Least Recently Used Access (LRUA)

• LRUA is a cache replacement strategy that evicts the least recently used object upon a cache miss, ensuring recency-based prioritization.
• It employs analytic methods such as Che's approximation and closed-form cubic formulations to estimate hit rates under stationary and non-stationary traffic.
• Static analysis techniques like antichain/ZDD-based methods enable precise cache hit/miss classification, offering significant performance improvements in real-world benchmarks.

Least Recently Used Access (LRUA) refers to cache replacement mechanisms and analysis strategies based on the Least Recently Used (LRU) policy. In LRU replacement, upon a cache miss and when the cache is full, the object that has not been requested for the longest time—the "least recently used"—is evicted. LRU is the canonical stack-based replacement policy, notable for strong recency exploitation and a rich mathematical analysis tradition. LRUA encompasses both analytic modeling of cache hit/miss rates under various demand models and static program analysis for classification of memory accesses.

1. Formal Definition and Behavioral Semantics

For a cache of capacity $C$ and a catalog of $N$ distinct objects, the LRU policy is defined as follows:

• On a request for object $o$, if $o$ is present in cache (a hit), $o$ is promoted to the most recently used (MRU) position.
• If $o$ is not present (a miss), and the cache is full, the least recently used object is evicted; $o$ is inserted as MRU.
• The cache can be modeled as an ordered list of up to $C$ items, with the leftmost as MRU and rightmost as LRU.

In static analysis, the set of possible cache states is

$$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$

where $s_1$ denotes the MRU and $N$0 the LRU, and $N$1 marks an empty line. The LRU update for state $N$2 upon access to $N$3 is defined as:

• If $N$4 for some $N$5, move $N$6 to $N$7, shift $N$8 to $N$9 rightward, keep $o$0 to $o$1 unchanged.
• If $o$2, insert $o$3 at $o$4 and shift $o$5 to $o$6 rightward, evict $o$7.

A hit occurs when $o$8 is already in cache; a miss otherwise (Maïza et al., 2018).

2. Analytic Modeling under Stationary and Non-Stationary Demand

2.1 Independent Reference Model (IRM) with Power-Law Demand

Under the IRM (i.i.d. requests), object ranks follow a Zipf-like law: $o$9, $o$0, $o$1.

Che et al.'s approximation introduces a mean-field "characteristic time" $o$2 such that the steady-state probability object $o$3 is in cache is

$o$4

with $o$5 determined by enforcing the average cache occupancy constraint

$o$6

Traditionally, finding $o$7 requires $o$8 numerical root finding (0705.1970).

2.2 Closed-Form Cubic Approximation

A constant-time closed-form for $o$9 is achieved via Taylor expansion and truncation, yielding a cubic equation in $o$0:

$o$1

with explicit coefficients in terms of $o$2, $o$3, and generalized harmonic numbers $o$4. The selected $o$5 is the smallest real root $o$6, and the per-object hit rate is computed as above.

Aggregate cache hit ratio follows immediately:

$o$7

The closed-form is $o$8 in $o$9 and $o$0, as harmonic numbers can be approximated by $o$1 (0705.1970).

2.3 Non-Stationary Traffic Patterns

For non-stationary demand (objects have finite lifetimes and time-varying popularity), a Poisson content arrival process of rate $o$2 is assumed. Each object $o$3 is published at time $o$4 with total request volume $o$5, and its request process is $o$6 for shape function $o$7 ($o$8).

Defining $o$9 as the "eviction time," Che's approximation under non-stationarity yields:

• Probability object $o$0 is in cache at $o$1:

$o$2

• Marginalizing over $o$3 and object arrivals gives integral equations: for expected cache occupancy $o$4 and hit probability $o$5,

$o$6

$o$7

where $o$8 is the moment-generating function of $o$9. Asymptotic regimes simplify the expressions in the small-cache and large-cache regimes (Ahmed et al., 2013).

3. Algorithmic Analysis and Program Classification

3.1 Classical Age-Based Abstract Interpretation

For static cache analysis, the age-based abstraction maps each block $C$0 to an interval $C$1. Transfer functions update ages according to LRU rules; at each control-flow join, intervals are merged. Hits and misses are classified as:

• Always-hit: $C$2
• Always-miss: $C$3
• Unknown: otherwise

This method scales as $C$4 but may yield 15–40% "unknown" classifications (Maïza et al., 2018).

