More	  Bang	  for	  your	  Buck	  Sharon	  Wagner	  Co-­‐Founder	  &	  CEO,	  Cloudyn	  Microso:	  Cloud	  Day	  April	  	...
Who	  am	  I?	  •  Director,	  Products,	  Oblicore	  	  •  Sr.	  principal,	  Cloud	  Business	  Unit,	  CA	  •  Co-­‐Fou...
How	  do	  you	  measure	  your	  cloud	  efficiency?	  
Overwhelming	  informaOon	  	  #	  of	  deployment	  cycles,	  idle	  servers,	  costs,	  %	  u9liza9on,	  jobs,	  applica...
Cool.	  	  Now	  we	  have	  a	  good	  grasp	  on	  our	  cloud!	  
Usage	  trend:	  compute	  uOlizaOon	  Most virtual servers are significantlyunderutilized. 	  •  Average	  yearly	  CPU	 ...
Usage	  trend:	  pricing	  Most customers use the leasteconomical pricing modules.	  •  26% reserve	  capacity	  upfront• ...
Actual	  usage,	  perfect	  and	  over	  provisioning	  
How	  do	  you	  measure	  your	  car	  efficiency?	  
Keep	  it	  simple	  
The	  only	  thing	  that	  maSers	  is	  the	  	  business impact!	  
User	  Experience	  =	  Workload	  Capacity	  OK,	  lets	  keep	  it	  really	  simple	  ?
Houston,	  we	  have	  a	  problem	  Cloud	  capacity	  is	  	  ∞	  
Cloud	  Value	   =	  Cost	  Compute	  Units	  The	  right	  metric	  
What	  does	  this	  mean?	  	  What	  should	  I	  do?	  
1.  Think	  small	  –	  rightsizing	  	  2.  Go	  green	  –	  recycle	  and	  reHre	  3.  Use	  your	  cash	  –	  reserve	...
How	  can	  we	  find	  opOmizaOon	  opportuniOes?	  Bringing	  real	  cloud	  usage	  data	  from	  500	  IaaS	  customers...
Recycling	  sample:	  storage/compute	  Surprise! You have storage.	  •  Typically	  represents	  15%	  of	  cloud	  spend...
Sizing	  sample	  One large cappuccino with an extra5GB of RAM please…	  Coffee	  customizaHon,	  Starbucks	  @	  IaaS	  If...
Sizing	  sample:	  compute	  Comparing large to x-large forRDBMS workload: 	  Spec	   	  Large	   X-­‐Large	  RAM	   7.5	 ...
Pricing	  opOmizaOon	  *Cloud vendors love charging less	  *	  Yep,	  this	  is	  not	  a	  typo,	  but	  they	  have	  to...
Pricing	  trend:	  reserved,	  on-­‐demand	  You will have to reserve your capacity	  93%	  of	  the	  on-­‐demand	  insta...
Price	  opOmizaOon	  Why they love charging you less?	  •  Because	  they	  have	  no	  choice	  •  Customer	  saHsfacHon	...
 $-­‐	  	  	  $200	  	  	  $400	  	  	  $600	  	  	  $800	  	  	  $1,000	  	  	  $1,200	  	  	  $1,400	  	  	  $1,600	  	 ...
Common	  reservaOon	  mistake	  Safe	  ReservaHon	  	  OpHmal	  ReservaHon	  
Summary:	  the	  cycle	  
Cloud	  Economics	  @cloudyn_buzz	  Cloudyn.com	  |	  blog.cloudyn.com	  sharon@cloudyn.com	  Thank	  you	  
Upcoming SlideShare
Loading in …5
×

How to get the most out of your cloud - Microsoft Cloud Day

1,902 views

Published on

Organized by Microsoft Israel R&D Center, this MS cloud event represents the first of its kind for start-ups in Israel. Cloud Day brought thought leaders together, including Cloudyn's CEO, Sharon Wagner, to provide real insights and know-how on all the latest cloud computing.

Published in: Technology, Business
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,902
On SlideShare
0
From Embeds
0
Number of Embeds
5
Actions
Shares
0
Downloads
41
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

