Delivering Schema as a Service

Copyright © 2015, Oracle and/or its affiliates. All rights reserved.
Enterprise Manager 12c
Pete Sharman
Database Architect, DBaaS EM Team
pete.sharman@oracle.com
February 2015
Schema as a Service

The Obligatory Slide About Me
• 20 years with Oracle
– DBA instructor / Education Centre Manager
– HA Consultant
– Internal services curriculum developer
– North America Sales and Consulting database lead
– RAC Development
– ST Curriculum DBA / EM team
– Product Manager, EM Product Suite
– Database Architect, DBaaS, EM Product Suite
• OakTable member
• Presented at conferences all over the place
• Owner, petewhodidnottweet.com ☺

EM 12c: Broadest, Most Complete Range of Enterprise
Services
Infrastructure-as-a-Service
(IaaS)
Infrastructure-as-a-Service
(IaaS)
Oracle VMOracle VM
VM VMVM DB DBDB
Self-Service Application/ APIs
App 1 App3App 2
Java Platform
Database-as-a-Service
(DBaaS)
Database-as-a-Service
(DBaaS)
Java Platform-as-a-Service
(Java PaaS)
Java Platform-as-a-Service
(Java PaaS)
Exadata/non-ExadataExadata/non-Exadata Exalogic/non-ExalogicExalogic/non-Exalogic
Increasing Enterprise Value

Consolidation: DBaaS Architectures
EM12c Supports Database Versions 10gR2 to 12c
Virtual Machines
Share servers
Dedicated Schema(s)
Share servers, OS
and database
Increasing Consolidation
Dedicated DBs
Share servers
and OS
Pluggable DBs
Share servers, OS
and database

EM12c DBaaS: Unmatched Architectural Choices
Virtual
Machines
Dedicated DB
Dedicated
Schemas
Pluggable DB
Consolidation
Density
Low-Moderate High Highest Highest
Management
Very complex (VM
Sprawl)
Easy
Easy to Involved
(based on required
resource isolation)
Easy
Isolation Excellent Good Least Good
Implementation
& Onboarding
Easy Easy Difficult Easy
Application
Suitability
Some (workload
dependent)
All
Home grown; requires
app validation
All but have to be
certified for
Database 12c

Cloud Taxonomy
Cloud
• Top level entity representing the collection
of all software and hardware resources
allocated to building and delivering a cloud
Zone
• A logical grouping of cloud infrastructure
resources based on functional, departmental
or geographic boundaries. Example, Finance
Zone, East Coast Zone
• Separate charge plans may be defined per
zone
• Zones can also be used to enforce access
control
Resource Pool
• A logical unit of homogeneous clustered or
non-clustered resources exhibiting common
characteristics. Example, Oracle VM server
Pool, Database Pool
Pool Pool
Pool
Pool
Pool
Pool
Pool
PoolPool
Zone
Zone
Zone
cloud

Additional concepts
• Software Library: A centralized repository of all reusable components/cloud
building blocks (scripts, assemblies, templates, profiles, etc.), accessible
from all OMSs in an EM installation.
• VM Templates: A VM image that can provision an Operating System image
• VM Assembly: One or more VM images with pre-defined associations and
late binding configuration (in OVF format)

Additional concepts
• Profile: A gold image used as a source of subsequent provisioning in
DBaaS/MWaaS on physical environments. This could be a DBCA template,
RMAN backup or export dump depending on the type of service
• Service Template: Is a standardized service definition that encapsulates the
payload (profile) and method of provisioning
• Service Instance: The state of an already deployed service. It can be in
“running” or “stopped” state.

Significance of Zones and Pools
• Significance of Zones
– Zones define the infrastructure boundary and can be mapped to logical entities like a
department or LOB
– Zones are what Self-Service Cloud user chooses to provisions into
– Chargeback can be associated with zones
– Zones can be used to define access control on physical infrastructure
• Significance of Pools
– Pools define the boundary within which a service is placed using placement policies
– Pools in Oracle VM based IaaS define the boundary of live migration and HA
– Targets within the same member of a pool must share the same downtime. Example:
Instances sharing an ORACLE_HOME must be down when the binaries are patched

Infrastructure as a Service (IaaS) using Oracle VM x86
ApplicationApplication
PlatformPlatform
cloudUser
cloud
Provider
IaaS cloud
• IaaS cloud consists of one or more
Oracle VM zones, which in turn contain
one or more Oracle VM server pools.
• Oracle VM server pool is a clustering
of Oracle VM servers. It provides a
boundary for live migration and fault
tolerance.
Server Pool
OVM Zone
Server Pool

IaaS: Sample Use Cases
Use Case 1: Zones based on Software Lifecycle
IT wants to setup a cloud to allow testers to request Linux machines. It also wants to
allow developers to deploy pre-configured application environments.
Server Pool 1
Dev Zone
Server Pool 2
Solution:
- Provide IaaS cloud with separate
Oracle VM zones for Developers and
Testers
- Setup role-based access so testers
can only access Test Zone and
developers can only access Dev
Zone
- Publish separate application
assemblies for Developers in the
Self Service Portal
IaaS cloud
Test Zone
Server Pool 1

IaaS: Sample Use Cases
Use Case 2: Zones based on LOBs
IT wants to allow business users in the Finance and HR departments to be able to request
Linux machines with various OS versions and configurations. IT wants to charge business
for the resources they consume.
Server Pool 1
Finance Zone
Server Pool 2
Solution:
- Provide IaaS cloud with separate
Oracle VM zones for HR and
Finance department
- Setup separate charge plans for HR
Zone and Finance Zone
IaaS cloud
HR Zone
Server Pool 1

