BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Efficient out-of-memory analytics on heterogeneous hardware throug
 h workload-aware storage interfaces.
DTSTART:20220624T140000
DTEND:20220624T160000
DTSTAMP:20260502T155957Z
UID:bf0ccf31cea25b4adf2cef51baf6834bdbdfa4e84ca72de2c1cc32b3
CATEGORIES:Conferences - Seminars
DESCRIPTION:Hamish Nicholson\nEDIC candidacy exam\nExam president: Prof. E
 douard Bugnion\nThesis advisor: Prof. Anastasia Ailamaki\nCo-examiner: Pro
 f. Anne-Marie Kermarrec\n\nAbstract\nModern servers increasingly feature h
 eterogeneous\nstorage (e.g.\, HDD\, SSD\, DRAM and emerging memory technol
 ogies).\nStorage systems\, which have historically been designed\nfor one 
 or two classes of storage devices\, must now make\ndata placement decision
 s on multiple device classes. Good data\nplacement decisions can lower sto
 rage costs and improve system\nresponse times. Most data placement decisio
 ns follow one of\ntwo approaches. The system can passively monitor request
 s to\nmake caching decisions based on frequency and recency of use\,\nor a
  human database administrator can manually place data\ninto a specific tie
 r. Passively monitoring requests discards useful\nsemantic information\, w
 hich can provide information on data reuse\nand co-accessed data\, while t
 he manual intervention does not\nscale as the decision space of data-place
 ment decisions increases.\nIn order to more effectively utilize heterogene
 ous modern storage\,\ndata systems need to provide expressive interfaces t
 hat allow\nimportant information to propagate across layers.\n\n\nBackgrou
 nd papers\n- Mosaic: a budget-conscious storage engine for relational data
 base systems\n- hStorage-DB: heterogeneity-aware data management to exploi
 t the full capability of hybrid storage systems\n- OctopusFS: A Distribute
 d File System with Tiered Storage Management\n 
LOCATION:BC 229 https://plan.epfl.ch/?room==BC%20229
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR