{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,11]],"date-time":"2026-01-11T04:30:10Z","timestamp":1768105810436,"version":"3.49.0"},"reference-count":20,"publisher":"Association for Computing Machinery (ACM)","issue":"11","license":[{"start":{"date-parts":[[2016,10,28]],"date-time":"2016-10-28T00:00:00Z","timestamp":1477612800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["Commun. ACM"],"published-print":{"date-parts":[[2016,10,28]]},"abstract":"<jats:p>Datacenter workloads demand high computational capabilities, flexibility, power efficiency, and low cost. It is challenging to improve all of these factors simultaneously. To advance datacenter capabilities beyond what commodity server designs can provide, we designed and built a composable, reconfigurable hardware fabric based on field programmable gate arrays (FPGA). Each server in the fabric contains one FPGA, and all FPGAs within a 48-server rack are interconnected over a low-latency, high-bandwidth network.<\/jats:p>\n          <jats:p>We describe a medium-scale deployment of this fabric on a bed of 1632 servers, and measure its effectiveness in accelerating the ranking component of the Bing web search engine. We describe the requirements and architecture of the system, detail the critical engineering challenges and solutions needed to make the system robust in the presence of failures, and measure the performance, power, and resilience of the system. Under high load, the large-scale reconfigurable fabric improves the ranking throughput of each server by 95% at a desirable latency distribution or reduces tail latency by 29% at a fixed throughput. In other words, the reconfigurable fabric enables the same throughput using only half the number of servers.<\/jats:p>","DOI":"10.1145\/2996868","type":"journal-article","created":{"date-parts":[[2016,10,31]],"date-time":"2016-10-31T12:29:14Z","timestamp":1477916954000},"page":"114-122","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":39,"title":["A reconfigurable fabric for accelerating large-scale datacenter services"],"prefix":"10.1145","volume":"59","author":[{"given":"Andrew","family":"Putnam","sequence":"first","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Adrian M.","family":"Caulfield","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Eric S.","family":"Chung","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Derek","family":"Chiou","sequence":"additional","affiliation":[{"name":"Microsoft and University of Texas at Austin"}]},{"given":"Kypros","family":"Constantinides","sequence":"additional","affiliation":[{"name":"Amazon Web Services, Boston, MA"}]},{"given":"John","family":"Demme","sequence":"additional","affiliation":[{"name":"Columbia University, New York, NY"}]},{"given":"Hadi","family":"Esmaeilzadeh","sequence":"additional","affiliation":[{"name":"Georgia Institute of Technology, Atlanta, GA"}]},{"given":"Jeremy","family":"Fowers","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Gopi Prashanth","family":"Gopal","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Jan","family":"Gray","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Michael","family":"Haselman","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Scott","family":"Hauck","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA and University of Washington, Seattle"}]},{"given":"Stephen","family":"Heil","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Amir","family":"Hormati","sequence":"additional","affiliation":[{"name":"Google, Inc., Mountain View, CA"}]},{"given":"Joo-Young","family":"Kim","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Sitaram","family":"Lanka","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"James","family":"Larus","sequence":"additional","affiliation":[{"name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL), Lausanne, Switzerland"}]},{"given":"Eric","family":"Peterson","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Simon","family":"Pope","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Aaron","family":"Smith","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Jason","family":"Thong","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Phillip Yi","family":"Xiao","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]},{"given":"Doug","family":"Burger","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA"}]}],"member":"320","published-online":{"date-parts":[[2016,10,28]]},"reference":[{"key":"e_1_2_1_1_1","volume-title":"Nios II Processor Reference Handbook, 13.1.0 edition","author":"Altera","year":"2014","unstructured":"Altera . Nios II Processor Reference Handbook, 13.1.0 edition , 2014 . Altera. Nios II Processor Reference Handbook, 13.1.0 edition, 2014."},{"key":"e_1_2_1_2_1","volume-title":"Stratix V Device Handbook, 14.01.10 edition","author":"Altera","year":"2014","unstructured":"Altera . Stratix V Device Handbook, 14.01.10 edition , 2014 . Altera. Stratix V Device Handbook, 14.01.10 edition, 2014."