How to get the most out of your cloud - Microsoft Cloud Day

  1. 1. More  Bang  for  your  Buck  Sharon  Wagner  Co-­‐Founder  &  CEO,  Cloudyn  Microso:  Cloud  Day  April    23rd  2013  How  to  get  the  most  out  of  your  cloud  
  2. 2. Who  am  I?  •  Director,  Products,  Oblicore    •  Sr.  principal,  Cloud  Business  Unit,  CA  •  Co-­‐Founder  &  CEO,  Cloudyn™  Cloud  Economics  @cloudyn_buzz  Cloudyn.com  |  blog.cloudyn.com  sharon@cloudyn.com  
  3. 3. How  do  you  measure  your  cloud  efficiency?  
  4. 4. Overwhelming  informaOon    #  of  deployment  cycles,  idle  servers,  costs,  %  u9liza9on,  jobs,  applica9on  performance,  speed,  databases,  network  bandwidth,  server  performance,  end  user  performance,  applica9on  response  9me,  revenue,  green  benefits.      
  5. 5. Cool.    Now  we  have  a  good  grasp  on  our  cloud!  
  6. 6. Usage  trend:  compute  uOlizaOon  Most virtual servers are significantlyunderutilized.  •  Average  yearly  CPU  uHlizaHon  of  17%    •  Max  RAM  uHlizaHon  of  64%By  looking  at  CPU,  Memory,  I/O,  Network:  
  7. 7. Usage  trend:  pricing  Most customers use the leasteconomical pricing modules.  •  26% reserve  capacity  upfront•  71% runs  on-­‐demand  virtual  instances  How  customers  pay  cloud  vendors:  93%  of  the  on-­‐demand  instances  should  be  reserved.  
  8. 8. Actual  usage,  perfect  and  over  provisioning  
  9. 9. How  do  you  measure  your  car  efficiency?  
  10. 10. Keep  it  simple  
  11. 11. The  only  thing  that  maSers  is  the    business impact!  
  12. 12. User  Experience  =  Workload  Capacity  OK,  lets  keep  it  really  simple  ?
  13. 13. Houston,  we  have  a  problem  Cloud  capacity  is    ∞  
  14. 14. Cloud  Value   =  Cost  Compute  Units  The  right  metric  
  15. 15. What  does  this  mean?    What  should  I  do?  
  16. 16. 1.  Think  small  –  rightsizing    2.  Go  green  –  recycle  and  reHre  3.  Use  your  cash  –  reserve  capacity  Three  steps  to  get  more  out  of  your  cloud  Nego?a?on  has  zero  impact…  
  17. 17. How  can  we  find  opOmizaOon  opportuniOes?  Bringing  real  cloud  usage  data  from  500  IaaS  customers  into  the  mix:  ~2.5mVirtual  instances,  thousands  of  databases  and  billions  of  storage  objects  monitored  in  the  survey.  
  18. 18. Recycling  sample:  storage/compute  Surprise! You have storage.  •  Typically  represents  15%  of  cloud  spend    •  Only  12%  are  using  cheaper  storage  opHons  •  15% of  volumes  are  disconnected  Does  it  make  sense  to  keep  the  light  on  when  you  leave  the  room?    
  19. 19. Sizing  sample  One large cappuccino with an extra5GB of RAM please…  Coffee  customizaHon,  Starbucks  @  IaaS  If  you  do  it  to  your  coffee,  why  not  treat  your  instances  the  same?  It’s  20%  of  your  monthly  cost.  
  20. 20. Sizing  sample:  compute  Comparing large to x-large forRDBMS workload:  Spec    Large   X-­‐Large  RAM   7.5  GB   15  GB  CPU   4  EC2  CU   8  EC2  CU  Storage   850  GB   1690  GB  I/O  Performance   Moderate   High  Large  instance,  volume-­‐opHmized  with  500  MBPS  provisioned  IOPS  performed  beSer  than  single  x-­‐large  and  cost  less!  
  21. 21. Pricing  opOmizaOon  *Cloud vendors love charging less  *  Yep,  this  is  not  a  typo,  but  they  have  to  run  a  business  too…    So,  how  does  it  work?  
  22. 22. Pricing  trend:  reserved,  on-­‐demand  You will have to reserve your capacity  93%  of  the  on-­‐demand  instances  should  be  reserved.  •  Requires  one  Hme  payment,  pay  less  for  each  hour  •  Resource  availability  is  guaranteed  •  71%  of  instances  runs  on-­‐demand,  26%  runs  reserved  
  23. 23. Price  opOmizaOon  Why they love charging you less?  •  Because  they  have  no  choice  •  Customer  saHsfacHon  •  The  upfront  payment  •  Capacity  projecHon  •  The  Jevons  paradox   Hi,  It’s  me,  William  Stanley  Jevons  
  24. 24.  $-­‐      $200      $400      $600      $800      $1,000      $1,200      $1,400      $1,600      $1,800      $2,000    Jan   Feb   Mar   Apr   May   Jun   Jul   Aug   Sep   Oct   Nov  VM  Cost  Time  On-­‐Demand   Reserve  X  Month   Reserve  Y  Month  Large  instance:  on-­‐demand  vs.  reserved  capacity  comparison  Savings  Breakeven  aoer  4mon,  30%  RunHme  
  25. 25. Common  reservaOon  mistake  Safe  ReservaHon    OpHmal  ReservaHon  
  26. 26. Summary:  the  cycle  
  27. 27. Cloud  Economics  @cloudyn_buzz  Cloudyn.com  |  blog.cloudyn.com  sharon@cloudyn.com  Thank  you  

×