Back to search

Institutional Repository of Vilnius University

Write-invalidate strategijos sukeliamo „thrashing“ efekto priklausomybė nuo užklausų pasiskirstymo netolygumo /

Abstract

dc:description

This thesis analyzes distributed cache expiry and invalidation strategies, evaluating their performance under mixed workloads. A critical problem is identified: using Write‑Invalidate methods in systems with skewed request traffic (Zipf distribution) triggers a cache thrashing effect. Excessive deletion and rewriting of “hot” keys drastically degrades the cache hit ratio and increases system latency. Through discrete‑event simulation and empirical testing in a “Valkey” environment, the study demonstrates that the Write‑Invalidate strategy becomes inefficient when write operations exceed 15 % of the workload (at Zipf s = 1.20). The thesis proposes and implements an adaptive strategy that detects high‑intensity windows and dynamically falls back to a temporary TTL mode for affected keys. Results confirm that, in simulation, the adaptive method recovers the hit ratio by up to 28 pp and reduces invalidation traffic by ∼3.8×. In the empirical “Valkey” experiment under a 50 % write load, the hit ratio rises from ∼44 % (Write‑Invalidate) to ∼78 %(adaptive), while tail latencies across all strategies remain similar (∼21 ms, bounded by the simulated database write delay).

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Migauskas, Rokas,

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
lit

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Migauskas, Rokas,. Write-invalidate strategijos sukeliamo „thrashing“ efekto priklausomybė nuo užklausų pasiskirstymo netolygumo /. Institutional Repository of Vilnius University, 2026. https://repository.vu.lt/VU:ELABAETD308338082&prefLang=en_US