Service Dependencies

dependee A depends on dependency B

dependency B is depended on dependee A

• static vs dynamic
• visible vs hidden
• explicit vs implicit
• strict vs tolerant

Dependencies in software

• code (executable)
• configuration (declarative)
• data (binary)

Importance

• if B goes down, A goes down too
• hidden cost

Dependency manifestation

• tight coupling / fixed binding
• loose coupling / late binding

Dependency type

• inter-service
• intra-service (e.g. libs)

Dependency consideration

• transitivity (recursive algorithms)
• data sources (DaaS model)
• deployment (dependency affinity)
• security
• consistency (strict, sequential, casual, eventual)

Dependency analysis

Matrix

Matrix structure for binary or qualified binary depends/not-depends relationship:

  1 2 3 n
1 .   .
2   . . .
3 .   . .
m       .

Graph

Associative graph structures: Directed acyclic, weighted or unweighted graph:

ServiceA -> ServiceB -> ServiceD
            ServiceC ->

Slower but more space-efficient than matrix and sub-structure capable.

Dependency Resolution Techniques

A depends on B if

• declared as such
• A invoked B (traceable)
• logic of A depends on state withing B (non-traceable)

Analyse statically (description) or dynamically (execution).

Dynamic analysis requires

• correlation vs causation
• small data sets
• clustering of services

Peddycord

Passive network monitoring and analysis -> Network service dependency miner

Logarithm based ranking scheme -> Ratio weight(A->B)/weight(A) or logarithm log(weight(A))*weight(A->B)

Frequency inference -> confidence in dependency candidates

Dependency resolution

Immediate resolution

• simple transitive inclusion
• solution not guaranteed

Interactive resolution

• with feedback
• iterative improvements
• may complement immediate resolution upon escalation

Composite services

Offer a single service interface and distribute requests to multiple services within the composition (parallel, serial or more complex routing): requires knowledge of dependencies.

Advantages of services

• higher availability (QoService)
• flexibility to switch (QoExperience)

Techniques

Service Orchestration

• Multiple services form another one
• Centrally managed and executed
• Based on workflows that require service interfaces

• Examples:

  • Business Process Execution Language (BPEL)
  • Yet Another Workflow Language (YAWL)
  • Dynamic Service Orchestration Language (DSOL)
  • AWS Step Functions

Resource Orchestration

• Workflows require resources
• Resource allocation performed centrally

• Examples:

  • Heat Orchestration Template (HOT)
  • Docker Compose
  • Kubernetes Descriptors

Choreography

• Interaction protocol between autonomous services
• No central point of control
• Difficult decentralised enactment

• Example:

  • BPMN 2.0
  • Chor
  • CHOReVOLUTION

Bundling

• Multiple services used in a bundle
• Less administrative overhead
• Less flexibility for exchanging single services

• Example:

  • USDL

Multiplexing

• Multiple services used in parallel handling partial requests
• Flexible redundancy systems
• More administrative overhead
• Higher cost due to multiple candidates

Mashup Service Bus/Mash

Artefacts

Bundles: WAR, SAR, BPR, ZIP, Helm charts

Executable code: .class, .py

Service descriptions: WSDL, …

SLA templates: WSLA, WS-Agreement

Deployment descriptors: HOT, PDD, web.xml, k8s yaml/json, Help templates

Dependencies by reference (e.g. Docker Compose): need resolution before execution

Dependencies by embedding (e.g. Helm charts): no resolution needed but transitivity