Re:platforming // The/Datacenter// With/Apache/Mesos/

Transcription

Christos(Kozyrakis(

Why your ASF project should run on Mesos

tributed/storage/(HDD,/Flash)/x10/MWatt/x100/M /(

developers ops automation(automation(performance((efficiency((

① Datacenter/past/

Static/Partitioning/

/memcached/

/memcached/

/memcached/

/memcached/

Static/Partitioning/developers ops " automation(" automation(! performance(" efficiency(((

② Datacenter/present/

(management(at(scale((cgroups,(docker,( )(

/AuthN/Slave/Servers/Executor/Executor/Task/Task/ Jenkins(AuthZ/Slave/ ands(of(servers(

k/ Jenkins(AuthZ/Slave/ (of(the(master(((

k/ Jenkins(AuthZ/Slave/ (of(the(framework(

k/ Jenkins(AuthZ/Slave/ (of(the(slave(process(

(launching(

dependencies(Queuing(&(priorities((

nd(scalable(Mesos(master(

(needed((

d/buy less machines or run more applications!

Service/Discovery/① Watch(ZK(for(((master(changes(Mesos(Master(② � e( Slave((nginx.marathon.mesos(#(10.13.17.95(nginx. tcp.marathon.mesos(#10.13.17.95:8181((

③ Datacenter/future/

Utilization/Reality/Twitter (Mesos)[Delimitrou’14]Google (Borg)[Barroso’09]

ns,(load(spikes,(software(&(platform(changes(

Oversubscription/Frameworks/Marathon( Jenkins(② offer s1,(4cores,( ocator/AuthZ/AuthN/AuthZ/① offer s1,(4cores,( Executor/Executor/Task/Task/

Oversubscription/Frameworks/Marathon( Jenkins(③ launch tasks,(s1,(4cores,( ocator/AuthZ/AuthN/AuthZ/④ launch tasks,(4cores,( Executor/Executor/Task/Task/

Oversubscription/Frameworks/Marathon( Jenkins(② offer s1,(BE,(2cores,( ocator/AuthZ/AuthN/AuthZ/① offer s1,(BE,2cores,( Executor/Executor/Task/Task/

Oversubscription/Frameworks/Marathon( Jenkins(③ launch tasks,(BE,(s1,(2cores,( ocator/AuthZ/AuthN/AuthZ/④ launch tasks,(BE,(2cores,( Executor/Executor/Task/Task/

Oversubscription/Frameworks/Marathon( Jenkins(② Status task,(killed,( ocator/AuthZ/AuthN/AuthZ/① Status task,(killed,( Executor/Executor/Task/Task/

Interference/#/Performance/Loss/Impact of interference on websearch’s latency300%L3 Cache 300%( 300%( 300%( 300%( 300%( 300%( 300%( 264%( 123%(DRAM 300%( 300%( 300%( 300%( 300%( 300%( 300%( 270%( 122%(HyperThread110%( 107%( 114%( 115%( 105%( 117%( 120%( 136%( 300%(CPU power124%( 107%( 116%( 109%( 115%( 105%( 101%( 100%( %OK64%(0%10% 20% 30% 40% 50% 60% 70% 80% 90%Load[Lo’15]

y((cgroups)(Disk(Network(Cache(Power( ((

Isolators/#/Performance/QoS/[Lo et al’15] /bestAeffort/task/ 90%/HW/utilization/No/latency/SLO/problems/

ves/

e( ata(locality,(pricing, ((

g(

Mesos/Datacenter/developers ops ! automation(! automation(! performance(! efficiency(((

(4.15pm(–(Mesos( (Yarn( (Myriad(

Questions?/((((

References/ http://mesos.apache.org/( http://www.mesosphere.com/( https://github.com/mesosphere/mesosLdns( https://www.cs.berkeley.edu/ alig/papers/mesos.pdf( https://www.cs.berkeley.edu/ alig/papers/drf.pdf( 16ED2V01Y201306CAC024( http://web.stanford.edu/ cdel/2014.asplos.quasar.pdf( http://web.stanford.edu/ davidlo/resources/2014.heracles.isca.pdf(

((Simple,(stable,(and(scalable(Mesos(master