data systems 101 prof. Stratos Idreos class 2

Transcription

1 class 2 data systems 101 prof. Stratos Idreos

2 big data V s (it is not about size only) volume velocity variety veracity actually none of that is really new new: our ability to gather and store machine generated data broad understanding that we cannot just manually get value out of data Stratos Idreos 2 /55

3 a data system stores data and provides access to data & makes knowledge generation easy data system X data analysis knowledge Stratos Idreos 3 /55

4 applications sql database kernel algorithms/operators cpu memory data data data disk Stratos Idreos 4 /55

5 declarative interface ask what you want data system the system decides how to best store and access data Stratos Idreos 5 /55

6 SQL complex legacy tuning expensive nosql simple clean just enough Gartner: DBMS Market = ~36Billion nosql/hadoop =~700Million SQL = ~rest as apps become more complex as apps need to be more scalable newsql hadoop spark flink goal: understand why, how and what is next Stratos Idreos 6 /55

7 more applications more data the need for new systems more h/w Stratos Idreos 7 /55

8 1 soon everyone will need to be a data scientist hmm, my data is too big :( how far away are we from a future where a data system sits in the critical path of everything we do? new applications/requirements Stratos Idreos 8 /55

9 2 data exploration not always sure what we are looking for (until we find it) Stratos Idreos 9 /55

10 3 daily data years [IBMbigdata] data* daily skills data years [StratosGuess] data system design, set-up, tune, use Stratos Idreos 10 /55

11 sql cpu - cpu - cpu - cpu cpu registers smaller/faster caches memory disk - disk - disk - disk + flash + non volatile memory memory hierarchy system where db runs Stratos Idreos 11 /55

12 Jim Gray, IBM, Tandem, DEC, Microsoft ACM Turing award ACM SIGMOD Edgar F. Codd Innovations award 100Kx disk 100x memory 10x on board cache 2x on chip cache registers Pluto 2 years New York 1.5 hours this building 10 min this room 1 min my head ~0 Stratos Idreos 12 /55

13 cheaper faster CPU registers on chip cache on board cache memory disk SRAM DRAM cache miss: looking for something which is not in the cache ~1ns ~10ns ~100ns speed memory wall memory miss: looking for something which is not in memory cpu mem time Stratos Idreos 13 /55 The term static differentiates SRAM from DRAM (dynamic random-access memory) which must be periodically refreshed. SRAM is faster and more expensive than DRAM; it is typically used for CPU cache while DRAM is used for a computer's main memory.

14 and touch/access only what you need design of storage/access methods/algorithms should minimize: data misses + instruction misses Stratos Idreos 14 /55

15 random access & page-based access CPU need to only read x but have to read all of page 1 registers data value x on chip cache page1 page2 page3 data move on board cache memory disk Stratos Idreos 15 /55

16 bytes query x<5 scan scan (size=120 bytes) memory level N memory level N page size: 5x8 bytes Stratos Idreos 16 /55 try to read something and use it fully and never ever read it again to avoid data transfer costs - reading data from disk has been the major effort

17 an oracle gives us the positions bytes query x<5 oracle oracle (size=120 bytes) memory level N memory level N page size: 5x8 bytes Stratos Idreos 17 /55 it does not make a difference - in fact we have to query and maintain the oracle

18 when does it make sense to have an oracle how can we minimize the cost e.g., query x< Stratos Idreos 18 /55

19 design space it all starts with how we store data every bit matters Stratos Idreos 19 /55

20 in parallel/prefetching sequential access: read one block; consume it completely; discard it; read next what is next? hardware can better predict/buffer sequential pages to be read e.g., 2MB buffers in modern DRAM amortize cost of moving disk arms Stratos Idreos 20 /55

21 random access: read one block; consume it partially; discard it; might have to read it again in future; read random next; Stratos Idreos 21 /55 disk mechanical arms - high cost to move arms - so when you move them exploit it and read everything in there memory buffers

22 C/C++ Stratos Idreos 22 /55

23 MAIN-MEMORY OPTIMIZED DATA SYSTEMS Stratos Idreos 23 /55

24 a simple example assume an array of N integers: find all positions where value>x qualifying positions select operator exists in all systems: sql, nosql, newsql data Stratos Idreos 24 /55 make it like a key-value store

25 assume an array of N integers: find all positions where value>x res=new array[data.size] what if only 1% qualifies? j=0 for (i=0; i<data.size; i++) if data[i]>x res[j++]=i data res copy res but how can we know? memory Stratos Idreos 25 /55

26 assume an array of N integers: find all positions where value>x and we haven t even started discussing about how to find the qualifying values res=new array[data.size] what if 90% qualifies? j=0 for (i=0; i<data.size; i++) if data[i]>x res[j++]=i result size= qualifying values*x bytes bit vector for res? if statements= bad, bad, bad Stratos Idreos 26 / vs

27 assume an array of N integers: find all positions where value>x thread1, core1 res=new array[data.size] j=0 for (i=0; i<data.size; i++) if data[i]>x res[j++]=i logically partition thread2, core2 thread3, core3 NUMA architectures? SIMD functionality? & what about result writing? not as simple as spinning off N threads Stratos Idreos 27 /55

28 assume an array of N integers: find all positions where value>x N>>1 queries in parallel res=new array[data.size] j=0 for (i=0; i<data.size; i++) if data[i]>x res[j++]=i q1,q2,q3 q4 q4 Stratos Idreos 28 /55

