Research Proposal Part 4: Writing the Review of Releated Literature


Guidelines for Writing the Review

Specifics are for our running example.

  1. Organize by Themes or Categories: Group related work into logical categories (e.g., general Kubernetes scheduling, heterogeneous systems, ML workload optimization) to make the review structured and easy to follow.
  2. Focus on Relevance: Prioritize papers directly related to Kubernetes scheduling, heterogeneous hardware, and ML workloads. Include seminal works and recent advancements (within the last 3-5 years, i.e., 2020-2025).
  3. Critically Analyze: For each work, summarize its contributions, strengths, and limitations, explicitly linking limitations to your research gap.
  4. Use Authoritative Sources: Cite top-tier conference and journal papers (e.g., OSDI, SOSP, NSDI, EuroSys, ASPLOS) and relevant industry whitepapers or open-source documentation.
  5. Be Concise and Targeted: Aim for 1-2 paragraphs per category, focusing on key works (6-10 papers total) to avoid overwhelming the reader.
  6. Highlight Your Contribution: Conclude by summarizing how your work addresses the gaps identified in the literature.
  7. Use Clear Citations: Follow a consistent citation style (e.g., IEEE, ACM) and include DOIs or URLs for accessibility.

Practical Steps for Writing

  1. Gather Sources:
    • Search IEEE Xplore, ACM Digital Library, and arXiv for papers on Kubernetes scheduling, heterogeneous systems, and ML workloads (2020-2025).
    • Check Kubernetes GitHub issues and documentation for practical insights.
    • Monitor X for recent discussions (e.g., #Kubernetes, #MLSystems) to identify preprints or community critiques.
  2. Organize and Summarize:
    • Create a table or spreadsheet to track papers, their contributions, and limitations.
    • Group papers into the three categories above (or adjust based on findings).
    • Summarize each paper in 2-3 sentences, focusing on relevance to your gap.
  3. Draft Critically:
    • For each category, write a paragraph summarizing 2-3 key works, emphasizing their limitations.
    • Use transitions to connect categories (e.g., "While Kubernetes schedulers address general workloads, they fall short in heterogeneous settings…").
  4. Synthesize and Conclude:
    • Write a final paragraph that ties the limitations to your research gap and briefly previews your solution.
  5. Revise for Clarity:
    • Ensure technical terms (e.g., pod, accelerator) are defined or clear from context.
    • Verify citations are complete and follow the required style (e.g., [Author, Conference Year]).


Tips for Computer Systems Context

  • Emphasize Systems Metrics: Discuss how existing works measure performance (e.g., throughput, latency, resource utilization) and where they fall short for ML workloads.
  • Address Practicality: Highlight gaps in real-world applicability, especially for open-source platforms like Kubernetes.
  • Consider Scalability and Robustness: Note limitations in handling large-scale or failure-prone clusters, as these are critical in systems research.
  • Stay Current: Use recent papers (2020-2025) and check preprints on arXiv or discussions on X to ensure your review reflects the latest trends.

Part 5 will focus on the Title.

Acknowledgement

This article was made with the help of Grok (accessed 2025-07-24)