unified-buffer-management · research · also UBM buffer management

Unified Buffer Management (UBM)

Unified Buffer Management classifies file-block accesses and allocates cache space among sequential, looping and other-reference partitions.

Detection, replacement and allocation are separate modules. Per-file consecutive blocks identify sequences; repeated sequences establish loops. Sequential blocks use MRU, loops use period-based replacement with MRU ties, and the paper uses LRU for other references. [1]

For a loop of l blocks with period p, the model gives Hit(n)=min(l,n)/p. Its marginal gain is 1/p through n=l and zero beyond saturation. Time here counts all block references, not seconds. Comparing partitions requires the same observation clock. [1]

In a constructed four-block loop with period 20, two buffers yield 2/20=0.1 hits per global reference, although 2/4=0.5 of that loop's accesses hit in the model. Doubling the period to 40 halves the former rate without changing the latter fraction. A fifth buffer adds no hits after all four loop blocks fit. [2]

A newly detected sequence initially lacks loop evidence. Under cache pressure, its first repeated access can miss before the detector recognizes looping; cold-cache spare space can conceal this cost. Detection therefore does not promise a first-reuse hit. [1]

Relations

related
Program-counter-based classification (PCC), Locality of reference

Limitations

  • Estimated reference-count gains do not price unequal I/O latency, dirty writeback or prefetch costs. [1]

Citations

  1. Kim et al. A Low-Overhead High-Performance Unified Buffer Management Scheme that Exploits Sequential and Looping References. OSDI 2000, sections 3-5. (accessed 2026-11-21)
  2. Hesela. Loop-allocation arithmetic and explicit global-reference denominators. (accessed 2026-11-21)

Further reading

  1. Hesela. Bounded loop expectation calculation and deterministic example generator. (accessed 2026-11-21)
  2. Hesela. Loop expectation boundary tests and exhaustive event-count oracle. (accessed 2026-11-21)
  3. Hesela. Reproduction instructions and model scope. (accessed 2026-11-21)
Sources & caveats
  • OSDI 2000 combines trace-driven simulation and a FreeBSD prototype. Its workload-specific improvements are not forecasts for modern systems. [1]
  • Synthetic arithmetic, not a native UBM implementation, measured trace or reproduction of the paper's reported performance. [2]

JSON: /data/terms/unified-buffer-management.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Unified Buffer Management (UBM). Hesela. 2026-11-21. https://hesela.dev/terms/unified-buffer-management/