17
!"#$%&'() +(, +,-'('./0+1() 234 '( $+0)" .5+$" .5"(+0'%. 6'$7"0/ 89.+(+: ;+/0'5< ;$'5=+0/: >=0'./%#=" ?"(0& !"$'@"0& -%,". A"0@"0 7+.", BC"7D E'@"F>! G'0/9+$ -+5='(" HA8 .15< A"0@"0 I+0- -%('/%0'() +(, 7"(5=-+0<'()

Deploying and administra1ng TAO in large scale scenarios Delivery

Embed Size (px)

Citation preview

!"#$%&'()*+(,*+,-'('./0+1()*234*'(*$+0)"*.5+$"*.5"(+0'%.*

6'$7"0/*89.+(+:*;+/0'5<*;$'5=+0/:*>=0'./%#="*?"(0&*

!"$'@"0&*-%,".*

•  A"0@"0*7+.",*BC"7D*•  E'@"F>!*•  G'0/9+$*-+5='("*

•  HA8*.15<*

•  A"0@"0*I+0-*-%('/%0'()*+(,*7"(5=-+0<'()*

A"0@"0*7+.",*,"$'@"0&*

•  A'()$"*."0@"0*."/9#*•  !9+$*."0@"0*."/9#*•  A5+$+7$"*."0@"0*I+0-*C'/=*+,@+(5",*-%('/%0'()*

A'()$"*."0@"0*."/9#*

Clients

Complete TAO installation

1Gbit/s

!9+$*."0@"0*."/9#*B+9/=%0*J*,"$'@"0&D*

Clients

del1.tao.lu

mod1.tao.lu

1Gbit/s

Complete TAO installation

Complete TAO installation

A"0@"0*I+0-K*LMMN*

del1.tao.lu del2.tao.lu

mod1.tao.lu

Fiber channel EMULEX 375

xraid1.tao.lu

portal.tao.lu

portal1

db1.tao.lu

1Gbit/s

A"0@"0*I+0-*LMMO*;?;*$%+,*7+$+(5'()*

;?;*$%+,*7+$+(5'()*+(,*7"&%(,*

•  P%/*/="*7"./*$%+,*7+$+(5'()*.%$91%(K*– +*9."0*5%--9('5+/".*C'/=*%("*."0@"0*

•  Q="(*+*."0@"0*,'".:*/="*9."0.*%I*/=+/*.&./"-*0"-+'(*./95<*

ARHS!*$%+,*7+$+(5'()*

•  H('T9"*,"$'@"0&*+55"..*#%'(/*B#%0/+$D*•  A/+15*,+/+*'.*5+5=",*'(*."0@"0*U3V*I%0*

I+./"0*,"$'@"0&*B.T9',*="+,D*

•  !&(+-'5*,+/+*'.*$%+,F7+$+(5",*7"/C""(*W*7+5<"(,*."0@"0.*

•  A=+0",*9."0*."..'%(*S!X.*•  Y+'$%@"0*#0%/"51%(*•  Z+.'$&*.5+$+7$"*•  V9$1#$"F.'/"*$%+,*7+$+(5'()*J*I+'$%@"0*

http://www.squid-cache.org/

ARHS!*A"0@"0*I+0-*LM[M*

?')=*+@+'$+7'$'/&*•  ARHS!*="+,*I+'$%@"0*#0%/"51%(*B#%0/+$[*\*#%0/+$LD*•  W*ARHS!*7+5<"(,.*C'/=*5="5<*+$'@"*I9(51%(*•  !+/+7+."*5$9./"0*•  L*-%,9$"*."0@"0.*B."#+0+/",*I0%-*,"$'@"0&*."0@"0.D*•  V%('/%0'()*+##$'5+1%(*."0@'5".*B"-+'$*+$+0-D*

•  !+/+*%(*U3S!*,0'@".*BS(/"0(+$*U3S!*.&./"-.D*

•  39/%-+/",*7+5<9#*%I*("%*\*!8*."0@"0.*•  H;A*#%C"0*7+5<9#*J*"-"0)"(5&*#%C"0*)"("0+/%0*

•  3'0*5%(,'1%('()*C'/=*I+'$%@"0*#0%/"51%(*•  ?')=*."590'/&*#%$'5'".*B]0"C+$$:*+55"..*/%*."0@"0*0%%-D*

3,-'(*5%(.%$"*•  Database management (phpMyAdmin console)