},{"key":"e_1_2_1_3_1","first-page":"426","volume":"16","author":"Baxter R.","year":"2008","unstructured":"Baxter , R. , Booth , S. , Bull , M. , Cawood , G. , Perry , J. , Parsons , M. , Simpson , A. , Trew , A. , Mccormick , A. , Smart , G. , Smart , R. , Cantle , A. , Chamberlain , R. , Genest , G. Maxwell -- A 64 FPGA Supercomputer. Eng. Lett. 16 ( 2008 ), 426 -- 433 , 2008. Baxter, R., Booth, S., Bull, M., Cawood, G., Perry, J., Parsons, M., Simpson, A., Trew, A., Mccormick, A., Smart, G., Smart, R., Cantle, A., Chamberlain, R., Genest, G. Maxwell -- A 64 FPGA Supercomputer. Eng. Lett. 16 (2008), 426--433, 2008.","journal-title":"FPGA Supercomputer. Eng. Lett."},{"key":"e_1_2_1_4_1","volume-title":"BEE4 Hardware Platform, 1.0 edition","author":"Cube","year":"2011","unstructured":"BEE Cube . BEE4 Hardware Platform, 1.0 edition , 2011 . BEECube. BEE4 Hardware Platform, 1.0 edition, 2011."},{"key":"e_1_2_1_5_1","volume-title":"HotChips 2013 (August","author":"Blott M.","year":"2013","unstructured":"Blott , M. , Vissers , K. Dataflow architectures for 10Gbps line-rate key-value stores . In HotChips 2013 (August 2013 ). Blott, M., Vissers, K. Dataflow architectures for 10Gbps line-rate key-value stores. In HotChips 2013 (August 2013)."},{"key":"e_1_2_1_6_1","volume-title":"The Convey HC-2 Computer, conv-12-030.2 edition","author":"Convey","year":"2012","unstructured":"Convey . The Convey HC-2 Computer, conv-12-030.2 edition , 2012 . Convey. The Convey HC-2 Computer, conv-12-030.2 edition, 2012."},{"key":"e_1_2_1_7_1","volume-title":"Cray XD1 Datasheet, 1.3 edition","author":"Cray","year":"2005","unstructured":"Cray . Cray XD1 Datasheet, 1.3 edition , 2005 . Cray. Cray XD1 Datasheet, 1.3 edition, 2005."},{"key":"e_1_2_1_8_1","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.1974.1050511"},{"key":"e_1_2_1_9_1","doi-asserted-by":"publisher","DOI":"10.1145\/360276.360311"},{"key":"e_1_2_1_10_1","volume-title":"Worldwide and Regional Public IT Cloud Services 2014--2018 Forecast (Oct","author":"Gens F.","year":"2014","unstructured":"Gens , F. Worldwide and Regional Public IT Cloud Services 2014--2018 Forecast (Oct . 2014 ). Gens, F. Worldwide and Regional Public IT Cloud Services 2014--2018 Forecast (Oct. 2014)."},{"key":"e_1_2_1_11_1","doi-asserted-by":"publisher","DOI":"10.1109\/MCSE.2011.11"},{"key":"e_1_2_1_12_1","doi-asserted-by":"publisher","DOI":"10.1109\/ReConFig.2011.49"},{"key":"e_1_2_1_13_1","volume-title":"IBM PureData System for Analytics N2001","author":"IBM.","year":"2013","unstructured":"IBM. IBM PureData System for Analytics N2001 , WAD12353-USEN-01 edition, 2013 . IBM. IBM PureData System for Analytics N2001, WAD12353-USEN-01 edition, 2013."},{"key":"e_1_2_1_14_1","doi-asserted-by":"publisher","DOI":"10.1109\/L-CA.2013.17"},{"key":"e_1_2_1_15_1","volume-title":"ICCAD Special Session 7C (November","author":"Martin A.","year":"2013","unstructured":"Martin , A. , Jamsek , D. , Agarawal , K. FPGA-based application acceleration: Case study with GZIP compression\/decompression streaming engine . In ICCAD Special Session 7C (November 2013 ). Martin, A., Jamsek, D., Agarawal, K. FPGA-based application acceleration: Case study with GZIP compression\/decompression streaming engine. In ICCAD Special Session 7C (November 2013)."},{"key":"e_1_2_1_16_1","volume-title":"How Microsoft Designs Its Cloud-Scale Servers","author":"Microsoft","year":"2014","unstructured":"Microsoft . How Microsoft Designs Its Cloud-Scale Servers , 2014 . Microsoft. How Microsoft Designs Its Cloud-Scale Servers, 2014."},{"key":"e_1_2_1_17_1","doi-asserted-by":"publisher","DOI":"10.1145\/2082156.2082172"},{"key":"e_1_2_1_18_1","volume-title":"Hwu, W. QP: A Heterogeneous Multi-accelerator Cluster","author":"Showerman M.","year":"2009","unstructured":"Showerman , M. , Enos , J. , Pant , A. , Kindratenko , V. , Steffen , C. , Pennington , R. , Hwu, W. QP: A Heterogeneous Multi-accelerator Cluster . 2009 . Showerman, M., Enos, J., Pant, A., Kindratenko, V., Steffen, C., Pennington, R., Hwu, W. QP: A Heterogeneous Multi-accelerator Cluster. 2009."},{"key":"e_1_2_1_19_1","volume-title":"MAPstation Systems, 70000 AH edition","author":"SRC.","year":"2014","unstructured":"SRC. MAPstation Systems, 70000 AH edition , 2014 . SRC. MAPstation Systems, 70000 AH edition, 2014."},{"key":"e_1_2_1_20_1","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM.2012.28"}],"container-title":["Communications of the ACM"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/2996868","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/2996868","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,18]],"date-time":"2025-06-18T19:05:49Z","timestamp":1750273549000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/2996868"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2016,10,28]]},"references-count":20,"journal-issue":{"issue":"11","published-print":{"date-parts":[[2016,10,28]]}},"alternative-id":["10.1145\/2996868"],"URL":"https:\/\/doi.org\/10.1145\/2996868","relation":{},"ISSN":["0001-0782","1557-7317"],"issn-type":[{"value":"0001-0782","type":"print"},{"value":"1557-7317","type":"electronic"}],"subject":[],"published":{"date-parts":[[2016,10,28]]},"assertion":[{"value":"2016-10-28","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}