Rethinking SQL for Big Data with Apache Drill

1. © 2015 MapR Technologies 1© 2015 MapR Technologies

2. © 2015 MapR Technologies 2
Topics
• Motivation
• Using Drill
• SQL + NoSQL = ???
• Security Controls
• Resources

3. © 2015 MapR Technologies 3
Empowering “as it happens”
businesses by speeding up the
data-to-action cycle

4. © 2015 MapR Technologies 4© 2015 MapR Technologies
Motivation

5. © 2015 MapR Technologies 5
SEMI-STRUCTURED
DATA
STRUCTURED DATA
1980 2000 20101990 2020
Data is Doubling Every Two Years
Unstructured data will account
for more than 80% of the data
collected by organizations
Source: Human-Computer Interaction & Knowledge Discovery in Complex Unstructured, Big Data
TotalDataStored

6. © 2015 MapR Technologies 6
1980 2000 20101990 2020
Fixed schema
DBA controls structure
Dynamic / Flexible schema
Application controls structure
NON-RELATIONAL DATASTORESRELATIONAL DATABASES
GBs-TBs TBs-PBsVolume
Database
Data Increasingly Stored in Non-Relational Datastores
Structure
Development
Structured Structured, semi-structured and unstructured
Planned (release cycle = months-years) Iterative (release cycle = days-weeks)

7. © 2015 MapR Technologies 7
How To Bring SQL to Non-Relational Data Stores?
Familiarity of SQL Agility of NoSQL
• ANSI SQL semantics
• BI (Tableau, MicroStrategy,
etc.)
• Low latency
• No schema management
– HDFS (Parquet, JSON, etc.)
– HBase
– …
• No transformation
– No silos of data
• Ease of use

8. © 2015 MapR Technologies 8
Industry's First
Schema-free SQL engine
for Big Data

9. © 2015 MapR Technologies 9
Enabling “As-It-Happens” Business with Instant Analytics
Hadoop data Data modeling Transformation
Data
movement
(optional)
Users
Hadoop data Users
Traditional
approach
Exploratory
approach
New Business questionsSource data evolution
Total time to insight: weeks to months
Total time to insight: minutes

10. © 2015 MapR Technologies 10
Evolution Towards Self-Service Data Exploration
Data Modeling and
Transformation
Data Visualization
IT-driven
IT-driven
IT-driven
Self-service
IT-driven
Self-service
Optional
Self-service
Traditional BI
w/ RDBMS
Self-Service BI
w/ RDBMS
SQL-on-Hadoop
Self-Service
Data Exploration
Zero-day analytics

11. © 2015 MapR Technologies 11
Common Use Cases
Raw Data Exploration JSON Analytics DWH offload
Hive HBaseFiles Directories
…
{JSON}, Parquet
Text Files …

12. © 2015 MapR Technologies 12
Drill Supports Schema Discovery On-The-Fly
• Fixed schema
• Leverage schema in centralized
repository (Hive Metastore)
• Fixed schema, evolving schema or
schema-less
• Leverage schema in centralized
repository or self-describing data
2Schema Discovered On-The-FlySchema Declared In Advance
SCHEMA ON
WRITE
SCHEMA
BEFORE READ
SCHEMA ON THE
FLY

13. © 2015 MapR Technologies 13
- Sub-directory
- HBase namespace
- Hive database
Drill Enables ‘SQL-on-Everything’
SELECT * FROM dfs.yelp.`business.json`
Workspace
- Pathnames
- Hive table
- HBase table
Table
- DFS (Text, Parquet, JSON)
- HBase/MapR-DB
- Hive Metastore/HCatalog
- Easy API to go beyond Hadoop
Storage plugin instance

14. © 2015 MapR Technologies 14
Drill’s Data Model is Flexible
JSON
BSON
HBase
Parquet
Avro
CSV
TSV
Dynamic
schema
Fixed schema
Complex
Flat
Flexibility
Name Gender Age
Michael M 6
Jennifer F 3
{
name: {
first: Michael,
last: Smith
},
hobbies: [ski, soccer],
district: Los Altos
}
{
name: {
first: Jennifer,
last: Gates
},
hobbies: [sing],
preschool: CCLC
}
RDBMS/SQL-on-Hadoop table
Apache Drill table
Flexibility

15. © 2015 MapR Technologies 15
Reuse Existing SQL Tools and Skills
Leverage SQL-compatible tools
(BI, query builders, etc.) via Drill’s
standard ODBC, JDBC and ANSI
SQL support
Enable business analysts, technical
analysts and data scientists to
explore and analyze large volumes
of real-time data

