Accelerating Redis Performance Using VMware vSphere 8 and NVIDIA BlueField DPUs

A shift to modern distributed workloads, along with higher networking speeds, has increased the overhead of infrastructure services. There are fewer CPU cycles…

Tim Lustig
9 min readadvanced
--
View Original

Overview

The article discusses how deploying NVIDIA BlueField Data Processing Units (DPUs) with VMware vSphere 8 can significantly enhance the performance of Redis, a popular NoSQL database. It highlights the benefits of offloading networking tasks to DPUs, resulting in improved transaction speeds, reduced CPU utilization, and increased energy efficiency.

What You'll Learn

1

How to benchmark Redis performance using NVIDIA BlueField DPUs

2

Why offloading networking tasks to DPUs improves application performance

3

How to achieve energy efficiency in data centers with DPUs

Prerequisites & Requirements

  • Understanding of distributed workloads and NoSQL databases
  • Familiarity with VMware vSphere and NVIDIA BlueField DPUs(optional)

Key Questions Answered

How does using NVIDIA BlueField DPUs enhance Redis performance?
Using NVIDIA BlueField DPUs allows for offloading networking tasks from the CPU, which significantly boosts Redis performance by increasing transactions per second (TPS) and reducing application latency. The benchmarks showed nearly 20 million TPS with full DPU acceleration compared to 12.75 million TPS with a standard NIC.
What are the energy efficiency benefits of using DPUs with Redis?
The use of DPUs can lead to a reduction in power consumption by 6%-40% per million TPS compared to a regular NIC. This efficiency allows data centers to run more workloads on fewer servers, ultimately saving on operational costs.
What were the benchmark results for Redis on vSphere with DPU acceleration?
The benchmark results indicated that with full DPU acceleration, Redis achieved 20 million TPS using 80 instances, while the default DPU offload mode achieved 17.74 million TPS. In contrast, the standard NIC peaked at 12.75 million TPS with only 30 instances.
How does DPU offloading affect CPU utilization?
DPU offloading frees up CPU cores by transferring network processing tasks to the DPU, allowing more cores to be dedicated to business applications. This results in a more efficient system with reduced CPU utilization for networking tasks.

Key Statistics & Figures

Transactions per second (TPS) with full DPU acceleration
20 million TPS
Achieved with 80 Redis instances using full DPU acceleration.
Transactions per second (TPS) with default DPU offload
17.74 million TPS
Achieved with 80 Redis instances using default DPU offload.
Transactions per second (TPS) with standard NIC
12.75 million TPS
Achieved with only 30 Redis instances.
Power consumption reduction
6%-40% fewer watts per million TPS
Compared to a regular NIC when using full DPU acceleration.
Reduction in CPU cores needed
14-18% fewer CPU cores
To support the same Redis workload when using DPU acceleration.

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

Hardware
Nvidia Bluefield Dpu
Used to offload networking tasks and enhance performance in vSphere environments.
Software
Vmware Vsphere 8
Platform for running distributed workloads and managing virtual machines.
Database
Redis
NoSQL database used for testing performance improvements with DPU acceleration.

Key Actionable Insights

1
Consider implementing NVIDIA BlueField DPUs in your VMware vSphere environment to enhance Redis performance.
By offloading networking tasks to DPUs, you can significantly increase transaction throughput and reduce application latency, which is crucial for high-demand applications.
2
Evaluate the energy efficiency of your data center by comparing power consumption metrics with and without DPU acceleration.
Understanding the power savings achieved through DPU usage can help justify the investment in this technology and improve overall operational costs.
3
Utilize the NVIDIA LaunchPad to experiment with DPU capabilities before full deployment.
This platform allows you to test and benchmark Redis performance with DPU acceleration, providing insights into potential improvements without the need for immediate hardware investment.

Common Pitfalls

1
Overlooking the need for proper benchmarking before deploying DPUs.
Without thorough testing, organizations may not fully understand the performance benefits or potential issues that could arise from integrating DPUs into their existing infrastructure.
2
Assuming all workloads will benefit equally from DPU acceleration.
Different applications may have varying degrees of performance improvement when offloading tasks to DPUs, so it's essential to evaluate specific use cases.

Related Concepts

Distributed Systems And Their Performance Challenges
Nosql Database Optimization Techniques
Energy Efficiency In Data Centers
Networking Offloading Strategies