How to configure Vertex AI Reduction Server worker pools?

You are training a custom language model for your company using a large dataset. You plan to use the Reduction Server strategy on Vertex AI. You need to configure the worker pools of the distributed training job. What should you do?

  1. Configure the machines of the first two worker pools to have GPUs, and to use a container image where your training code runs. Configure the third worker pool to have GPUs, and use the reductionserver container image.
  2. Configure the machines of the first two worker pools to have GPUs and to use a container image where your training code runs. Configure the third worker pool to use the reductionserver container image without accelerators, and choose a machine type that prioritizes bandwidth. Source Reference Answer
  3. Configure the machines of the first two worker pools to have TPUs and to use a container image where your training code runs. Configure the third worker pool without accelerators, and use the reductionserver container image without accelerators, and choose a machine type that prioritizes bandwidth.
  4. Configure the machines of the first two pools to have TPUs, and to use a container image where your training code runs. Configure the third pool to have TPUs, and use the reductionserver container image.

Community Votes

B
100%

100% of anonymous learners picked answer B. Votes are pick records left by other test-takers — they are not the verified answer.

Community Insight

This question tests the architecture of the Reduction Server strategy, specifically the requirement for GPUs in training pools and high-bandwidth CPUs without accelerators in the reduction pool, with the common trap being assuming TPUs are supported or that the reduction pool needs GPUs.

When configuring Vertex AI Reduction Server for distributed training, training worker pools require GPUs, while the dedicated reduction server pool requires high-bandwidth CPUs without accelerators. The community confirms that TPUs are not supported for this specific strategy, making Option B the correct configuration.

Selecting Option A or D because users often mistakenly assume the Reduction Server worker pool requires GPUs or TPUs for computation, failing to recognize that it relies on high-bandwidth CPUs for gradient aggregation.

Community Discussion (8 comments)

Pau1234 👍 1 Selected: B
Reduction server strategy: 1. Only GPUs 2. You do not use GPUs for the Reduction Server nodes. https://cloud.google.com/vertex-ai/docs/training/distributed-training
lunalongo 👍 1 Selected: B
B is the right answer because: - Reduction Server strategy is generally implemented with GPUs, not TPUs. - First 2 pools' replicas perform model training; need GPUs for faster processing - Container image should contain your custom training code. - 3rd pool contains reduction server, no GPU is needed here; prioritize network bandwidth instead!
wences 👍 1 Selected: B
The real reason for answer B is the custom model, which means it was not suited well for TPU
fitri001 👍 2 Selected: B
GPUs for Training: Configure the first two worker pools with GPUs to leverage the hardware acceleration capabilities for your custom language model training code. Reduction Server without GPUs: The third worker pool should use the reductionserver container image. This image is pre-configured for Reduction Server functionality and doesn't require GPUs. High-Bandwidth CPU: Choose a machine type with high bandwidth for the third pool since Reduction Server focuses on communication and gradient reduction.
pinimichele01 👍 1 Selected: B
https://cloud.google.com/blog/topics/developers-practitioners/optimize-training-performance-reduction-server-vertex-ai In this article, we introduce Reduction Server, a new Vertex AI feature that optimizes bandwidth and latency of multi-node distributed training on NVIDIA GPUs for synchronous data parallel algorithms.
shadz10 👍 3 Selected: B
TPUs are not supported for reductionserver so B
winston9 👍 2 Selected: B
bandwidth is important for the reduction server
pikachu007 👍 2 Selected: B
Worker Pools 1 and 2: These pools are responsible for the actual model training tasks. They require GPUs (or TPUs, if applicable to your model) to accelerate model computations. They run the container image containing your training code. Worker Pool 3: This pool is dedicated to the reduction server. It doesn't require accelerators (GPUs or TPUs) for gradient aggregation. Prioritize machines with high network bandwidth to optimize gradient exchange. Use the specific reductionserver

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

Submitted for moderation before publishing. Keep it helpful and respectful.

Expert Analysis

Why the Answer Is Correct

Option B is correct because the Reduction Server strategy on Vertex AI is designed for NVIDIA GPU-based training. The first two worker pools handle the actual model training and thus require GPUs and the custom training code container. The third worker pool runs the reductionserver container, which handles gradient aggregation; this process does not require GPUs but benefits significantly from high network bandwidth to minimize latency.

Why the Other Options Are Wrong

Options A and D are incorrect because they assign GPUs or TPUs to the third worker pool (Reduction Server), which is unnecessary and inefficient as the reduction process is network-bound, not compute-bound. Options C and D are incorrect because they suggest using TPUs; Google documentation explicitly states that Reduction Server is supported only for NVIDIA GPUs, not TPUs.

Community Comment Notes

Community members emphasize that TPUs are not supported for the Reduction Server strategy, immediately ruling out options C and D. Several comments highlight the importance of selecting a machine type that prioritizes bandwidth for the third pool, as the reduction server's primary role is to optimize network communication during distributed training.

Official Reference

Exam Strategy

Remember that the Reduction Server strategy is exclusive to NVIDIA GPUs and does not support TPUs. When configuring the pools, ensure the training nodes have accelerators while the reduction nodes are optimized for network bandwidth without accelerators.

Related Analysis

You work for a social media company. You want to create a no

A-Grade

You are creating a model training pipeline to predict sentim

A-Grade

You are implementing a batch inference ML pipeline in Google

S-Grade

Practice All PMLE Questions

Access 65 questions with complete answers and detailed explanations.

View Full PMLE Practice Test →

← Back to PMLE Study Guide