•  Usage statistics / log management •  Repository for TAO master files on neo.tao.lu

–  PHP code (updates from SVN directory) –  Compiled tests (cache) –  Result files

•  Synchronization -  Deploy TAO PHP code (rsync: neo => servers) -  Deploy test caches (rsync: neo => servers) -  Fetch results from all servers (rsync: servers => neo)

Questions ?

!"#$%&'()*+(,*+,-'('./0+1()*234*'(*$+0)"*.5+$"*.5"(+0'%.*

V%('/%0'()*\*7"(5=-+0<'()*

-%('/%0'()*

3(/?'$$*7"(5=-+0<'()*

What sort of tool is Nagios?

!  detection

!  performance data

Detection

Resources "  no disk space left "  zombie processes taking 100% cpu "  no memory left "  network bottleneck

High-level services "  database unavailable "  php sessions unavailable "  user connection unavailable

Performance data tool

!  disk space !  CPU usage !  system load !  Network bandwidth

How do we benchmark?

Nagios: supervisor

AntHill: a sort of DOS tool

AntHill benchmarking, keypoints:

!  Payload: what we ask to the services –  file downloads –  user connection –  real condition test –  results upload

!  Repartitions: how we ask it – Gaussian density –  Uniform density –  'Dot' density

Gaussian distributions

Dot distribution : all at the same time Uniform distribution : constant density

Overview of benchmarking

1000 simultaneous downloads – 'dot' distribution 10 files 1Mio each

Number of processes, function of the time client side

TSF:batch-1210077443

Number of processes, function of the time server side

TSF:batch-1210077443

1000 simultaneous downloads – 'dot' distribution 10 files 1Mio each

500 results sent, gaussian distribution, 1 hour

Number of processes, function of the time client side

TSF:batch-1209980295

Duration of the processes, function of when launched client side

TSF:batch-1209980295

500 results sent, gaussian distribution, 1 hour

Nagios and AntHill are:

#  A way to know what's going wrong #  A way to watch the server farm #  A way to benchmark TAO & TSF

Questions ?

!"#$%&'()*+(,*+,-'('./0+1()*234*'(*$+0)"*.5+$"*.5"(+0'%.*

4/="0*,"$'@"0&*-%,".*

!"$'@"0&*V%,".*•  E'@">!*

2

•  LiveCd ?

•  Used in PISA study, running on computers from schools.

Advantages

•  Open-sources liveCd solutions available

•  Independent from local operating systems, configuration and security policies.

•  Complete control on the environment (prevents user from accessing the operating system, going on the internet, etc.)

•  Secured file system possible (instruments results)

•  Supports very large scale

Drawbacks

•  Bios Issue

•  Only common hardware detected.

•  Keyboard layout

•  No storage (requires to be combined with usb key use)

!"$'@"0&*V%,".*•  G'0/9+$*V+5='("*

•  Virtual machine ?

•  Used in PIAAC study

Advantages

- Open source solutions available (VirtualBox)

- Complete control on the environment

- Independent from the host machine configuration

- Secured file systems possible (instruments and results)

- Supports large scale

Drawbacks

•  Requires installation of a VM player

•  User may still access the host machine

•  Keyboard layout

!"$'@"0&*V%,".*•  HA8*<"&*B>$'"(/*.',"*^*A"0@"0*.',"_D*

Advantages

• Control the environment (web browser only)

•  No installation required

•  Keyboard layout setup from the host machine

Drawbacks

•  Securing instruments and results challenging

•  Local webserver configuration is dependant on the host configuration

4/="0*!"$'@"0&*V%,".*

•  E%5+$*A"0@"0*^*H.7*<"&*>$'"(/.*

•  H.7*<"&*A"0@"0*^*H.7*`"&*>$'"(/.*

•  ;+#"0*;"(*•  V%7'$"*E+7*

Questions ?