fusion-io-mysql-multi-instances-report.pdf

7/27/2019 fusion-io-mysql-multi-instances-report.pdf

http://slidepdf.com/reader/full/fusion-io-mysql-multi-instances-reportpdf 1/14

Fusion-io MySQL multi-instances report

Percona Inc.,Vadim Tkachenko

Oct-2011



Contents

0 Introduction 3

1 Testing methodology 4

2 Results 62.1 2400W, 120GB memory . . . . . . . . . . . . . . . . . 62.2 2400W, 64GB memory . . . . . . . . . . . . . . . . . 72.3 1200W, 120GB memory . . . . . . . . . . . . . . . . . 82.4 1200W, 64GB memory . . . . . . . . . . . . . . . . . 9

3 Conclusions 10

A Systems Specifications 11

A.1 Server configuration . . . . . . . . . . . . . . . . . . . 11A.2 Client configuration . . . . . . . . . . . . . . . . . . . . 11

B MySQL configuration 12

2



0 Introduction

The goal of this research is get most performance from Fusion-io ioDrivecard, used to run MySQL database server.

It is known that MySQL due internal limitations is not able to utilizeall CPU and IO resources available on modern hardware. Idea is to runmultiple instances of MySQL to gain better performance.

The research is sponsored by Fusion-io, Inc

3



1 Testing methodology

For tests we usedtpcc-mysql package, which generates TPCC-like workloadon MySQL systems.

1. Server hardware: Dell PowerEdge R815, details in Appendix A.1

2. Storage: Fusion-io ioDrive Duo 640GB MLC. Fusion-io driver version:2.3.1 build 123; Firmware v5.0.7, rev 101971

3. Software: Percona Server 5.5.15, configuration in Appendix B

4. Client hardware: IBM x3650, details in Appendix A.2

tpcc-mysql tests were run for following combinations:

• 2400W, big buffer pool

– 1 MySQL instance, 2400 warehouses (220GB of data), 120GB buffer pool

– 2 MySQL instances, 1200 warehouses (110GB of data) each, 60GB buffer pool each


• 2400W, small buffer pool

– 1 MySQL instance, 2400 warehouses (220GB of data), 64GB bufferpool



• 1200W, big buffer pool

– 1 MySQL instance, 1200 warehouses (110GB of data), 120GB buffer pool


• 1200W, small buffer pool

– 1 MySQL instance, 1200 warehouses (110GB of data), 64GB bufferpool

4



– 2 MySQL instances, 600 warehouses (55GB of data) each, 32GB

buffer pool each

The purpose of different combination of data and memory sizes, was tocheck how data/memory ratio affects results.

We used 48 user sessions and we performed 2700 sec long run, gatheringdata for New Order Transaction each 10 seconds.

That is, for each set of user sessions, we take 270 throughput measure-ments.

Based on this, we constructed the following metric, which we use asfinal result:

• Median Throughput for last 900 sec, to avoid warm-up influence onresults.

5



2 Results

2.1 2400W, 120GB memory

Table 1: Results for 2400W, 120GB memory

Instances 1 2 4Median Throughput 11184 15668 18738

Ratio - 1.4x 1.68x

FusionIO, 48 threads, 2400W − 120GB BP

Instances

T h r o u g h p u t , N O T

/ 1 0 s e c

0

5000

10000

15000

11184

15668

18738

1 2 4

6



2.2 2400W, 64GB memory


Instances 1 2 4Median Throughput 4810 8786 11952

Ratio - 1.83x 2.48x


Instances

T h r o u g h p u t , N O T / 1 0 s e c

0

2000

4000

6000

8000

10000

12000

4810

8786

11952

1 2 4

7



2.3 1200W, 120GB memory


Instances 1 2Median Throughput 12069 17590

Ratio - 1.46x


Instances


0

5000

10000

15000

12069

17590

1 2

8



2.4 1200W, 64GB memory


Instances 1 2Median Throughput 11362 16070

Ratio - 1.41x


Instances


0

5000

10000

15000

11362

16070

1 2

9



3 Conclusions

In conclusion we can highlight:

• Running multiple instances shows good improvement in throughput.1.4x-1.8x for 2 instances and 1.6-2.4x for 4 instances.

• If you have sharding environment which allows you separate databaseinto multiple instances you may try 2-4 instances setup to get betteroverall throughput from your MySQL setup

10



A Systems Specifications

A.1 Server configuration

# P e rco na T o o l ki t S ys te m Summary Re po rt ######################Date | 2011−10−06 2 1 : 0 1 : 4 5 UTC ( l o c a l TZ : PDT −0700)

System | Dell Inc . ; PowerEdge R815 ;Se r vi c e Tag | FD9M6Q1

P l at f o rm | Lin uxR e l ea s e | Red Hat E n t e r p r i s e L inux S e r ve r r e l e a s e 6 . 1 ( S a nt i

K e r n e l | 2 . 6 . 3 2 −10 0. 34 .1 . e l 6u e k . x86_64A r c h i t e c t u r e | CPU = 64− b i t , OS = 64− b i t

Threading | NPTL 2. 12SELinux | Disabled

V i r tu a l iz e d | No v i r t u a l i z a t i o n d et ec te d# P ro ce s s o r ##################################################

P ro ce ss or s | p hy si ca l = 4 , c or es = 4 8 , v i r t u a l = 4 8 ,| h y p e r t h r e a d i n g = n o

