BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
BEGIN:VEVENT
UID:834@lincs.fr
DTSTART;TZID=Europe/Paris:20240522T150000
DTEND;TZID=Europe/Paris:20240522T160000
DTSTAMP:20240601T083251Z
URL:https://www.lincs.fr/events/caching-and-pre-fetching-the-role-of-hazar
 d-rates/
SUMMARY:Caching and pre-fetching: the role of hazard rates
DESCRIPTION:Consider a local memory system receiving requests from a given
 catalog: the typical caching objective is to maximize its hit rate\, i.e.
 the number of items that are served directly from the cache. Two main
 alternatives have been studied: replacement policies (such as LRU) and
 timer-based policies (TTL).\nIn previous work\, we have explored the role
 of the hazard rate function of the inter-request times in the optimal TTL
 policy. Recently\, Towsley et al. also identified that the optimal
 non-anticipative replacement policy involves the hazard rate function.\nIn
 this talk we will discuss how these two concepts relate: we show that\, in
 large scale systems\, the optimal non-anticipative policy converges to a
 fixed threshold policy\, and that this threshold also characterizes the
 optimal TTL policy. Thus TTL caching is asymptotically optimal\, and we can
 approximate the maximal hit rate by that of the TTL policy fluid limit.
 This enables us to compute new a universal performance bound for caching
 performance. We also remark that\, in the case of increasing hazard rates\,
 caching is far from optimal and introduce the notion of optimal timer-based
 pre-fetching.\n\nhttps://aferragu.github.io
CATEGORIES:Seminars,Youtube
LOCATION:Room 4B01\, 19 place Marguerite Perey\, Palaiseau\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=19 place Marguerite Perey\,
 Palaiseau\, France;X-APPLE-RADIUS=100;X-TITLE=Room 4B01:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:DAYLIGHT
DTSTART:20240331T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR