OpenStack architecture: a high-level view

An overview of the OpenStack cloud platform and its internal architecture. Presented at the DC Python Meetup on Sept. 3, 2013.

All slides are licensed CC BY SA 3.0 unless explicitly credited to another source. See http://creativecommons.org/licenses/by-sa/3.0/us/

Transcript

1. OpenStack  architecture:     a  high-­‐level  view   Lorin  Hochstein   Nimbis  Services   lorin@nimbisservices.com   @lhochstein  
2. Where  we’re  going  today   •  What  is  OpenStack?   •  What  does  the  architecture  look  like?   •  How  do  I  get  access  to  an  OpenStack  cloud?   •  How  do  I  get  involved  with  the  project?  
3. What  is  OpenStack?  
4. OpenStack  provides  users  access  to   compute,  networking  and  storage  
5. Access  compute  resources   Start 5 servers, each with: • 2 CPUs • 4 GB of RAM • 20 GB of storage User Suspend server "X" Snapshot server "X"
6. Access  networking  resources   Create private network with subnet 10.2.12.0/24 User Attach IP address 203.0.113.19 to server "X" Allow access to ports 80 and 443 for the "web" group
7. Access  block  storage  resources,     which  aGach  to  servers  like  disks   Create a 100 GB data volume and attach it to server "X" User Take a snapshot of volume "Q"
8. Access  object  storage  resources,   which  are  accessed  over  HTTP   Store file family.png in container "photos" User Make file logo.png in container "acme" visible to everyone
9. OpenStack  includes  a  web-­‐based   Dashboard  service   Credit:  OpenStack  Compute  Admin  Guide  
10. Users  can  also  make  requests  with   command-­‐line  tools  or  REST  API   $ nova boot --image precise --flavor m1.small myinstance Client HTTP POST /servers { "server": { "imageRef": ... "flavorRef": ... "name": "myinstance", } }
11. Also...  Python  APIs!   import novaclient.v1_1.client as nvclient creds = ... nova = nvclient.Client(**creds) image = nova.images.find(name="precise") flavor = nova.flavors.find(name="m1.small") nova.servers.create(image=image, flavor=flavor, name="myinstance")
12. ECOSYSTEM SIZE 
 (Members + Sponsors + Supporters) CUMULATIVE 
 CONTRIBUTORS AVERAGE MONTHLY 
 CONTRIBUTORS PATCHES
 MERGED 998 230 7,260 (Grizzly Release Cycle – Six Months) 209
 Companies Members: 24 Total
 Sponsors: 35 Total Supporters: 150 Total TOP 10 Countries 1) United States
 2) China
 3) India
 4) Great Britain
 5) Australia 9,511 6) France
 7) Russia
 8) Canada
 9) Ireland
 10) Germany INDIVIDUAL MEMBERS Project  is  very  acRve  with  many   contributors   Credit:  hGp://www.openstack.org/assets/markeRng/presentaRons/GenericPlaTormDeckMay-­‐23-­‐2013.pptx   Data  from  May  2013  
13. Large  industry  players  among   OpenStack  plaRnum  &  gold  members   8 Credit:  hGp://www.openstack.org/assets/markeRng/presentaRons/GenericPlaTormDeckMay-­‐23-­‐2013.pptx  
14. No  single  company  dominates  the   project   Credit:  stackalyRcs.com   Metric:  lines  of  code   Release:  Havana   Projects:  all  
15. What  does  the  architecture  look   like?  
16. System  is  implemented  as  a  collecRon   of  RESTful  web  services   Credit:  hGp://www.solinea.com/2013/06/15/   openstack-­‐grizzly-­‐architecture-­‐revisited/  
17. Compute  service  manages  virtual   machines  on  compute  nodes  
18. Image  service  provides  a  catalog  of   virtual  machine  images  
19. IdenRty  service  handles  authenRcaRon   and  maintains  a  catalog  of  all  services  
20. Network  service  manages  network   connecRvity  for  virtual  machines  
21. Block  storage  service  manages  block   devices  that  aGach  to  virtual  machines  
22. Object  storage  service  manages  binary   objects  (files)  accessed  via  HTTP  
