Phoenix SNAPSHOT Performance

Phoenix Performance Result :: Tue Jul 16 14:54:47 PDT 2013

AGGREGATION_TEST_TABLE
Various server side grouped/un-grouped aggregation functions. Data is pre-split in 9 regions over a 4 region server cluster.

ROWS 4000000 [4M]

DDL CREATE TABLE $TABLE (HOST CHAR(2) NOT NULL,DOMAIN VARCHAR NOT NULL,FEATURE VARCHAR NOT NULL,DATE DATE NOT NULL,USAGE.CORE BIGINT,USAGE.DB BIGINT,STATS.ACTIVE_VISITOR INTEGER CONSTRAINT PK PRIMARY KEY (HOST, DOMAIN, FEATURE, DATE)) SPLIT ON ('CSGoogle','CSSalesforce','EUApple','EUGoogle','EUSalesforce','NAApple','NAGoogle','NASalesforce')

DATA GENERATOR
STRING :: Values: NA|CS|EU
STRING :: Values: Salesforce.com|Apple.com|Google.com|Yelp.com
STRING :: Values: Login|Report|Dashboard|Sales|UI
DATE :: Values: SEQUENTIAL
INTEGER :: Max. Value: 100 Values: RANDOM
INTEGER :: Max. Value: 2000 Values: RANDOM
INTEGER :: Max. Value: 10000 Values: RANDOM

QUERIES

select count(*) from AGGREGATION_TEST_TABLE

2.36s PHOENIX-1.1

2.29s PHOENIX-1.2.0

select count(1) from AGGREGATION_TEST_TABLE group by HOST

2.42s PHOENIX-1.1

2.37s PHOENIX-1.2.0

select MIN(CORE) from AGGREGATION_TEST_TABLE group by DOMAIN

2.73s PHOENIX-1.1

2.69s PHOENIX-1.2.0

select core, count(1) from AGGREGATION_TEST_TABLE group by CORE

2.79s PHOENIX-1.1

2.58s PHOENIX-1.2.0

select core, host, sum(ACTIVE_VISITOR) from AGGREGATION_TEST_TABLE group by core, host

4.41s PHOENIX-1.1

4.17s PHOENIX-1.2.0

select AVG(DB), MIN(CORE), MAX(ACTIVE_VISITOR) from AGGREGATION_TEST_TABLE where host='CS' and DOMAIN<>'Google.com'

5.26s PHOENIX-1.1

5.64s PHOENIX-1.2.0

select round(date, 'day',50), sum(core)+2 from AGGREGATION_TEST_TABLE where (CORE<10 or CORE>90) group by round(date, 'day',50), host

1.84s PHOENIX-1.1

1.96s PHOENIX-1.2.0

select trunc(date, 'hour', 1000), count(DB), count(ACTIVE_VISITOR) from AGGREGATION_TEST_TABLE where feature='Login' and (core<50 or db<20 or active_visitor > 200) group by trunc(date, 'hour', 1000)

3.64s 200)) SERVER AGGREGATE INTO DISTINCT ROWS BY [ROUND(DATE)] CLIENT MERGE SORT'>PHOENIX-1.1

2.56s 200)) SERVER AGGREGATE INTO DISTINCT ROWS BY [ROUND(DATE)] CLIENT MERGE SORT'>PHOENIX-1.2.0

select count(active_visitor) count from AGGREGATION_TEST_TABLE

3.79s PHOENIX-1.1

3.54s PHOENIX-1.2.0

select count(distinct active_visitor) count from AGGREGATION_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select percentile_cont(0.9) within group (order by active_visitor asc) from AGGREGATION_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select percentile_cont(0.1) within group (order by active_visitor desc) from AGGREGATION_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select count(core), count(db), count(active_visitor) from AGGREGATION_TEST_TABLE

5.41s PHOENIX-1.1

4.99s PHOENIX-1.2.0

select count(distinct core), count(distinct db), count(distinct active_visitor) from AGGREGATION_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select percentile_cont(0.5) within group (order by core asc), percentile_cont(0.5) within group (order by db asc), percentile_cont(0.5) within group (order by active_visitor asc) from AGGREGATION_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select core, count(db) count from AGGREGATION_TEST_TABLE group by CORE

4.02s PHOENIX-1.1

3.89s PHOENIX-1.2.0

select core, count(distinct db) count from AGGREGATION_TEST_TABLE group by CORE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select core, percentile_cont(0.9999) within group (order by db asc) from AGGREGATION_TEST_TABLE group by CORE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select count(core) from AGGREGATION_TEST_TABLE where core < 10 and active_visitor < 1000

2.91s PHOENIX-1.1

1.5s PHOENIX-1.2.0

select count(core) from AGGREGATION_TEST_TABLE where core < 50 and active_visitor < 5000

