The Cache Inference Problem and its Application to Content and Request Routing

Abstract

Abstract-In many networked applications, independent caching agents cooperate by servicing each other's miss streams, without revealing the operational details of the caching mechanisms they employ. Inference of such details could be instrumental for many other processes. For example, it could be used for optimized forwarding (or routing) of one's own miss stream (or content) to available proxy caches, or for making cache-aware resource management decisions. In this paper, we introduce the Cache Inference Problem (CIP) as that of inferring the characteristics of a caching agent, given the miss stream of that agent. While CIP is insolvable in its most general form, there are special cases of practical importance in which it is, including when the request stream follows an Independent Reference Model (IRM) with generalized power-law (GPL) demand distribution. To that end, we design two basic "litmus" tests that are able to detect LFU and LRU replacement policies, the effective size of the cache and of the object universe, and the skewness of the GPL demand for objects. Using extensive experiments under synthetic as well as real traces, we show that our methods infer such characteristics accurately and quite efficiently, and that they remain robust even when the IRM/GPL assumptions do not hold, and even when the underlying replacement policies are not "pure" LFU or LRU. We exemplify the value of our inference framework by considering example applications.

Temporal locality in today's content caching

“…We show that the IRM leads to considerable errors, rendering its use in this context misguided at best. We believe this to be an important result, because, due its tractability, the IRM is still the standard assumption made in many works [5,6,12,13]. As our second contribution, we propose a novel replacement to the IRM called the Shot-Noise Model (SNM).…”

“…The most common approach to evaluating the performance of caching systems is to assume that content requests are generated under the IRM, with a request distribution following a generalised Zipf law [6,10,11,12,18]. The IRM considers a fixed population of N contents, such that the sequence of content requests arriving at the cache is characterised by the following set of fundamental properties: the probability of a request for a given content n, for 1 ≤ n ≤ N , is constant (i.e.…”

Integrated prefetching and caching for adaptive video streaming over HTTP

“…In addition, we assume that suitable mechanisms for request routing (e.g., [17] Figure 1: The architecture of iPac. The green lines denote the video traffic, and the rest denote the request traffic.…”

An optimal replication strategy for data grid systems

“…In the future work, we will deal with the heterogeneous case that sub grid can take different size. Without loss of generality, we assume that the nodes belong to sub grid g i are Just like some earlier works on replication strategy in P2P network [12][13][14]16], we assume different files have the same unit size to avoid adding 0/1-knapsacktype complexities to a problem that is already combinatorial [17]. In fact, our subsequent analysis can be extended to this case by treating the jth file as a group of c j files with the same query rate, where c j is the size of file f j .…”