Solutions
Solutions Home
Systems
Solutions
Performance Overview
Reference Architecture Overview
Tuning Guides for Cloud Native Processing
Tutorials
Workload Briefs Overview
Where to Try
EN
Test Notes for Big Data Workloads
Scalability Tests were performed on a 9 node cluster of bare metal nodes.
- CPU - Single socket per server with Ampere Altra N1 processor with 80 cores @3.0GHz
- Memory - Samsung DDR4 512GB DRAM 3200 MHz
- Storage - 4 Micron 7300 NVME Drives each with 3.5 TB storage space
- Network Cards - Mellanox CX-6, 2 x 100Gb/s with a bonded configuration
Terasort Hadoop and Spark tests were done on Ampere Altra and x86 VM's with the following configurations.
- Intel - Skylake 2000 GHz
- AMD - Rome 2249 MHz
- AMD - Milan 2449 MHz
- Ampere - Altra processors 3000 MHz
- vCPU's - 16 ARM Cores or 16 x86 threads.
- Memory - 64 GB
- Storage - 1TB storage with 1000 MB/sec read/write performance
Note: TPC-DS tests were done on Intel Ice Lake processors(Graph 2).
Few yarn and spark configuration details used on VM's listed below:
| dfs.block.size | 512M |
| yarn.scheduler.maximum-allocation-mb | 40g |
| yarn.scheduler.maximum-allocation-vcores | 15 |
| yarn.nodemanager.resource.cpu-vcores | 16 |
| yarn.nodemanager.resource.memory-mb | 64g |
| mapreduce.map.memory.mb | 1024 |
| mapreduce.reduce.memory.mb | 3072 |
| mapred.reduce.parallel.copies | 16 |
| mapreduce.reduce.shuffle.parallelcopies | 14 |
| mapreduce.map.java.opts | Xmx2048m |
| spark.master | yarn |
| spark.executor.cores | 5 |
| spark.executor.memory | 18g |
| spark.executor.instances | 3 |
Created At : May 15th 2022, 7:14:43 pm
Last Updated At : March 20th 2023, 6:51:53 pm
© 2023 Ampere Computing LLC. All rights reserved. Ampere, Altra and the A and Ampere logos are registered trademarks or trademarks of Ampere Computing.
