Resilient Cloud Application

Service Lifecycle

• Service Offering: Implementation, Description, Registration
• Service Discovery: Request, Matchmaking, Negotiation
• Service Usage: Activation, Monitoring, Optimization
• Service Termination: Withdrawal, Deactivation, Feedback

Quality dimensions

SQuaRE (Software product Quality Requirements and Evaluation - ISO/IEC 25000)

• Functional suitability
• Performance efficiency
• Compatibility
• Usability
• Maintainability
• Reliability
• Security
• Portability

Formula: Reliability = Maturity + Availability + Fault tolerance + Recoverability + Compliance

Reliability (IEEE)

The portability of failure-free software operation for a specified period of time in a specified environment.

Resiliency

The ability of an app to recover from certain types of failure and yet remain functional from the customer perspective.

Service Quality Model

Non-functional characteristics (performance, cost, compliance to standards), QoS

Quality specifications

• UML with OCL (Object Constraint Language)
• CQML+ (Component Quality Modelling Language)
• OASIS/TOSCA
• SLA (Service Level Agreement) as languages (LegalXML, XACML, WSAG, WSLA, SLAng)

Quality dimensions

• Relevance
• Added value
• Confidence
• Objectivity
• Timeliness
• Consistency
• Precision
• Completeness
• Correctness

Functionality: what a service is doing

Non-functional property: How a service is doing what it is doing

Measurable quality

• Described metrics: Authoritative vs. non-authoritative
• Measured metrics: Monitoring, attestation services
• Estimated metrics: Prediction algorithms
• Static analysis: Descriptions (completeness, timeliness)
• Dynamic analysis: Measurements (correctness)

Service metaquality (SMQ): how well is a service described

SQM caclulation and decision: Table with quality dimensions and NFPs - weighted quality dimension per metric

Replication

Replica: Repeated software component or data

• Faster response time through multiplexing
• Improved safety through backup
• Increased storage/computation/bandwidth

Taxonomy

• partial (decomposed/erasures)
• full (n-fold)

Placement

• Equal distribution
• Proportional distribution
• Absolute distribution

Optimal algorithm

• Staggered determination
• Guarantee of availability, price, capacity
• Deadline for calculation

Data

• Singular copies
• Incremental copies, backup with deltas, Hanoi strategy
• Distributed versioned replication
• Erasure coding

Erasure coding

c = a XOR b

• if a fails, recover a = c XOR b
• if b fails, recover b = c XOR a
• if c fails, recalc c = a XOR b
• 50% redundancy (2 significant, 1 redundant)

d = a XOR b XOR c

• 33% redundancy (3 significant, 1 redundant)

z = erasure(a,b,…) in Galois Field GF(2)

• arbitrary redundancy

Resilience

Fault injection

Test before releases, find issues -> Chaos engineering, live system, simulation/emulation

Software level

• Fuzzing: Invalid data input
• Data modifications
• e.g. jFuzz concolic whitebox fuzzer

System level

• Resource exhaustion
• Limits exhaustion
• Random process signaling/termination
• e.g. Chaosmonkey

Network level

• Artificial slowdown
• Random port blocking
• Protocol mutation
• e.g. Chaosmonkey

Emulation

MC-EMU: Multi cloud emulator

Behavioural models

• convergence
• incident
• random
• replay/library

Properties

• availability
• slowness
• popularity

Targets

• No-op
• Web/file server (storage)
• OS container (compute)
• L4 proxy (network)