16. © 2015 MapR Technologies 16© 2015 MapR Technologies
Using Drill

17. © 2015 MapR Technologies 17
Business dataset {
"business_id": "4bEjOyTaDG24SY5TxsaUNQ",
"full_address": "3655 Las Vegas Blvd SnThe StripnLas Vegas, NV 89109",
"hours": {
"Monday": {"close": "23:00", "open": "07:00"},
"Tuesday": {"close": "23:00", "open": "07:00"},
"Friday": {"close": "00:00", "open": "07:00"},
"Wednesday": {"close": "23:00", "open": "07:00"},
"Thursday": {"close": "23:00", "open": "07:00"},
"Sunday": {"close": "23:00", "open": "07:00"},
"Saturday": {"close": "00:00", "open": "07:00"}
},
"open": true,
"categories": ["Breakfast & Brunch", "Steakhouses", "French", "Restaurants"],
"city": "Las Vegas",
"review_count": 4084,
"name": "Mon Ami Gabi",
"neighborhoods": ["The Strip"],
"longitude": -115.172588519464,
"state": "NV",
"stars": 4.0,
"attributes": {
"Alcohol": "full_bar”,
"Noise Level": "average",
"Has TV": false,
"Attire": "casual",
"Ambience": {
"romantic": true,
"intimate": false,
"touristy": false,
"hipster": false,
"classy": true,
"trendy": false,
"casual": false
},
"Good For": {"dessert": false, "latenight": false, "lunch": false,
"dinner": true, "breakfast": false, "brunch": false},
}
}

18. © 2015 MapR Technologies 18
Reviews dataset
{
"votes": {"funny": 0, "useful": 2, "cool": 1},
"user_id": "Xqd0DzHaiyRqVH3WRG7hzg",
"review_id": "15SdjuK7DmYqUAj6rjGowg",
"stars": 5,
"date": "2007-05-17",
"text": "dr. goldberg offers everything ...",
"type": "review",
"business_id": "vcNAWiLM4dR7D2nwwJ7nCA"
}

19. © 2015 MapR Technologies 19
Zero to Results in 2 minutes
$ tar -xvzf apache-drill-1.0.0.tar.gz
$ bin/sqlline -u jdbc:drill:zk=local
$ bin/drill-embedded
> SELECT state, city, count(*) AS businesses
FROM dfs.yelp.`business.json`
GROUP BY state, city
ORDER BY businesses DESC LIMIT 10;
+------------+------------+-------------+
| state | city | businesses |
+------------+------------+-------------+
| NV | Las Vegas | 12021 |
| AZ | Phoenix | 7499 |
| AZ | Scottsdale | 3605 |
| EDH | Edinburgh | 2804 |
| AZ | Mesa | 2041 |
| AZ | Tempe | 2025 |
| NV | Henderson | 1914 |
| AZ | Chandler | 1637 |
| WI | Madison | 1630 |
| AZ | Glendale | 1196 |
+------------+------------+-------------+
Install
Query files
and
directories
Results
Launch shell
(embedded
mode)

20. © 2015 MapR Technologies 20
Directories are implicit partitions
SELECT dir0, SUM(amount)
FROM sales
GROUP BY dir1 IN (q1, q2)
sales
├── 2014
│ ├── q1
│ ├── q2
│ ├── q3
│ └── q4
└── 2015
└── q1

21. © 2015 MapR Technologies 21
Intuitive SQL Access to Complex Data
// It’s Friday 10pm in Vegas and looking for Hummus
> SELECT name, stars, b.hours.Friday friday, categories
FROM dfs.yelp.`business.json` b
WHERE b.hours.Friday.`open` < '22:00' AND
b.hours.Friday.`close` > '22:00' AND
REPEATED_CONTAINS(categories, 'Mediterranean') AND
city = 'Las Vegas'
ORDER BY stars DESC
LIMIT 2;
+------------+------------+------------+------------+
| name | stars | friday | categories |
+------------+------------+------------+------------+
| Olives | 4.0 | {"close":"22:30","open":"11:00"} | ["Mediterranean","Restaurants"]
|
| Marrakech Moroccan Restaurant | 4.0 | {"close":"23:00","open":"17:30"} |
["Mediterranean","Middle Eastern","Moroccan","Restaurants"] |
+------------+------------+------------+------------+
Query data
with any
levels of
nesting