29 assume an array of N integers: find all positions where value>x res=new array[10] j=0 for (i=0; i<10; i++) if data[i]>x res[j++]=i vs res=new array[10] j=0 if data[0]>x res[j++]=i if data[1]>x res[j++]=i if data[2]>x res[j++]=i if data[3]>x res[j++]=i if data[4]>x res[j++]=i if data[5]>x res[j++]=i if data[6]>x res[j++]=i if data[7]>x res[j++]=i if data[8]>x res[j++]=i if data[9]>x res[j++]=i Stratos Idreos 29 /55 30vs50

30 assume an array of N integers: find all positions where value>x option 1: scan all data option 2: use a tree (do not consider tree generation costs) which one is best Stratos Idreos 30 /55

31 cost: data touched & computation speed cpu mem time Stratos Idreos 31 /55

32 a simple example build a key-value store similar to the ones Facebook, Google, etc use interface supported: put, get, scan, count, get range, load unique key-value pairs, r>>w but w>>0 data how to store and access Stratos Idreos 32 /55

33 design logical design physical design system design Stratos Idreos 33 /55

34 essential steps in using a database system experts/system admins clean schema load tune query user/apps Stratos Idreos 34 /55

35 professors (id,name, ) key table/relation relational model+sql courses (id,name, profid, ) column/attribute database students (id,name, ) create table for professors: create table professors (id:integer, name: char(40), telephone: char(10), ) insert into professors ( , john smith, ) give me the names of all students: select name from students where GPA>3.0 Stratos Idreos 35 /55

36 employee (id:int, name:varchar(50), office:char(5), telephone:char(10), city:varchar(30), salary:int) data schema (1, name1, office1, tel1, city1, salary1) (2, name2, office2, tel2, city2, salary2) (3, name3, office3, tel3, city3, salary3) (4, name4, office4, tel4, city4, salary4) (5, name5, office5, tel5, city5, salary5) (6, name6, office6, tel6, city6, salary6) (7, name7, office7, tel7, city7, salary7) (8, name8, office8, tel8, city8, salary8) (9, name9, office9, NULL, city9, salary9) SQL:insert into employee (1, name1, office1, tel1, city1, salary1) cardinality=9 value does not exist Stratos Idreos 36 /55

37 professors (id,name, ) enrolled (studentid, courseid, ) relational model+sql courses (id,name, profid, ) foreign key database students (id,name, ) give me all students enrolled in cs265 select student.name from students, enrolled, courses where courses.name= cs165 and enrolled.courseid=course.id and student.id=enrolled.studentid join Stratos Idreos 37 /55

38 star schema dimension table 1 (id1, ) fact table (id1,id2, ) dimension table 2 (id2, ) Stratos Idreos 38 /55

39 key-value store vs relational can we store a document collection in a relational systems? can we store a relational database in a key-value store? Stratos Idreos 39 /55

40 design logical design physical design system design Stratos Idreos 40 /55

41 essential steps in using a database system experts/system admins clean schema load tune query user/apps Stratos Idreos 41 /55

42 declarative interface ask what you want so do db systems just work? db system Stratos Idreos 42 /55

43 declarative interface ask what you want DBA indexes/views/tuning knobs but db cracking, adaptive* ideas db system Stratos Idreos 43 /55

44 design logical design physical design system design Stratos Idreos 44 /55

45 select min(a) from R where B<10 and C<80 algorithms/operators database kernel data data data parser optimizer execution storage Stratos Idreos 45 /55

46 logistics Stratos Idreos 46 /55

47 projects option 1: systems project basic key-value store functionality - work individually single machine - multi-core design basic design as in Facebook, LinkedIn, Mongo, etc. can lead to research option 2: research project self-designing data systems + shape-shifting access methods research with DASlab researchers - groups of 3 available for cs165 students or otherwise advanced students next generation adaptive Key-value stores (with Facebook) Stratos Idreos 47 /55

48 C/C++ no libraries unless we explicitly allow it we expect you build everything from scratch so you can control storage and access 100% Stratos Idreos 48 /55

49 midway check-in (10%) special class (2-3 hour?) in mid March: 1) design docs 2) at least one performance example 3) presentation/poster Stratos Idreos 49 /55

50 workload? two reading sessions and two hacking sessions per week? so maybe a minimum of 15 hours per week (if you already have decent hacking and data structure/algorithms experience) Stratos Idreos 50 /55

51 how can I prepare? 1) start browsing some basic texts Get familiar with the very basics of traditional database architectures: Architecture of a Database System. By J. Hellerstein, M. Stonebraker and J. Hamilton. Foundations and Trends in Databases, 2007 Get familiar with very basics of modern database architectures: The Design and Implementation of Modern Column-store Database Systems. By D. Abadi, P. Boncz, S. Harizopoulos, S. Idreos, S. Madden. Foundations and Trends in Databases, 2013 Get familiar with the very basics of modern large scale systems: Massively Parallel Databases and MapReduce Systems. By Shivnath Babu and Herodotos Herodotou. Foundations and Trends in Databases, ) play with basic data structures implementation in C (linked list/hash table/tree) Stratos Idreos 51 /55

52 two more lectures next week and then we go into discussion mode wed: db architectures basics fri: projects next week: paper signup + systems project is already online Stratos Idreos 52 /55

53 Action steps: 1) Read the syllabus & website carefully, 2) Register to Piazza, 3) Do P0 if you have not taken CS165 and check self-test, 4) Register for paper presentation (week 2), 5) Start submitting your paper reviews (week 3) web site: piazza: piazza.com/harvard/spring2017/cs265/home office hours: Stratos: Wed/Thur/Fri, 3-4pm, MD139 TF office hours: Mon?, Tue, 3-4pm, MD 136 textbook: nope research papers will be available from the Harvard network Stratos Idreos 53 /55

54 class 2 data systems 101 BIG DATA SYSTEMS prof. Stratos Idreos next time modern main-memory optimized data systems