The goals of this document is to investigate about document storage, what is the impact in the performance point of view, how it could be used to adjust flexibility of Tosca Service Template Handling, and define a clear roadmap how to reach all those goals.

Introduction

There are different approaches that could be used to Implement a document storage: convert a the concept model to Json text or using a database document oriented. A service template stored as document, it will make simple to have more than one service template into a db.

Also it will be possible to implement a more complex solution using namespace and imports.

Roadmap

• Convert all applications to document storage

  • performance improvements

  • Keep our current APIs, all changes will be internal

  • Provide an upgrade path to the new data structure and a rollback to the current structure

  • Old ORM layer will be not removed during this step

  • namespace and imports could be defined but not implemented

  • create more the one service template (Multi-service templates) is initially disabled in ORM layer

• Add multi-service templates support (related to POLICY-3236 - adjust flexibility of Tosca Service Template Handling)
  
  

  • Remove old ORM layer

  • Enable namespace and Multi-service templates in ORM layer

  • Adapt the current APIs to handle Multi-service templates where it needs

  • move common code in policy-model

• Define and support any other complex features not implemented yet

  • Enable import support

Investigation about PostgreSQL and json types

PostgreSQL offers two types for storing JSON data: json and jsonb. The json data type stores an exact copy of the input text, which processing functions must reparse on each execution; while jsonb data is stored in a decomposed binary format that makes it slightly slower to input due to added conversion overhead, but significantly faster to process, since no reparsing is needed. jsonb also supports indexing, which can be a significant advantage.

Because the json type stores an exact copy of the input text, it will preserve semantically-insignificant white space between tokens, as well as the order of keys within JSON objects. Also, if a JSON object within the value contains the same key more than once, all the key/value pairs are kept. (The processing functions consider the last value as the operative one.) By contrast, jsonb does not preserve white space, does not preserve the order of object keys, and does not keep duplicate object keys. If duplicate keys are specified in the input, only the last value is kept.

Note

Using PostgreSQL special type as "json" or "jsonb" rather than "text", has as consequence that all applications will lose any compatibility with H2 and MariaDB, so that solution is not recommended. The table below shows the compatibility by type.

ORM Layer using Document Storage

ORM layer using document storage (PostgreSQL or MongoDB) could be organized in two layer:

• Document layer (Domain Model to be converted in Json) - implementation has no dependency from DB

• Persistence layer - (Domain Model depend of the DB used): Entities for PostgreSQL, Documents for MongoDB

An implementation on the Document layer could be found here: https://gerrit.nordix.org/c/onap/policy/models/+/13633

Example

In the example below DocToscaServiceTemplate should be serialized to Json.

Persistence Model

In the example below the implementation of JpaToscaServiceTemplate for PostgreSQL/MariaDB (full implementation could be found here: https://gerrit.nordix.org/c/onap/policy/clamp/+/13642)

Persistence Model

In the example below the implementation of JpaToscaServiceTemplate for MongoDB (full implementation could be found here: https://gerrit.nordix.org/c/onap/policy/clamp/+/13615)

Persistence Model

Converters

Jakarta and Spring do no support json Type, but we can use Converters to convert DocToscaServiceTemplate  to a Json String.

Converters

Note: Serialization and deserialization in Json is already used in policy-model (Un example could be found here [JpaToscaPolicy.java]). @Converter is just an elegant way to do the same thing.