3.2 Exact Antichain/ZDD-Based Analysis

To attain precision without model checking, the focused antichain/ZDD analysis is introduced:

• For each target block $C$5, the concrete cache state w.r.t. $C$6 is encoded as either a special "absent" symbol $C$7, or the set of all blocks younger than $C$8.
• The abstract domain consists of antichains of subsets, representing minimal (for may-hit) or maximal (for may-miss) sets of younger blocks.
• Transfer functions are given by set union and subsumption, and joins are set-wise unions with antichain minimization/maximization.

A worklist-driven fixpoint yields, for each program location, an exact determination: may-hit iff $C$9, may-miss iff $$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$0. Always-hit iff $$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$1, always-miss iff $$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$2.

On real benchmarks, this method offers 100% classification precision, with runtime and memory usage several orders of magnitude below model checking: average speedup of $$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$3 and maximum $$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$4, using typically $$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$5GB memory (Maïza et al., 2018).

Method	Mean Time	Peak Memory	Precision
Focused + Model Check	$$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$6s	$$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$7GB	exact
ZDD Fixed-Point	$$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$8s	$$S = \{ (s_1, \dots, s_C) \mid s_i \in A \cup \{\bot\}, \ s_i \ne s_j \text{ for } i \ne j \},$$9GB	exact

4. Computational Complexity and Practical Considerations

Classical hit/miss decision problems for LRU on acyclic control-flow graphs are NP-complete, both for may-hit and may-miss classification. Membership in NP follows by guessing an execution path and simulating the LRU cache, while NP-hardness is shown via reductions from SAT and Hamiltonian Path, exploiting the combinatorics of inserted and replaced blocks in LRU (Maïza et al., 2018).

Traditional numeric/simulation-based methods for cache models (e.g., exact Markov chain or iterative Che approximation) typically exhibit $s_1$0 or higher complexity. The closed-form cubic formulation (0705.1970) and the focused antichain analysis (Maïza et al., 2018) reduce complexity to $s_1$1 for occupancy/characteristic time and $s_1$2 for per-object statistics.

Accuracy for the closed-form cubic holds rigorously when $s_1$3 and Zipf exponent $s_1$4. For larger $s_1$5 or $s_1$6, proportional normalization of per-object probabilities restores precision. In non-stationary demand, error relative to simulation is within 2–5% across a broad range of traffic models (0705.1970, Ahmed et al., 2013).

5. Sensitivity to Traffic, Demand, and System Parameters

LRU cache performance is highly sensitive to several parameters:

• Content lifetime ($s_1$7): For small caches, hit probability $s_1$8. Short-lived, bursty objects increase hit rates for fixed $s_1$9.
• Volume distribution ($N$00, $N$01): The tail of the request volume distribution ($N$02 in a Pareto law, i.e., Zipfian exponent) strongly affects hit rates as $N$03 scales.
• Temporal profile ($N$04): Sharply peaked (square) temporal profiles (high $N$05) increase hit probability for small caches.
• System scale ($N$06): Relative cache size modulates approximation accuracy; normalization fixes are effective for $N$07 (Ahmed et al., 2013).

Validation against Monte Carlo simulations demonstrates that these analytic approaches robustly capture cache performance across parameter space, matching simulation results within a few percent for realistic $N$08 and widely varying workloads.

6. Limitations and Extensions

Che’s approximation and its closed-form descendants assume i.i.d. requests (IRM) or independent Cox processes for generalized traffic. These models neglect higher-order correlations (e.g., user sessions, non-Poisson arrivals, content dependency) and interactions across cache networks or hierarchies. Extensions to multi-class object types are feasible but require a-priori class-mix estimation from data.

For static analysis, all methods assume deterministic LRU behavior, single-level caches, and associative memory; they do not address hardware-specific behavior (e.g., set-associativity conflicts), nor do they handle instruction/data streams jointly.

A plausible implication is that as real-world cache and traffic models become increasingly nonstationary and high-volume, the computational efficiency and accuracy guarantees of closed-form and antichain-based analyses are necessary enablers for multi-cache, multi-object system design, yet must be augmented or hybridized to accommodate full system complexity (0705.1970, Ahmed et al., 2013, Maïza et al., 2018).