21
Cloudward Bound: Planning for Beneficial Migration of Enterprise Applications to the Cloud B95b01023 鍾鍾鍾

Cloudward Bound: Planning for Beneficial Migration of Enterprise Applications to the Cloud

Embed Size (px)

DESCRIPTION

Cloudward Bound: Planning for Beneficial Migration of Enterprise Applications to the Cloud. B95b01023 鍾宇彥. Migrating reachability policies. Abstraction and problem formulation ACL migration algorithm Granularity of policies in R Deriving Rnew Partitioning Rnew Installing Rnew. - PowerPoint PPT Presentation

Citation preview

Page 1: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Cloudward Bound: Planning for Beneficial Migration ofEnterprise Applications to the Cloud

B95b01023 鍾宇彥

Page 2: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Migrating reachability policies

• Abstraction and problem formulation• ACL migration algorithm– Granularity of policies in R– Deriving Rnew– Partitioning Rnew– Installing Rnew

Page 3: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Abstraction and problem formulation

• Correctness– Old and new must have the same result

• Migrating entity– At least: a server

• Reachability Matrix– Access Control List (ACL)– Avoid unwanted traffic

Page 4: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Reachability Matrix

a: access control list (ACL)

• Record direction, ACL must encounter– Like the directed graph presented by adjacency matrix

Page 5: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

ACL migration algorithm• Filter domain (Fa) – Set of origin-destination(OD) entity communication

pairs(i,j) filtered by ACL a– Scalability• w/o setting domain, setting rule between each pairs

Fa2 comprise:

source destination

BE1 fe1

BE1 fe2

BE2 fe1

BE2 fe2

INT fe1

INT fe2

Page 6: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

ACL migration algorithm

• Deriving Rnew

– Infer Rold from LDC– Translate :• Ex: IP mapping

• Partitioning Rnew

– Filtered by LDC or CDC– Filter the package before remote communication

LDC: Local Data CenterCDC: Cloud Data Center

Page 7: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Rnew matrix

Partition:Dotted: filtered in CDCGray: filtered in LDC

Two submatrices extracted from ACL a2

Page 8: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Installing Rnew - Submatrix extraction

• For scaling: keep number of submatrices small• Greedy choice maximum subset

Fa2(LDC,1)

Fa2(LDC,2)

Fa2(LDC,1)source destination

BE1 fe1

BE1 fe2

BE2 fe1

BE2 fe2

Fa2(LDC,2)

source destination

INT fe2

Page 9: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Installing Rnew

• Locating placement– Place ACL at edge-cut-set– Computing minimum edge-cut-set for each submatrices

Fa2(LDC,1)source destination

BE1 fe1

BE1 fe2

BE2 fe1

BE2 fe2

BE1

BE2

fe1

fe2

ACL a2

Theoretical ACL place location for Fa2(LDC,1)

Page 10: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Installing Rnew - Locating placementCDC ACL placement

LDC ACL placement

BE1

BE2

fe1

fe2

AR

BR

AR

AR Red: path to fe1Blue: path to fe2

Fa2(LDC,1) physical deployment presented by graph

INT

i/p

i/p

i/p

i/p

o/p

o/p

o/p

o/p

Page 11: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Installing Rnew - Locating placement• Edge-cut-set– Find minimum cut– Place ACL on the interface closer to source

Cut: separate the src and des

ACL a2

ACL a2

Page 12: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Installing Rnew - Generating ACL Configuration

• Old ACL configuration might fail1.Before migration (local) 2.After migration

Page 13: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Installing Rnew - Generating ACL Configuration

• Traffic domain D(L) : all OD pairs path by location, “L”, placed ACL– Separate legal and illegal OD pairs in D(L)– Scoped, Isolated form : different default permission

Page 14: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

ACL migration algorithm

Page 15: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Evaluation

• Deriving model parameter(linear programming)– Data flow, communication…

• Migration benefit calculation• Migration strategies– Delay V.S Cost– Security policy • Ex: keep database servers in local data center

Page 16: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Evaluation – case 1

Internal request ~80%External request ~20%

Page 17: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Evaluation – case 1

CDF: Cumulative Distribution Function

Page 18: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Evaluation – case 2

Page 19: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Evaluation – case 2

• Policy : w/o BE migration

Page 20: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Evaluation - summery

• The optimal migration not only consider the CPU usage and storage, but communication …etc

• Delay bound consideration• Relative size of transaction between different

components may determine the optimal strategy• Sensitivity to the cost of internet communication

Page 21: Cloudward  Bound:  Planning  for Beneficial Migration of Enterprise Applications to the Cloud

Conclusion

• Benefits of hybrid cloud deployment• Importance and feasibility of migration decision• Provide feasible reconfiguration of reachability

policies method