{"675748":{"#nid":"675748","#data":{"type":"event","title":"PhD Defense by Matthew Whitlock","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle:\u0026nbsp;\u003C\/strong\u003EDesigning and Automating Asynchronous, Localized, Multi-Level Fault-Tolerance at the Application Level\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EDate:\u0026nbsp;\u003C\/strong\u003EThursday, August 15, 2024\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ETime:\u003C\/strong\u003E\u0026nbsp;3pm - 5pm EST\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ELocation:\u0026nbsp;\u003C\/strong\u003EKlaus 3402\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ERemote\u003C\/strong\u003E: via \u003Ca href=\u0022https:\/\/teams.microsoft.com\/l\/meetup-join\/19%3ameeting_MDI0MGIyMWQtNzhkNC00MzFmLTg3NDItM2JjMWQ2MmI5YWZh%40thread.v2\/0?context=%7b%22Tid%22%3a%22482198bb-ae7b-4b25-8b7a-6d7f32faa083%22%2c%22Oid%22%3a%221ca2054b-a945-4131-9ad7-f3552a880051%22%7d\u0022\u003ETeams\u003C\/a\u003E\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EMatthew Whitlock\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EPh.D. Candidate in Computer Science\u003C\/p\u003E\u003Cp\u003ESchool of Computer Science\u003C\/p\u003E\u003Cp\u003ECollege of Computing\u003C\/p\u003E\u003Cp\u003EGeorgia Institute of Technology\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ECommittee:\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EDr. Vivek Sarkar (Advisor) - School of Computer Science, Georgia Institute\u0026nbsp;of Technology\u003C\/p\u003E\u003Cp\u003EDr. Keita Teranishi - Advanced Computing Systems Research Section, Oak Ridge National Laboratories\u003C\/p\u003E\u003Cp\u003EDr. Ada Gavrilovska - School of Computer Science, Georgia Institute of\u0026nbsp;Technology\u003C\/p\u003E\u003Cp\u003EDr. Umakishore Ramachandran - School of Computer Science, Georgia Institute\u0026nbsp;of Technology\u003C\/p\u003E\u003Cp\u003EDr. Tom Conte - School of Computer Science, Georgia Institute of Technology\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAbstract\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EMoore\u0027s law is dead or dying, but Rock\u0027s law of doubling costs for semiconductor fabrication is still going strong. It is becoming more expensive to meet ever-growing compute demands, and the general public is expressing growing concerns about the environmental impact of extreme-scale computing. Consequently, researchers in fields like machine learning and embedded computing are exploring reduced-reliability computing.\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003ESupercomputing facilities, however, are struggling to maintain high-reliability hardware to support the inefficient and unscalable global checkpoint\/restart (C\/R) mechanisms that most scientific computing applications continue to rely on. The performance cost of C\/R is rising faster than the performance of leading supercomputers. Applications\u0027 fault-tolerance must scale against higher parallelism and reduced hardware reliability for HPC to continue scaling while reigning in its environmental footprint. To avoid the exponential growth of C\/R overheads, applications must localize the cost of hardware faults. Further, fault tolerance must be flexible to application-specific refinements while managing application developers\u0027 reticence to implement complex resilience code.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EWe describe a layer-based resilience taxonomy that exposes the imperative configurability mechanisms to make fault-tolerance tools that can flexibly combine to utilize general application- and platform- tailored fault recovery. We prove this by extending contemporary resilience tools to enable flexible, easy-to-implement online recovery into applications with a multi-layered approach. Next, we define the key requirements of localized recovery by creating a general analytical model for local recovery. We prove that recovery can be localized using modern User-Level Fault Tolerance (ULFM) MPI features despite ULFM\u0027s collective recovery constraints. Finally, we prove that asynchrony via task-based parallelism can mitigate the non-local costs of recovery for applications which cannot strictly meet the requirements for localized recovery.\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EThese works build the path for HPC to maintain environmental accountability, meet growing compute demands, and benefit from novel upcoming hardware trends.\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EDesigning and Automating Asynchronous, Localized, Multi-Level Fault-Tolerance at the Application Level\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Designing and Automating Asynchronous, Localized, Multi-Level Fault-Tolerance at the Application Level"}],"uid":"27707","created_gmt":"2024-08-02 16:51:42","changed_gmt":"2024-08-02 16:52:22","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2024-08-15T15:00:00-04:00","event_time_end":"2024-08-15T17:00:00-04:00","event_time_end_last":"2024-08-15T17:00:00-04:00","gmt_time_start":"2024-08-15 19:00:00","gmt_time_end":"2024-08-15 21:00:00","gmt_time_end_last":"2024-08-15 21:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Klaus 3402","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"100811","name":"Phd Defense"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}