Servers & Systems: The Right Compute

Improved Big Data economics with HPE ProLiant DL325 Gen10

Get better price/performance for Big Data Analytics running on HPE Elastic Platform for Analytics (EPA) using HPE ProLiant DL325 Gen10 as a compute node powered by AMD EPYC processors.

shutterstock_506311120.jpgThe HPE Elastic Platform for Analytics (EPA) is a modular infrastructure foundation designed to deliver a scalable, multi-tenant platform by enabling independent scaling of compute and storage through infrastructure building blocks that are optimized for density and running disparate workloads. And with the availability of the recently announced HPE ProLiant DL325 Gen10 server, the HPE Elastic Platform for Big Data Analytics (EPA) adds another choice for its modular infrastructure.

This addition features a secure and versatile single processor (1P) 1U AMD EPYC™ based platform with a good balance of processor, memory and I/O for virtualization and data intensive workloads. With up to 32 cores, up to 16 DIMMs or 2 TB of memory capacity and support for up to 10 NVMe drives, this server delivers 2P performance with 1P economics.

Each AMD EPYC system on a chip (SoC) can have from 8 to 32 cores with access to industry leading amounts of I/O and up to 2TBs of memory regardless of the number of cores in use. In order to reach memory capacities of over 1.5TB RAM per processor with Intel® Xeon® Scalable processors you must use dual processors and the more expensive “M” processors that can address up to 1.5TB RAM each.

The HPE ProLiant DL325 provides a compelling platform as either a standard node in a traditional Hadoop cluster (where compute and storage are co-located and the workload requirements are more batch oriented), or as a building block node in the compute tier of an elastic cluster (where workload requirements drive unique and varying compute and storage building blocks). Edge to Core.png

Find more details about the HPE EPA architecture.

As we prepared to test the HPE ProLiant DL325 Gen10 servers in our lab with Hadoop workloads, we decided to look for the processors that provide a good value for rackmount servers for our customers. We started comparing the multitude of processor options by cost, power and performance. For initial system level performance evaluation we used SPEC CPU2017 results published at

We concluded that AMD EPYC 7551P provides great value for our workloads based on the high number of cores it provides plus access to 2TB RAM. These days an average compute node configuration starts with 256GB RAM, two SSDs for data, 25G networking and around 32 cores per node. We searched for a similar Intel configuration (same number of cores, same core frequency). The chart below shows the relative delta in system cost, power and performance between AMD and Intel for servers that provide a total of 32 cores, running at 2GHz, 256GB RAM, 2x 480GB MU SSDs for data, 2x 240GB RI SSDs for OS and 25G networking.SSDs.png

The cost deltas in the above chart take into calculation the performance delta between the two hardware configurations, so these are overall price/performance delta.

HPE Power Advisor, HPE suggested list prices and SPEC CPU2017 results were used in this comparison for AMD and Intel.

Here is the complete bill of materials

Qty       Part #               Description

1          P04852-L21       HPE DL325 Gen10 7551P AMD FIO Kit

1          P04654-B21      HPE DL325 Gen10 8SFF CTO Server

8          838083-B21       HPE 32GB 2Rx4 PC4-2666V-R Smart Kit

2          877740-B21       HPE 240GB SATA RI SFF SC DS SSD

2          875470-B21       HPE 480GB SATA MU SFF SC DS SSD

1          817749-B21       HPE Eth 10/25Gb 2P 640FLR-SFP28 Adptr

2          865408-B21       HPE 500W FS Plat Ht Plg LH Pwr Sply Kit

1          869079-B21       HPE Smart Array E208i-a SR G10 LH Ctrlr

1          P04983-B21      HPE DL325 Gen10 RPS Enablement Kit

1          866448-B21       HPE 1U Gen10 8SFF SAS Cable Kit

1          734807-B21       HPE 1U SFF Easy Install Rail Kit

1          H7J34A3           HP 3yr Foundation Care 24x7 Service

2          JL295A             HPE X240 25G SFP28 to SFP28 3m DAC

2          AF595A            HP 3.0M,Blue,CAT6 STP,Cable Data

1          G3J30AAE        RHEL Svr 2 Sckt/2 Gst 3yr 24x7 E-LTU

1          F5Z52AAE        HDP Ent 4N/50TB Rw Strg 1yr 24x7 E-LTU

As you can see from our research, HPE ProLiant DL325 Gen10 provides a great option for the HPE EPA compute tier, offering 17% cost savings over a similarly configured Intel server.

For configuring AMD servers, we recommend using the AMD tuning guides from the links at the end of this blog to ensure proper IRQ affinity as AMD processors have multiple dies on each socket.  Also please note the kernel option related to iommu (iommu=pt). Without it, the network performance suffers.

Learn more

Read the solution brief: Improve the economics of Big Data—HPE Elastic Platform for Analytics (EPA) with HPE ProLiant DL325 Gen10 Servers


Meet Servers : The Right Compute blogger, Daniel Pol, Data and Analytics ArDaniel Pol.jpgchitect, HPE.




0 Kudos
About the Author


Our team of Hewlett Packard Enterprise server experts helps you to dive deep into relevant infrastructure topics.