BEGIN:VCALENDAR PRODID;X-RICAL-TZSOURCE=TZINFO:-//Calagator//EN CALSCALE:GREGORIAN X-WR-CALNAME:Calagator METHOD:PUBLISH VERSION:2.0 BEGIN:VTIMEZONE TZID;X-RICAL-TZSOURCE=TZINFO:America/Los_Angeles BEGIN:DAYLIGHT DTSTART:20170312T020000 RDATE:20170312T020000 TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT CREATED;VALUE=DATE-TIME:20170518T200854Z DTEND;TZID=America/Los_Angeles;VALUE=DATE-TIME:20170531T190000 DTSTART;TZID=America/Los_Angeles;VALUE=DATE-TIME:20170531T180000 DTSTAMP;VALUE=DATE-TIME:20170518T200854Z LAST-MODIFIED;VALUE=DATE-TIME:20170518T200854Z UID:http://calagator.org/events/1250471938 DESCRIPTION:This paper will be presented by Johannes Hoff. \n\n\n\nThis p aper describes how Facebook stores 700 million data points (timestamp an d value) per minute and keeps 26 hours of it available in memory for fas t access. We will mainly look at the compression algorithms making this possible. Here is the abstract: \n\n\n"\;"\;"\; Large-scale internet services aim to remain highly available and responsive in the p resence of unexpected failures. Pro-viding this service often requires m onitoring and analyzing tens of millions of measurements per second acro ss a large number of systems\, and one particularly effective solution i s to store and query such measurements in a time series database (TSDB). A key challenge in the design of TSDBs is how to strike the right balan ce between efficiency\, scalability\, and relia-bility. In this paper we introduce Gorilla\, Facebook’s in-memory TSDB. Our insight is that user s of monitoring sys-tems do not place much emphasis on individual data p oints but rather on aggregate analysis\, and recent data points are of m uch higher value than older points to quickly detect and diagnose the ro ot cause of an ongoing problem. Gorilla op-timizes for remaining highly available for writes and reads\, even in the face of failures\, at the e xpense of possibly drop-ping small amounts of data on the write path. To improve query efficiency\, we aggressively leverage compression tech-ni ques such as delta-of-delta timestamps and XOR’d floating point values t o reduce Gorilla’s storage footprint by 10x. This allows us to store Gor illa’s data in memory\, reduc-ing query latency by 73x and improving que ry throughput by 14x when compared to a traditional database (HBase)-bac ked time series data. This performance improvement has unlocked new moni toring and debugging tools\, such as time series correlation search and more dense visualization tools. Gorilla also gracefully handles failures from a single-node to entire regions with little to no operational over head. "\;"\;"\; \n\n\nPDF: http://www.vldb.org/pvldb/vol8/p1 816-teller.pdf \n\n\nReminder: Papers We Love PDX has a code of conduct. \n\nTags: meetup:event=240093042\, meetup:group=Papers-We-Love-PDX\, com puter science\, computer programming\n\nImported from: http://calagator. org/events/1250471938 URL:https://www.meetup.com/Papers-We-Love-PDX/events/240093042/ SUMMARY:Papers We Love @ PDX - Gorilla: A Fast\, Scalable\, In-Memory Tim e Series Database LOCATION:Mozilla PDX: 1120 NW Couch St #320\, Portland OR 97209 us SEQUENCE:1 END:VEVENT END:VCALENDAR