ΠΑΝΕΠΙΣΤΗΜΙΟ ΙΩΑΝΝΙΝΩΝ - ΤΜΗΜΑ ΜΗΧΑΝΙΚΩΝ Η/Υ & ΠΛΗΡΟΦΟΡΙΚΗΣ SITE MAP | ΠΑΝΕΠΙΣΤΗΜΙΟ ΙΩΑΝΝΙΝΩΝ - ΤΜΗΜΑ ΜΗΧΑΝΙΚΩΝ Η/Υ & ΠΛΗΡΟΦΟΡΙΚΗΣ ΕΛΛΗΝΙΚΑ | ΠΑΝΕΠΙΣΤΗΜΙΟ ΙΩΑΝΝΙΝΩΝ - ΤΜΗΜΑ ΜΗΧΑΝΙΚΩΝ Η/Υ & ΠΛΗΡΟΦΟΡΙΚΗΣ ENGLISH
Πέμπτη, 08/01/2015   -   Ομιλίες
Σεμινάριο Τμήματος με τίτλο:"Efficiency and Consistency in Multistream Storage", Στέργιος Αναστασιάδης

Στο πλαίσιο της διοργάνωσης των εβδομαδιαίων σεμιναρίων του τμήματος θα πραγματοποιηθεί την Παρασκευή 16/01/2015 και ώρα 12:00 στην αίθουσα Σεμιναρίων του Τμήματος Μηχανικών Η/Υ και Πληροφορικής, ομιλία με τίτλο "Efficiency and Consistency in Multistream Storage". Ομιλητής θα είναι ο κ. Στέργιος Αναστασιάδης, Επίκουρος Καθηγητής του Τμήματος Μηχανικών Η/Υ & Πληροφορικής, Πανεπιστήμιου Ιωαννίνων.

ΠΕΡΙΛΗΨΗ

Synchronous small writes play critical role in system availability, because they are used to safely log recent state modifications for fast recovery from crashes. Demanding systems typically dedicate separate devices to logging for adequate performance during normal operation and redundancy during state reconstruction. However, storage stacks enforce page-sized granularity in data transfers from memory to disk and consume excessive storage bandwidth to handle small writes. The problem becomes worse in the typical case of multiple concurrent streams, which effectively generate random I/O traffic. In a journaled filesystem, we introduce Wasteless Journaling to coalesce synchronous concurrent small writes of data into full page-sized journal blocks. Additionally, we propose Selective Journaling to automatically activate wasteless journaling on data writes of size below a fixed threshold. We implemented a functional prototype of our design over a widely-used filesystem. We experimentally compare our methods with existing ones using microbenchmarks and application-level workloads on standalone servers and a multi-tier networked system. Coalescing small data updates to the journal preserves filesystem consistency, reduces consumed bandwidth up to several factors, and lowers write latency up to orders of magnitude.

©2010 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING - UNIVERSITY OF IOANNINA
created by data|SCIENCE

Valid XHTML 1.0 Strict Valid CSS!