23. Dashboard  service  provides  web   interface  for  end-­‐users  &  admins  
24. Let’s  look  at  an  example  
25. Alice  is  a  web  dev  who  needs  an  Ubuntu   server  to  deploy  her  app  for  tesRng  
26. Start  a  server  using  the  ‘ubuntu12.04’   image    and  a8ach  it  to  network  ‘mynet’  
27. 1.  Get  the  compute  endpoint  and  an   authenRcaRon  token  
28. 2.  Start  a  new  compute  server  
29. 3.  Download  the  ubuntu12.04  image  
30. 4.  Retrieve  the  image  from  the  object   store  
31. 5.  Get  info  to  determine  how  to  connect   the  virtual  machine  to  the  ‘mynet’  network  
32. 6.  Start  the  virtual  machine  
33. Alice  sees  that  her  server  is  now   running  and  connects  to  it  via  ssh.  
34. Each  OpenStack  service  is   implemented  as  Linux  daemons     Example:     Compute  Service  
35. Daemons  use  non-­‐blocking  I/O  with   eventlet  to  handle  mulRple  requests   OpenStack daemon Eventlet
36. Daemons  maintain  persistent  state   using  an  SQL  database  via  SQLAlchemy   OpenStack daemon SQLAlchemy
37. Daemons  communicate  using  remote   procedure  call  over  message  queue   OpenStack daemon kombu eventlet. green. zmq qpid message queue
38. Let’s  go  back  to  the  earlier  example,   and  zoom  in  on  the  Compute  service   Credit:  hGp://www.solinea.com/2013/06/   15/openstack-­‐grizzly-­‐architecture-­‐revisited/  
39. “Start  a  new  compute  server”  request   comes  in  to  the  nova-­‐api  daemon  
40. Ask  nova-­‐scheduler  to  find  a  compute   node  that  can  fulfill  the  request  
41. Dispatch  the  request  to  nova-­‐compute   on  a  node  that  has  enough  resources  
42. Download  the  image  from  the  Image   service  
43. Get  network  connecRon  info  from   Networking  service  
44. Ask  the  hypervisor  to  start  the  virtual   machine  
45. OpenStack  is  a  toolkit  for  building   clouds  using  third-­‐party  technologies   Compute   Storage   Network  
46. OpenStack  uses  driver-­‐based   architecture  to  support  different  techs   compute_driver=xenapi.XenAPIDriver volume_driver=cinder.volume.drivers.glusterfs.
 GlusterfsDriver core_plugin=neutron.plugins.ryu.
 ryu_neutron_plugin. RyuNeutronPluginV2 Compute  service:  Xen  driver   Block  storage  service:  Gluster  driver   Networking  service:  Ryu  plugin  
47. How  do  I  get  access  to  an   OpenStack  cloud?  
48. There  are  several  public  clouds  built   on  OpenStack  
49. TryStack.org  is  a  free  OpenStack   sandbox  for  community  use  
50. DevStack:  shell  scripts  to  deploy   OpenStack  on  a  single  node  
51. How  do  I  get  involved  with  the   project?  
52. Community  help  is  available  via     web,  email  &  IRC   •  DocumentaRon:  hGp://docs.openstack.org   •  Mailing  lists:  hGp://lists.openstack.org   – openstack@lists.openstack.org     – openstack-­‐operators@lists.openstack.org     •  QA  site:  hGp://ask.openstack.org   •  IRC:  #openstack  on  irc.freenode.net    
53. Project  is  very  open  to  new   contributors   •  hGps://wiki.openstack.org/wiki/ How_To_Contribute   •  #openstack-­‐101  on  irc.freenode.net:     help  for  new  contributors   •  Developer  mailing  list:   – openstack-­‐dev@lists.openstack.org   – hGp://lists.openstack.org/cgi-­‐bin/mailman/ lisRnfo/openstack-­‐dev  
54. Q:  Where  do  I  start?   A:  Log  a  bug  against  the  documentaRon  when   you  find  something  unclear  or  incorrect     1.  Create  an  account  on  launchpad.net   2.  Go  to  hGps://bugs.launchpad.net/                        openstack-­‐manuals/+bugs   3.  Click  on  “Report  a  bug”  in  the  top-­‐right  
55. Fin