with 2 x 480 GB Intel® SSD S4510 M.2 Serial ATA (SATA) (RAID1) | |
Storage adapter | Dell PERC H755N front NVMe RAID adapteri | |
Cache storage (optional) | 1 x 400 GB Intel® Optane™ SSD P5800X (PCIe Gen4) or 1 x 375 GB Intel® Optane SSD DC P4800X (PCIe Gen3)ii | |
Capacity storage | 1 x (up to 9 x) 3.84 TB Intel® SSD P5500 (PCIe Gen4, read intensive) | |
Network interface controller (NIC) | Intel® Ethernet Network Adapter E810-XXV for OCP3 (dual-port 25 Gb) |
Learn More
Written with Intel
Learn more about secure AI inferencing:
- “Intel and Consilient Join Forces to Fight Financial Fraud with AI”
- “Intel SGX Enables Magnit to Create a Trusted Computing Environment”
- “Spain’s largest hospitals connect for federated learning”
Contact your Dell or Intel account team. 1-877-289+-3355
- Read the Principled Technologies report: “Reap better SQL Server OLTP performance with next-generation Dell EMC PowerEdge MX servers.”
- View the Principled Technologies vSAN infographic.
i An NVM Express® (NVMe®) RAID adapter is optional, but it is recommended for configurations with a large number of capacity drives.
ii Cache storage is optional. Intel® Optane™ SSD P5800X drives are recommended when available, but the previous-generation Intel® Optane SSD DC P4800X can be used otherwise.
The information in this publication is provided as is. Dell Inc. makes no representations or warranties of any kind with respect to the information in this publication, and specifically disclaims implied warranties of merchantability or fitness for a particular purpose.