22. © 2015 MapR Technologies 22
ANSI SQL Compatibility
//Get top cool rated businesses
 SELECT b.name from dfs.yelp.`business.json` b
WHERE b.business_id IN
(SELECT r.business_id FROM dfs.yelp.`review.json` r
GROUP BY r.business_id HAVING SUM(r.votes.cool) > 2000 ORDER BY
SUM(r.votes.cool) DESC);
+------------+
| name |
+------------+
| Earl of Sandwich |
| XS Nightclub |
| The Cosmopolitan of Las Vegas |
| Wicked Spoon |
+------------+
Use familiar SQL
functionality
(Joins,
Aggregations,
Sorting, Sub-
queries, SQL data
types)

23. © 2015 MapR Technologies 23
Logical Views
//Create a view combining business and reviews datasets
> CREATE OR REPLACE VIEW dfs.tmp.BusinessReviews AS
SELECT b.name, b.stars, r.votes.funny,
r.votes.useful, r.votes.cool, r.`date`
FROM dfs.yelp.`business.json` b, dfs.yelp.`review.json` r
WHERE r.business_id = b.business_id;
+------------+------------+
| ok | summary |
+------------+------------+
| true | View 'BusinessReviews' created successfully in 'dfs.tmp' schema |
+------------+------------+
> SELECT COUNT(*) AS Total FROM dfs.tmp.BusinessReviews;
+------------+
| Total |
+------------+
| 1125458 |
+------------+
Lightweight file
system based
views for
granular and de-
centralized data
management

24. © 2015 MapR Technologies 24
Materialized Views AKA Tables
> ALTER SESSION SET `store.format` = 'parquet';
> CREATE TABLE dfs.yelp.BusinessReviewsTbl AS
SELECT b.name, b.stars, r.votes.funny funny,
r.votes.useful useful, r.votes.cool cool, r.`date`
FROM dfs.yelp.`business.json` b, dfs.yelp.`review.json` r
WHERE r.business_id = b.business_id;
+------------+---------------------------+
| Fragment | Number of records written |
+------------+---------------------------+
| 1_0 | 176448 |
| 1_1 | 192439 |
| 1_2 | 198625 |
| 1_3 | 200863 |
| 1_4 | 181420 |
| 1_5 | 175663 |
+------------+---------------------------+
Save analysis
results as
tables using
familiar CTAS
syntax

25. © 2015 MapR Technologies 25
Repeated Values Support
// Flatten repeated categories
> SELECT name, categories
FROM dfs.yelp.`business.json` LIMIT 3;
+------------+------------+
| name | categories |
+------------+------------+
| Eric Goldberg, MD | ["Doctors","Health & Medical"] |
| Pine Cone Restaurant | ["Restaurants"] |
| Deforest Family Restaurant | ["American (Traditional)","Restaurants"] |
+------------+------------+
> SELECT name, FLATTEN(categories) AS categories
FROM dfs.yelp.`business.json` LIMIT 5;
+------------+------------+
| name | categories |
+------------+------------+
| Eric Goldberg, MD | Doctors |
| Eric Goldberg, MD | Health & Medical |
| Pine Cone Restaurant | Restaurants |
| Deforest Family Restaurant | American (Traditional) |
| Deforest Family Restaurant | Restaurants |
+------------+------------+
Dynamically
flatten repeated
and nested data
elements as part
of SQL queries.
No ETL necessary

26. © 2015 MapR Technologies 26
Extensions to ANSI SQL to work with repeated values
// Get most common business categories
>SELECT category, count(*) AS categorycount
FROM (SELECT name, FLATTEN(categories) AS category
FROM dfs.yelp.`business.json`) c
GROUP BY category ORDER BY categorycount DESC;
+------------+------------+
| category | categorycount|
+------------+------------+
| Restaurants | 14303 |
…
| Australian | 1 |
| Boat Dealers | 1 |
| Firewood | 1 |
+------------+------------+

27. © 2015 MapR Technologies 27
Checkins dataset {
"checkin_info":{
"3-4":1,
"13-5":1,
"6-6":1,
"14-5":1,
"14-6":1,
"14-2":1,
"14-3":1,
"19-0":1,
"11-5":1,
"13-2":1,
"11-6":2,
"11-3":1,
"12-6":1,
"6-5":1,
"5-5":1,
"9-2":1,
"9-5":1,
"9-6":1,
"5-2":1,
"7-6":1,
"7-5":1,
"7-4":1,
"17-5":1,
"8-5":1,
"10-2":1,
"10-5":1,
"10-6":1
},
"type":"checkin",
"business_id":"JwUE5GmEO-sH1FuwJgKBlQ"
}