ToscaServiceTemplate Table

<sup>clampacm=# \d ToscaServiceTemplate
                    Table "public.toscaservicetemplate"
     Column      |          Type          | Collation | Nullable | Default
-----------------+------------------------+-----------+----------+---------
 name            | character varying(120) |           | not null |
 version         | character varying(20)  |           | not null |
 servicetemplate | text                   |           |          |
Indexes:
    "toscaservicetemplate_pkey" PRIMARY KEY, btree (name, version)


clampacm=# select * from public.ToscaServiceTemplate;
        name        | version | servicetemplate
--------------------+---------+-----------------
 PMSH_Test_Instance | 1.0.0   | 16505
(1 row)


clampacm=# select convert_from(lo_get(servicetemplate::oid), 'UTF8') from toscaservicetemplate;
{"tosca_definitions_version":"tosca_simple_yaml_1_3", ...</sup>

Proposals

The new ORM layer will be implemented into an additional package and the old layer will be not changed. My propose options are shown below:

1. Save ToscaServiceTemplace as Json String in a single Entity:

   • Each Service Template is stored as a JSON "LOB"

   • It is compatible with H2, MariaDB and PostgreSQL

   • It is an additional code for policy-models and a medium impact for all applications that are using it

   • That solution is compatible with Applications not implemented in Spring

2. MongoDB/Cassandra

   • Document oriented approach full supported by SpringBoot (not needs Converters)

   • Compatible only with MongoDB/Cassandra (MongoDB and Cassandra are not compatible to each other)

   • It is an additional code for policy-models and a huge impact for all applications that are using it (all repositories and persistence classes have to change to a Document oriented classes)

   • Unit tests need an Embedded Server (example for cassandra: EmbeddedCassandraServerHelper or CassandraContainer)

   • Spring Boot has own annotations for Documents. Eventually for application not in Spring Boot, it needs additional dao style implementation

Note

• Using document storage, it involves only the ORM layer, it does not change the functionality of the application (first step of Roadmap)

• Business logic could be optimized reducing the number of access to DB: read one time a service template and using it during the elaboration rather then fetch data from data from DB for each single search (load a property, load a list of ToscaNodeTemplate ecc..)

Benchmark Performance of runtime-acm

In order to generate the benchmark I have used (into a laptop) a Virtual Machine whit the follow configuration:

• 8192 Mb

• 2 CPU

For the tests:

• Jmeter to generate requests (same used by performance tests)

• Prometheus for monitoring

• DMaap simulator

• Participant simulator

• MariaDB/PostgreSQL/MongoDB

The existing system

Hibernate/Mariadb. Tosca Service template is saved as a schema entity relation.

Open

Using Json in MariaDB

Hibernate/Mariadb. Tosca Service Template is saved into a longtext as Json.

Open

Using Json in Postgres

Hibernate/PostgreSQL. Tosca Service Template is saved into a text type as Json.

Open

MongoDB

MongoDB. Tosca Service Template and all other entities (Participants and AutomationComposition) are saved as MongoDB Document.

Open

id cannot have dot '.' in MongoDB : solved with minimal configuration

Implementation (First step of Roadmap)

Current ORM layer for ToscaServiceTemplate

Open

ORM layer for ToscaServiceTemplate Proposed (during first step in Roadmap)

Simple-concepts entities will be maintained but not used.

Open

Namespace and Import service templates (second and third step in Roadmap)

In TOSCA language namespace is supposed to be a URI. Name and version of service template,  are not defined in TOSCA language, not used in yaml files, and not present in all examples neither Unit Tests, but they are used as primary key of the ToscaServiceTemplate table and they are used in REST endpoints as id of a resource.

• TOSCA language could be found here: https://docs.oasis-open.org/tosca/TOSCA-Simple-Profile-YAML/v1.3/os/TOSCA-Simple-Profile-YAML-v1.3-os.pdf

• A yaml file contains only one service template

• TOSCA Service Templates MUST always have, as the first line of YAML, the keyword “tosca_definitions_version” with an associated TOSCA Namespace Alias value.

• Each service template has a unique name space

• A service template could import other service templates

• namespace as URI could be an issue if used as id of a resource

• name could be extracted as host of the namespace, and version could be extracted from the postfix of the namespace.

