- Community Home
- >
- HPE Networking
- >
- Networking
- >
- Bringing the data processing unit (DPU) revolution...
Networking
1819691
Members
3548
Online
109605
Solutions
Forums
Categories
Company
Local Language
๎คซ
back
Forums
Discussions
Forums
- Data Protection and Retention
- Entry Storage Systems
- Legacy
- Midrange and Enterprise Storage
- Storage Networking
- HPE Nimble Storage
Discussions
Discussions
Discussions
Discussions
Forums
Forums
Discussions
๎คซ
back
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
- BladeSystem Infrastructure and Application Solutions
- Appliance Servers
- Alpha Servers
- BackOffice Products
- Internet Products
- HPE 9000 and HPE e3000 Servers
- Networking
- Netservers
- Secure OS Software for Linux
- Server Management (Insight Manager 7)
- Windows Server 2003
- Operating System - Tru64 Unix
- ProLiant Deployment and Provisioning
- Linux-Based Community / Regional
- Microsoft System Center Integration
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Blogs
Information
Community
Resources
Community Language
Language
Forums
Blogs
Subscribe to RSS Feed
|
Blog Options
- Subscribe to RSS Feed
- Mark as New
- Mark as Read
- Bookmark
- Receive email notifications
- Printer Friendly Page
- Report Inappropriate Content
Bringing the data processing unit (DPU) revolution to your data center
Mark_Berly
โ11-14-2023
07:55 AM
The data processing unit, or DPU, is a new class of programmable processor that enables servers to more efficiently move data, freeing up valuable CPU cycles and allowing services to be statefully embedded in the data center network. As the data center continues to evolve, the combination of CPU, GPU and DPU will be the pillars that power it.
The idea of adding dedicated processors to offload network workloads isnโt newโsmart network interface cards (SmartNICs) have provided network and security capabilities for years. A key difference is that SmartNICs are limited use case or fixed-function devices. For example, you might buy a SmartNIC to offload encryption and decryption in a server. While the SmartNIC boosts encrypt/decrypt performance, thatโs all it does. If your use case changes, or you want to add a new capability, youโd need to open those servers and swap in SmartNICs with the set of functions youโre looking for. Thatโs expensive, labor-intensive, and, letโs be honest, operationally not feasible for most data center operators.
DPUs are programmable, meaning new features and capabilities can be added via simple firmware updates. Now IT organizations get accelerated performance and flexibility in a single package. We already see DPUs supporting a variety of capabilities, including IPSec, NAT, stateful firewalling, micro segmentation, and telemetry.
We have seen hyperscalers realizing massive efficiency gains as they adopt DPUs and distribute critical functions throughout their infrastructure. This distributed model eliminates the need for dedicated hardware appliances and cuts down on demands for power, cooling, and physical space.
You are probably reading this and thinking โwell, that is great for the hyperscalers, but what about me?โ That is where the smart switch comes inโproviding an easy way to benefit from the DPU revolution without an entire forklift upgrade of your current infrastructure.
DPUs and switches: a potent pairing
Why put a DPU in a switch? This placement is ideal for introducing a modern, distributed services model to enterprise data centers. In this model, software-defined and policy-driven advanced segmentation and telemetry can be deployed inline, closer to the workloads. This ultimately provides better visibility into and control over east-west traffic. All without the need for fiber taps and additional network infrastructure, reducing the you are paying for this method which only provides limited visibility.
Distributed services
The distributed services model stands in contrast to a typical data center design, which relies on complicated traffic engineering to shunt packets and flows to a cluster of purpose-built appliances that hang off the data center. This adds latency to traffic flows and complexity for network operators.
Alternatively, you can deploy software agents on servers. While this can distribute services throughout the network fabric, it adds significant management and cost, as all agents need to be deployed, managed, and updated. Agents cannot be deployed everywhere, and where they can, consume server memory and CPU. With licensing and management costs this solution is expensive at best and unobtainable by most.
Managing a distributed services data center
In addition to physical hardware, HPE Aruba Networking also provides industry-leading software-defined networking capabilities and management. This network underlay, overlay and distributed services technology makes deploying a data center network as easy as answering a short list of basic questions and letting the system provision and manage daily tasks (e.g., moves, adds and deletes). There is no need to move VLAN, VNI or policies and when the workload moves so do its network and services attributes. These are all automated with no user intervention required.
Bringing the DPU revolution to your data center
DPUs take the benefits of SmartNICsโoffloading and accelerating essential servicesโand extend them with programmable, high-performance silicon. DPUs have proven themselves in hyperscaler environments. They are now available to help transform enterprise data centers by bringing essential functions and services, which are software-defined and policy-driven, closer to workloads. Meanwhile, the network can do what it does best: getting business data where it needs to go.
For more information
Labels:
About the Author
Mark_Berly
Over the last 30 years, Mark Berly has had the opportunity to be at the epicenter of Ethernet, specifically data center networking. Holding senior technical and executive roles at the industryโs most well-known and respected networking tech companies helping architect, implement, operate, and troubleshoot some of the worldโs largest networks. Mark holds multiple patents and is a trusted technical adviser for many of today's largest corporations, executives and IT teams.
- Back to Blog
- Newer Article
- Older Article
Labels
-
AI-Powered
23 -
AI-Powered Networking
17 -
Analytics and Assurance
4 -
Aruba Unplugged
7 -
Cloud
9 -
Corporate
3 -
customer stories
4 -
Data Center
15 -
data center networks
19 -
digital workplace
2 -
Edge
4 -
Enterprise Campus
9 -
Events
5 -
Government
10 -
Healthcare
2 -
Higher Education
2 -
Hospitality
4 -
Industries
1 -
IoT
8 -
Large Public Venue
1 -
Location Services
3 -
Manufacturing
1 -
midsize business
1 -
mobility
17 -
Network as a Service (NaaS)
12 -
Partner Views
4 -
Primary Education
1 -
Retail
1 -
SASE
21 -
SD-WAN
12 -
Security
94 -
small business
1 -
Solutions
7 -
Technical
5 -
Uncategorized
1 -
Wired Wireless WAN
82 -
women in technology
2
- « Previous
- Next »
Top Kudoed Authors Last 30 Days
The opinions expressed above are the personal opinions of the authors, not of Hewlett Packard Enterprise. By using this site, you accept the Terms of Use and Rules of Participation.
Company
Learn About
News and Events
Support
© Copyright 2025 Hewlett Packard Enterprise Development LP