3.47s PHOENIX-1.1

1.58s PHOENIX-1.2.0

select count(core) from AGGREGATION_TEST_TABLE where core < 100 and active_visitor < 10000

4.81s PHOENIX-1.1

1.64s PHOENIX-1.2.0

select count(core) from AGGREGATION_TEST_TABLE where core > 90

1.52s PHOENIX-1.1

1.58s PHOENIX-1.2.0

select count(core) from AGGREGATION_TEST_TABLE where core > 50

2.15s PHOENIX-1.1

2.18s PHOENIX-1.2.0

select count(core) from AGGREGATION_TEST_TABLE where core > 0

2.86s PHOENIX-1.1

2.82s PHOENIX-1.2.0

LOAD DATA [4000000 ROWS]

118.81s PHOENIX-1.1

145.82s PHOENIX-1.2.0

IN_LIKE_TABLE
Phoenix 1.2+ uses skip scan that gives significant performance improvement for IN/LIKE queries.

ROWS 4000000 [4M]

DDL CREATE TABLE IF NOT EXISTS $TABLE (K1 CHAR(1) NOT NULL, K2 VARCHAR NOT NULL, CF.A INTEGER, CF.B VARCHAR, CF.C VARCHAR CONSTRAINT PK PRIMARY KEY (K1,K2)) SPLIT ON ('B','C','D','E','F','G','H','I')

DATA GENERATOR
STRING :: Values: A|B|C|D|E|F|G|H|I
STRING :: Length: 10 Values: SEQUENTIAL
INTEGER :: Max. Value: 1000000 Values: RANDOM
STRING :: Length: 50 Values: RANDOM
STRING :: Length: 50 Values: RANDOM

QUERIES

select count(1) from IN_LIKE_TABLE where K1 in ('Z', 'D', 'F', 'X')

1.33s PHOENIX-1.1

1.25s PHOENIX-1.2.0

select * from IN_LIKE_TABLE where K1 in ('A','B') and K2 in ('xxx1000000','xxx2000000','xxx2000001')

1.65s PHOENIX-1.1

0.01s PHOENIX-1.2.0

select * from IN_LIKE_TABLE where K1 in ('A', 'C', 'E', 'F', 'H') and K2 like 'xxxxxx1%'

7.92s PHOENIX-1.1

0.02s PHOENIX-1.2.0

select count(*) from IN_LIKE_TABLE where K1 in ('X','A','B','C','D','E','F','G','H','I','Z') and K2 like '%1%'

1.74s PHOENIX-1.1

1.88s PHOENIX-1.2.0

select count(*) from IN_LIKE_TABLE where (K1 in ('A','E','F','G','H','I','Z') or K1 like 'E%') and (K2 like '%xx1%' or K2 like '%xxxxxx2%')

1.88s PHOENIX-1.1

1.95s PHOENIX-1.2.0

LOAD DATA [4000000 ROWS]

94.76s PHOENIX-1.1

123.02s PHOENIX-1.2.0

INDEXED_TEST_TABLE
Indexing test. Phoenix 2.0+ feature. Data is pre-split in 9 regions over a 4 region server cluster.

ROWS 2000000 [2M]

DDL CREATE TABLE $TABLE (HOST CHAR(2) NOT NULL,DOMAIN VARCHAR NOT NULL,FEATURE VARCHAR NOT NULL,DATE DATE NOT NULL,USAGE.CORE BIGINT,USAGE.DB BIGINT,STATS.ACTIVE_VISITOR INTEGER CONSTRAINT PK PRIMARY KEY (HOST, DOMAIN, FEATURE, DATE)) IMMUTABLE_ROWS=true SPLIT ON ('CSGoogle','CSSalesforce','EUApple','EUGoogle','EUSalesforce','NAApple','NAGoogle','NASalesforce');CREATE INDEX INDEX_CORE ON $TABLE (CORE);CREATE INDEX INDEX_DB ON $TABLE (DB);CREATE INDEX INDEX_ACTIVE_VISITOR ON $TABLE (ACTIVE_VISITOR);CREATE INDEX INDEX_ALL ON $TABLE (CORE,DB,ACTIVE_VISITOR);

DATA GENERATOR
STRING :: Values: NA|CS|EU
STRING :: Values: Salesforce.com|Apple.com|Google.com|Yelp.com
STRING :: Values: Login|Report|Dashboard|Sales|UI
DATE :: Values: SEQUENTIAL
INTEGER :: Max. Value: 100 Values: RANDOM
INTEGER :: Max. Value: 2000 Values: RANDOM
INTEGER :: Max. Value: 10000 Values: RANDOM