  • A pattern could be like a http://{host_namespace}/service/template/{x}/{y}/{z}. As example "http://tosca/service/template/0/0/1" we can extract name: "tosca" and version: "0.0.1"

  • Version could be omitted: http://{host_namespace}/service/template. As example "http://tosca/service/template/onap/org" we can extract name: "tosca" and default version: "0.0.1"

• http://tosca/service/template/onap/org could be used as default namespace

• name and version will be not present into the yaml file (service template) as now, but we can continue to use them as id of a resource

• host will be used as key to reference to external service template, and with "namespace_prefix" will be used to avoid collisions with same host but different version

• about type and type_version backward compatibility still valid

• type_version will be deprecated, type will be used as formatted string as follow:

  • {type}

  • {type}:{type_version}

  • {host_namespace}:{type}:{type_version}

tosca001.yaml

tosca001.yaml

ServiceTemplate002.yaml

ServiceTemplate002.yaml

ServiceTemplate003.yaml

ServiceTemplate003.yaml

Tosca Service Template Handling

The current work in progress is available here: https://gerrit.nordix.org/c/onap/policy/clamp/+/13755

Validation

Validation in current ORM layer

As the first step of Roadmap is backward compatible, the current validation will be maintained as it is, but it should be adapted to new DocToscaEntities. In the second step do not need other changes.

"type_version" is optional and "0.0.0" is the default value. The "Key" is used for the validation to find if the ToscaEntity exists.

Examples:

"derivedFrom" is referenced to a ToscaEntity of the same type and placed in same collection.

Validation in third step

In the last step, due the introduction of imports and namespace support, it needs to extend the validation.

"type_version" is optional and "0.0.0" is the default value, "host_namespace" is optional and "tosca" is the default value, type could be used as formatted string: {host_namespace}:{name}:{version}. The "Key" is used for the validation to find if the ToscaEntity exists in same ServiceTemplate or in other one.

Examples:

"derivedFrom" should have same logic as before and it could be used as formatted string: {host_namespace}:{name}:{version}.

Policy-Api

Enabling namespace and multi-service templates

Update functionality (second and third step of Roadmap)

There are two kind of update:

• values (example "description", "metadata", ...)

• data structure (example Create or Delete of a ToscaEntity, or Update a reference as "type", "type_version" or "host_namespace")

Proposals

• We could consider Service templates references to each other as a DAG ( directed acyclic graph), where Service templates are nodes, and a references as edges. Cyclic references are not allowed.

• Should be available a functionality to load a full version of the service template that contains own tosca entities and also all tosca entities referenced from other service templates (this functionality in for validation or business logic purpose and will be not visible to the end user)

• A service template cannot be deleted if there are references to it

• Any Tosca Entity in a common service template cannot be deleted if there are references to that service template

• Any Tosca Entity in a common service template cannot change "host_namespace", "type" and "type_version" if there are references to that service template

• Any Tosca Entity in a common service template cannot change values if there are automation compositions referenced to that service template (and references to that service template?)

• A Tosca Entity cannot be add in a common service template if there are automation compositions referenced to that service template

• Functionality as create, update and delete of a ToscaEntity in a service template should be validated and restricted to the service template itself. (Example: update property values of a service template should never update property values from other service templates referenced)

• To reduce the complexity, could be useful to save additional information about the service template:

  • A full version of the service template that contains own tosca entities and also all tosca entities referenced from other service templates. it will be used in read-only for business logic purpose

  • A table to save relations between service templates

  • A table to save relations between service templates and automation compositions (with all service templates referenced)

Conclusion

• Move to document storage does not change the functionality of the application. Example if clampacm move to document storage, and policyadmin not yet, those applications will continue to work fine.

• Enable namespace and enable to handle more than one service template, impacts functionalities and applications dependency. Example if clampacm has namespace enabled, and policyadmin not yet, due the dependency, they could have some issues.

• Document storage Optional?: using SpringBoot profiles and interfaces it could be possible to have a double implementation and run only one, (based on parameter in properties file), but it does not make sense due the second point.