28. © 2015 MapR Technologies 28
Supports Dynamic / Unknown Columns
> SELECT KVGEN(checkin_info) checkins
FROM dfs.yelp.`checkin.json` LIMIT 1;
+------------+
| checkins |
+------------+
| [{"key":"3-4","value":1},{"key":"13-5","value":1},{"key":"6-6","value":1},{"key":"14-
5","value":1},{"key":"14-6","value":1},{"key":"14-2","value":1},{"key":"14-3","value":1},{"key":"19-
0","value":1},{"key":"11-5","value":1},{"key":"13-2","value":1},{"key":"11-6","value":2},{"key":"11-
3","value":1},{"key":"12-6","value":1},{"key":"6-5","value":1},{"key":"5-5","value":1},{"key":"9-
2","value":1},{"key":"9-5","value":1},{"key":"9-6","value":1},{"key":"5-2","value":1},{"key":"7-
6","value":1},{"key":"7-5","value":1},{"key":"7-4","value":1},{"key":"17-5","value":1},{"key":"8-
5","value":1},{"key":"10-2","value":1},{"key":"10-5","value":1},{"key":"10-6","value":1}] |
+------------+
> SELECT FLATTEN(KVGEN(checkin_info)) checkins FROM
dfs.yelp.`checkin.json` limit 6;
+------------+
| checkins |
+------------+
| {"key":"3-4","value":1} |
| {"key":"13-5","value":1} |
| {"key":"6-6","value":1} |
| {"key":"14-5","value":1} |
| {"key":"14-6","value":1} |
| {"key":"14-2","value":1} |
+------------+
Convert Map with
a wide set of
dynamic columns
into an array of
key-value pairs

29. © 2015 MapR Technologies 29
Makes it easy to work with dynamic/unknown columns
// Count total number of checkins on Sunday midnight
> SELECT SUM(checkintbl.checkins.`value`) as SundayMidnightCheckins
FROM
(SELECT FLATTEN(KVGEN(checkin_info)) checkins
FROM dfs.yelp.checkin.json`) checkintbl
WHERE checkintbl.checkins.key='23-0';
+------------------------+
| SundayMidnightCheckins |
+------------------------+
| 8575 |
+------------------------+

30. © 2015 MapR Technologies 30© 2015 MapR Technologies
SQL + NoSQL = Accessible & Linearly Scalable

31. © 2015 MapR Technologies 31
180 Tables
NOT SHOWN!

32. © 2015 MapR Technologies 32
MusicBrainz on NoSQL
• Artists, albums, tracks and labels are key objects
• Reality check:
– Add works (compositions), recordings, release, release group
• 7 tables for artist alone
• 12 for place, 7 for label, 17 for release/group, 8 for work
– (but only 4 for recording!)
– Total of 12 + 7 + 17 + 8 + 4 = 48 tables
• But wait, there’s more!
– 10 annotation tables, 10 edit tables, 19 tag tables, 5 rating tables, 86
link tables, 5 cover art tables and 3 tables for CD timing info (138 total)
– And 50 more tables that aren’t documented yet

33. © 2015 MapR Technologies 33
236 tables
to describe 7 kinds of things

34. © 2015 MapR Technologies 34
artist
id
gid
name
sort_name
begin_date
end_date
ended
type
gender
area
being_area
end_area
comment
list<ipi>
list<isni>
list<alias>
list<release_id>
list<recording_id>
artist
id
gid
name
sort_name
begin_date
end_date
ended
type
gender
area
being_area
end_area
comment
list<ipi>
list<isni>
list<alias>

35. © 2015 MapR Technologies 35
artist
id
gid
name
sort_name
begin_date
end_date
ended
type
gender
area
being_area
end_area
comment
list<ipi>
list<isni>
list<alias>
list<release_id>
list<recording_id>

36. © 2015 MapR Technologies 36
Further Reductions
• All 86 link tables become properties on artists, releases and
other entities
• All 44 tag, rating and annotation tables become list properties
• All 5 cover art tables become lists of file references
• Current score: 162 tables become 4
• You get the idea

37. © 2015 MapR Technologies 37
Is This Good?
• Expressivity
– The JSON data model is at least as expressive as the original relational
model
• Many cases easier to describe in nested data
• No cases are harder
• Efficiency
– Inlining can increase data size. Locality improves, however
– Sessionizing can substantially decrease data size
– Inlining back-references is more efficient than ordinary indexes
– Inlined columnar data allows 1000x speedup for time series
• Introspection (you decide)