QUERIES

select count(*) from INDEXED_TEST_TABLE

1.16s PHOENIX-1.1

1.18s PHOENIX-1.2.0

select count(1) from INDEXED_TEST_TABLE group by HOST

1.2s PHOENIX-1.1

1.24s PHOENIX-1.2.0

select MIN(CORE) from INDEXED_TEST_TABLE group by DOMAIN

1.38s PHOENIX-1.1

1.42s PHOENIX-1.2.0

select core, count(1) from INDEXED_TEST_TABLE group by CORE

1.38s PHOENIX-1.1

1.35s PHOENIX-1.2.0

select core, host, sum(ACTIVE_VISITOR) from INDEXED_TEST_TABLE group by core, host

2.28s PHOENIX-1.1

2.24s PHOENIX-1.2.0

select AVG(DB), MIN(CORE), MAX(ACTIVE_VISITOR) from INDEXED_TEST_TABLE where host='CS' and DOMAIN<>'Google.com'

2.68s PHOENIX-1.1

2.83s PHOENIX-1.2.0

select round(date, 'day',50), sum(core)+2 from INDEXED_TEST_TABLE where (CORE<10 or CORE>90) group by round(date, 'day',50), host

0.98s PHOENIX-1.1

0.99s PHOENIX-1.2.0

select trunc(date, 'hour', 1000), count(DB), count(ACTIVE_VISITOR) from INDEXED_TEST_TABLE where feature='Login' and (core<50 or db<20 or active_visitor > 200) group by trunc(date, 'hour', 1000)

1.82s 200)) SERVER AGGREGATE INTO DISTINCT ROWS BY [ROUND(DATE)] CLIENT MERGE SORT'>PHOENIX-1.1

1.28s 200)) SERVER AGGREGATE INTO DISTINCT ROWS BY [ROUND(DATE)] CLIENT MERGE SORT'>PHOENIX-1.2.0

select count(active_visitor) count from INDEXED_TEST_TABLE

1.9s PHOENIX-1.1

1.85s PHOENIX-1.2.0

select count(distinct active_visitor) count from INDEXED_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select percentile_cont(0.9) within group (order by active_visitor asc) from INDEXED_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select percentile_cont(0.1) within group (order by active_visitor desc) from INDEXED_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select count(core), count(db), count(active_visitor) from INDEXED_TEST_TABLE

2.67s PHOENIX-1.1

2.58s PHOENIX-1.2.0

select count(distinct core), count(distinct db), count(distinct active_visitor) from INDEXED_TEST_TABLE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select core, count(db) count from INDEXED_TEST_TABLE group by CORE

2s PHOENIX-1.1

1.91s PHOENIX-1.2.0

select core, count(distinct db) count from INDEXED_TEST_TABLE group by CORE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select core, percentile_cont(0.9999) within group (order by db asc) from INDEXED_TEST_TABLE group by CORE

N/A for PHOENIX-1.1

N/A for PHOENIX-1.2.0

select count(core) from INDEXED_TEST_TABLE where core < 10 and active_visitor < 1000

1.48s PHOENIX-1.1

0.76s PHOENIX-1.2.0

select count(core) from INDEXED_TEST_TABLE where core < 50 and active_visitor < 5000

1.74s PHOENIX-1.1

0.78s PHOENIX-1.2.0

select count(core) from INDEXED_TEST_TABLE where core < 100 and active_visitor < 10000

2.38s PHOENIX-1.1

0.8s PHOENIX-1.2.0

select count(core) from INDEXED_TEST_TABLE where core > 90

0.77s PHOENIX-1.1

0.82s PHOENIX-1.2.0

select count(core) from INDEXED_TEST_TABLE where core > 50

1.08s PHOENIX-1.1

1.12s PHOENIX-1.2.0

select count(core) from INDEXED_TEST_TABLE where core > 0

1.44s PHOENIX-1.1

1.45s PHOENIX-1.2.0

LOAD DATA [2000000 ROWS]

61.29s PHOENIX-1.1

80.68s PHOENIX-1.2.0

CF_TABLE_14CF
Data is split in 14 column families. Phoenix 1.2+ uses Essential Column Family filter feature which fetches only required columns when filtering rows.