S p e e d s | 4 8 x 1 9 0 0 . 0 0 7Models | 48xAMD Opteron (tm) Pro ce sso r 6168C ache s | 48x 512 KB

# Memory ######### ######## ######### ######## ######### ######### #

T o t a l | 1 5 7 . 7 1 GA.2 Client configuration

# P e rco na T o o l ki t S ys te m Summary Re po rt ######################Date | 2011−10−06 2 1 : 0 6 : 5 3 UTC ( l o c a l TZ : EDT −0400)

System | IBM; System x3650 M2 −[7947AC1]− ;P l at f o rm | Lin ux

R e l ea s e | CentOS r e l e a s e 5 . 6 ( F i n a l )K e r n e l | 2 . 6 . 1 8 −2 3 8 . 1 2 . 1 . e l 5

A r c h i t e c t u r e | CPU = 64− b i t , OS = 64− b i t

Threading | NPTL 2. 5C om pi le r | GNU CC v e r s i o n 4 . 1 . 2 2 0 08 0 70 4 ( Red Hat 4 . 1 . 2 −5 0 ) .S ELin ux | P e rmi s s i ve

V i r tu a l iz e d | No v i r t u a l i z a t i o n d et ec te d# P ro ce s s o r ##################################################

P ro ce ss or s | p hy si ca l = 2 , c or es = 8 , v i r t u a l = 1 6 ,| h y p e r t h r e a d i n g = y e s

S p e e d s | 1 6 x 1 5 9 6 . 0 0 0Models | 16 xI nt el (R) Xeon(R ) CPU X5570 @ 2. 93GHz

11



C ache s | 16x 8192 KB

# Memory ######### ######## ######### ######## ######### ######### #T o t a l | 9 4 . 3 9 G

B MySQL configuration

[mysqld]gdb

datadir=/mlc/data

# f o r SSD#innodb_read_ahead = nonei n n o db_fl u s h _n ei gh bo r_pages = 0i n n odb_adapt i ve _f l u sh i n g_me t h od = ke e p_averagei n n o d b _ b u f f e r _ p o o l _ r e s t o r e _ a t _ s t a r t u p = 3 0 0

#innodb_adaptive_hash_index_num = 16

# # ## # f i x e d i nn od b o p t i o n si n n o d b _ f i l e _ p e r _ t a b l e = t r u ei n n o db_ dat a _f i l e _ pat h = i bdat a 1 :1 0M: au t o ex t e n d

i n n od b _ f l us h _ l og _ a t _ t rx _ c o m mi t = 2innod b_fl ush_m ethod = O_DIRECTi n n o d b _ l o g _ b u f f e r _ s i z e = 2 56M

#innodb_max_dirty_pages_pct=30

i n n o d b _ b u f f e r _ p oo l _ s i z e = 1 10G

i n n o d b _ l o g _ f i le _ s i z e = 4Gi n n o d b _ l o g _ f i l e s _ i n _ g r o u p = 2i n n o db_l o g_bl o ck_s i ze = 4096

#i n n o db_do ubl e wri t e = t ru ei n n odb_dou bl ewri t e = 0

# # ## # p l u g in o p t i o n si n n od b _ r ea d _ i o _t h r e ad s = 1 6i n n o d b _ w ri t e _ i o _ t hr e a d s = 1 6i n n o d b _ i o _ c a p a c it y = 1 0 00 0

12



# n o t i nn od b o p t i o n s ( f i x e d )

p o r t = 3 3 0 6 back_log = 50max _co nn e ct i on s = 400max _co n ne ct _e rrors = 10t a b l e _ c a c h e = 2 0 4 8max_allowed_packet = 16M b i n l o g _ c a c h e _ s i z e = 16Mmax _h e ap_t abl e_s i ze = 64M

s o r t _ b u f f e r _ s i z e = 4M j o i n _ b u f f e r _ s i z e = 4Mt h r e a d _ c ac h e _ s i z e = 1 00 0q u er y _ ca c h e_ s i ze = 0q u e ry_cach e _t ype = 0q u e r y _ c a c h e _ l i m i t = 2Mf t _mi n_word_le n = 4memlock# d e f a u l t _ t a b l e _ t y p e = InnoDBt h r e a d _ st a c k = 1 92K

t m p _ t a b l e _ s i z e = 6 4M

s e rve r−i d = 1 0

#∗∗∗ MyISAM Sp ec i fi c opti onsk e y _ b u f f e r _ s i z e = 8Mr e a d _ b u f f e r _ s i z e = 1Mr e a d _ r n d _ b u f f e r _ s i z e = 4M b u l k _ i n s e r t _ b u f f e r _ s i z e = 8Mm y i s a m _ s o r t _ b uf f e r _ s i z e = 8Mmyi s am_max _s o rt _f i l e _s i ze = 10G

# m y i s a m _ m a x _ e x t r a _ s o r t _ f i l e _ s i z e = 1 0Gmyi s am_re pai r_t h re ads = 1myisam_recover

soc ket =/var/ l i b /mysql/mysql . socku s e r= ro o tskip−grant−t a b l e s

[mysql]

13



no−auto−rehash

soc ket =/var/ l i b /mysql/mysql . sock

[ c l i e n t ]soc ket =/var/ l i b /mysql/mysql . sock

Documents

fusion-io-mysql-multi-instances-report.pdf