38. © 2015 MapR Technologies 38
Searching for Elvis
// Find discs where Elvis was credited
> SELECT distinct album_id, name FROM
(SELECT id album_id, artist_id, name, FLATTEN(credit) FROM release) albums
join
(SELECT distinct artist_id FROM
(SELECT id artist_id, FLATTEN(alias) FROM artist
where name like 'Elvis%Presley’)
) artists
USING artist_id;

39. © 2015 MapR Technologies 39
Benefits
• Extended relational model allows massive simplification
– On a real example, we see >20x reduction in number of tables
• Simplification drives improved introspection
– This is good
• Apache Drill gives very high performance execution for extended
relational problems
• You can try this out today

40. © 2015 MapR Technologies 40© 2015 MapR Technologies
Security Controls

41. © 2015 MapR Technologies 41
Access Controls that Scale
PAM Authentication +
User Impersonation
Fine-grained row and
column level access control
with Drill Views – no
centralized security
repository required
Files HBase Hive
Drill
View 1
Drill
View 2
UUU
U
U

42. © 2015 MapR Technologies 42
Granular Security via Drill Views
Name City State Credit Card #
Dave San Jose CA 1374-7914-3865-4817
John Boulder CO 1374-9735-1794-9711
Raw File (/raw/cards.csv)
Owner
Admins
Permission
Admins
Business Analyst Data Scientist
Name City State Credit Card #
Dave San Jose CA 1374-1111-1111-1111
John Boulder CO 1374-1111-1111-1111
Data Scientist View (/views/maskedcards.csv)
Not a physical data copy
Name City State
Dave San Jose CA
John Boulder CO
Business Analyst View
Owner
Admins
Permission
Business
Analysts
Owner
Admins
Permission
Data
Scientists

43. © 2015 MapR Technologies 43
Ownership Chaining
Combine Self Service Exploration with Data Governance
Name City State Credit Card #
Dave San Jose CA 1374-7914-3865-4817
John Boulder CO 1374-9735-1794-9711
Raw File (/raw/cards.csv)
Name City State Credit Card #
Dave San Jose CA 1374-1111-1111-1111
John Boulder CO 1374-1111-1111-1111
Data Scientist (/views/V_Scientist)
Jane (Read)
John (Owner)
Name City State
Dave San Jose CA
John Boulder CO
Analyst(/views/V_Analyst)
Jack (Read)
Jane(Owner)
RAWFILEV_ScientistV_Analyst
Does Jack have access to V_Analyst? ->YES
Who is the owner of V_Analyst? ->Jane
Drill accesses V_Analyst as Jane (Impersonation hop 1)
Does Jane have access to V_Scientist ? -> YES
Who is the owner of V_Scientist? ->John
Drill accesses V_Scientist as John (Impersonation hop 2)
John(Owner)
Does John have permissions on raw file? -> YES
Who is the owner of raw file? ->John
Drill accesses source file as John (no impersonation here)
Jack queries the view V_Analyst
*Ownership chain length (# hops) is configurable
Ownership
chaining
Access
path

44. © 2015 MapR Technologies 44
Security Summary
Logical
– No physical data copies/silos
Granular
– Row level and column level security controls
De-centralized
– User impersonation respecting storage system permissions
– No separate permission repository for granular controls
– Integrated with Hadoop File System permissions and LDAP
Self-service w/ governance
– If you have access to data, you control who and how widely can access it
– Audits

45. © 2015 MapR Technologies 45© 2015 MapR Technologies
Wrapping it up

46. © 2015 MapR Technologies 46
Drill is Top-Ranked SQL-on-Hadoop
Source: Gigaom Research, 2015
Key:
• Number indicates companies relative strength across all vectors
• Size of ball indicates company’s relative strength along individual vector
“Drill isn’t just about
SQL-on-Hadoop.
It’s about SQL-on-
pretty-much-
anything,
immediately, and
without formality.”

47. © 2015 MapR Technologies 47

48. © 2015 MapR Technologies 48
Recommendations for Getting Started with Drill
New to Drill?
– Get started with Free MapR On Demand training
– Test Drive Drill on cloud with AWS
– Learn how to use Drill with Hadoop using MapR sandbox
Ready to play with your data?
– Try out Apache Drill in 10 mins guide on your desktop
– Download Drill for your cluster and start exploration
– Comprehensive tutorials and documentation available
Ask questions
– user@drill.apache.org

49. © 2015 MapR Technologies 49
Q&A
@kingmesal maprtech
jscott@mapr.com
Engage with us!
MapR
maprtech
mapr-technologies