ROWS 4000000 [4M]

DDL CREATE TABLE IF NOT EXISTS $TABLE (K1 CHAR(1) NOT NULL, K2 VARCHAR NOT NULL, CF1.A INTEGER, CF2.B INTEGER, CF3.C INTEGER, CF4.D INTEGER, CF5.E INTEGER, CF6.F INTEGER, CF7.G INTEGER, CF8.H INTEGER, CF9.I INTEGER, CF10.J INTEGER, CF11.K INTEGER, CF12.L INTEGER, CF13.M INTEGER, CF14.N INTEGER CONSTRAINT PK PRIMARY KEY (K1,K2)) SPLIT ON ('B','C','D')

DATA GENERATOR
STRING :: Values: A|B|C|D
STRING :: Length: 10 Values: SEQUENTIAL
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM

QUERIES

select count(1) from CF_TABLE_14CF where K1 in ('A', 'B', 'C', 'D')

3.5s PHOENIX-1.1

3.26s PHOENIX-1.2.0

select count(1) from CF_TABLE_14CF where K1 in ('A', 'C')

3.16s PHOENIX-1.1

3.18s PHOENIX-1.2.0

select count(*) from CF_TABLE_14CF where N < 1000

5.23s PHOENIX-1.1

1.67s PHOENIX-1.2.0

select a,b,c,d,e,f,g,h,i,j,k,l,m,n from CF_TABLE_14CF where (B>1000 and B<2000) and (M>1000 and M<2000)

131.79s PHOENIX-1.1

2.33s PHOENIX-1.2.0

select sum(a),sum(e),sum(j),sum(n) from CF_TABLE_14CF group by K1

11.05s PHOENIX-1.1

10.37s PHOENIX-1.2.0

LOAD DATA [4000000 ROWS]

193.3s PHOENIX-1.1

255.7s PHOENIX-1.2.0

SALTED_TABLE
This is Phoenix 1.2+ feature. Data is split in 4 regions equal to number of region servers in test cluster by adding salted byte to each key.

ROWS 4000000 [4M]

DDL CREATE TABLE $TABLE (K1 CHAR(1) NOT NULL, K2 VARCHAR NOT NULL, CFA.A VARCHAR, CFA.B CHAR(100), CFA.C INTEGER, CFB.D INTEGER CONSTRAINT PK PRIMARY KEY (K1,K2)) SALT_BUCKETS=4

DATA GENERATOR
STRING :: Values: A|B|C|D
STRING :: Length: 10 Values: SEQUENTIAL
STRING :: Length: 100 Values: RANDOM
STRING :: Length: 100 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM

QUERIES

select count(1) from SALTED_TABLE where K1 in ('A', 'B', 'C', 'D')

4.86s PHOENIX-1.1

1.92s PHOENIX-1.2.0

select count(*) from SALTED_TABLE where D < 1000

6.69s PHOENIX-1.1

0.92s PHOENIX-1.2.0

select avg(c),avg(d) from SALTED_TABLE group by K1

9.82s PHOENIX-1.1

4.06s PHOENIX-1.2.0

LOAD DATA [4000000 ROWS]

167.77s PHOENIX-1.1

162.62s PHOENIX-1.2.0

CF_TABLE_1CF
Same data as in 14 column family table but this one only has a single column family. Performance should be worse in most cases.

ROWS 4000000 [4M]

DDL CREATE TABLE IF NOT EXISTS $TABLE (K1 CHAR(1) NOT NULL, K2 VARCHAR NOT NULL, CF.A INTEGER, CF.B INTEGER, CF.C INTEGER, CF.D INTEGER, CF.E INTEGER, CF.F INTEGER, CF.G INTEGER, CF.H INTEGER, CF.I INTEGER, CF.J INTEGER, CF.K INTEGER, CF.L INTEGER, CF.M INTEGER, CF.N INTEGER CONSTRAINT PK PRIMARY KEY (K1,K2)) SPLIT ON ('B','C','D')

DATA GENERATOR
STRING :: Values: A|B|C|D
STRING :: Length: 10 Values: SEQUENTIAL
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM
INTEGER :: Max. Value: 1000000 Values: RANDOM

QUERIES

select count(1) from CF_TABLE_1CF where K1 in ('A', 'B', 'C', 'D')

4.42s PHOENIX-1.1

4.15s PHOENIX-1.2.0

select count(1) from CF_TABLE_1CF where K1 in ('A', 'C')

4.4s PHOENIX-1.1

4.12s PHOENIX-1.2.0

select count(*) from CF_TABLE_1CF where N < 1000

5.64s PHOENIX-1.1

5.55s PHOENIX-1.2.0

select a,b,c,d,e,f,g,h,i,j,k,l,m,n from CF_TABLE_1CF where B>1000 and B<2000 and M>1000 and M<2000

24.69s PHOENIX-1.1

6.57s PHOENIX-1.2.0

select sum(a),sum(e),sum(j),sum(n) from CF_TABLE_1CF group by K1

12.67s PHOENIX-1.1

12.08s PHOENIX-1.2.0

LOAD DATA [4000000 ROWS]

164.79s PHOENIX-1.1

219.71s PHOENIX-1.2.0

Note: Hover over version label to see Phoenix Query Explain Plan.