Platform as a Service (PaaS)
ApplicationApplication
CloudUser
Cloud
Provider
PaaS cloud
• PaaS Cloud consists of PaaS Infrastructure Zones.
A PaaS Infrastructure Zone can either be physical or
virtual and comprises the hardware (hosts) that
make the cloud
• A virtual PaaS Zone contains one or more OVM
Zones. This way a PaaS cloud can be built on top of
Oracle VM environments.
• A PaaS Zone can contain Database Pools and/or
Middleware Pools.
• A Database Pool is a collection of homogeneous (4
digit version)
• Single Instance or RAC Oracle Homes
(Database as a Service)
• Single Instance or RAC Databases (for Schema
as a Service)
• A Middleware Pool is a collection of homogeneous
Fusion Middleware Oracle Homes.
Database Pool
PaaS Zone
Middleware Pool
PaaS Zone
OVM Zone OVM Zone

PaaS: Sample Use Cases
Use Case 1: Database as a Service (DBaaS)
IT wants to setup a database cloud to allow DBAs to request databases on demand.
There are also some developers in the Finance BU who would like to request database
schemas.
Solution:
- Provide a PaaS cloud with 2 PaaS Zones.
One for DBAs and one for Developers.
- Setup role-based access so DBAs can only
access DBA Zone and developers can only
access Dev Zone using the Self service
portal
- Publish service templates to DBAs can
request database instances and developers
can request individual schemas
PaaS cloud
Database Pool
DBA Zone
Database Pool
Database Pool
Dev Zone

Use Case 2: Schema as a Service
IT wants to setup a database cloud to allow developers to deploy applications. The Pool1
runs 11.2.0.4 Database and accommodates empty schemas for HTMLDB (APEX)
applications while Pool 2 hosts the schemas acting as data-sources for in-house Java
application
Solution:
- Provide a PaaS cloud with 2 PaaS Pools.
- Setup role-based access for individual
development groups
- Publish service templates. Service Template
associated with Pool 1 creates empty
schemas while Service Template associated
with Pool 2 creates schemas with seed data
- The databases on the two pools could be
configured with different resource
management configurations, thereby having
different QOS
PaaS cloud
Database Pool 1
Developer Zone
Database Pool 2

Use Case 3: Database as a Service (DBaaS) on Exadata
IT wants to setup a database cloud on Exadata to allow developers to request databases
of two different versions on demand.
Solution:
- Treat Exadata as a PaaS Zone
- Create Database Pools by grouping Oracle
Homes on various Exadata compute nodes
- Pool 1: Runs 11.1.0.7
- Pool 2: Runs 11.2.0.3
- Publish service templates into the Self
Service Portal for developers to deploy
database instances in the PaaS Zone
- Setup charge plans based on the zone
Database Pool 1
Compute Nodes
5 -8
Database Pool 2
Compute Nodes
1-4
PaaS Zone
Exadata

Use Case 4: Middleware as a Service (MWaaS) on Physical infrastructure
IT wants to setup a cloud to allow developers in the Finance LOB to request WebLogic
domains and databases for custom app deployment . Also developers in the HR LOB
want to request WLS domains. Resource usage by each LOB needs to be tracked.
Solution:
- Provide a PaaS cloud with 2 PaaS Zones,
one for each LOB.
- Setup role-based access to both zones.
- Publish MW service templates to the Self
Service Portal
PaaS cloud
Database Pool
Finance Zone
MIddleware Pool
MIddleware Pool
HR Zone

Use Case 5: Middleware as a Service (MWaaS) on Virtual infrastructure
IT wants to setup a cloud to allow developers and testers to request WebLogic domains
for custom app development. Resource usage for developers and testers needs to be
tracked.
Solution:
- Setup a PaaS cloud that contains one
or more PaaS Zones. Each PaaS
Zone in turn consists of one or more
Oracle VM Zones.
- Allow developers and testers to
access the PaaS Zone through the
Self Service Portal
- Setup charge plan for the PaaS Zone
and meter resource usage for every
tester or developer
PaaS cloud
PaaS Zone
Server Pool 1
OVM Zone 1
Server Pool 2
OVM Zone 2
Server Pool 1
PaaS cloud
PaaS Zone

Demo

Schema as a Service Considerations
• Security Isolation
• Operational Isolation
• Resource Isolation
• Fault Isolation
• Scalability
• High Availability

Security Isolation
• In the Schema as a Service environment, the effect of granting a privilege
or role is contained to the schema where the grant was made, thus
ensuring greater security.
– One of the main issues (if not the main issue) in a Schema as a Service environment is
namespace collision (covered next).
• May be mistakenly resolved by creating public synonyms, which of course is not recommended from
a security perspective. The result is that while by itself, Schema as a Service does not lead to reduced
security, the decisions of the administrator can end up meaning security is decreased when
compared to PDBaaS.
• For most configurations, Oracle’s out of the box database security profiles
are sufficient to limit access to data in the Schema as a Service
environment. However, it is also possible to provide deeper security using
functionality such as encryption, Database Vault and Audit Vault.

Operational Isolation
• From a backup and recovery perspective, Schema as a Service tablespaces
can be both backed up and recovered separately to each other, even
recovered to different points in time.
– Increases the operational isolation significantly.
– As more schemas are consolidated into a single database, operations that affect an
ORACLE_HOME will affect more schemas.
• However, this is offset to a certain extent by the ease with which transportable tablespaces can be
used to move the schema(s) to a different database.
• Having said that, moving a schema in this way is not quite as straightforward as moving a pluggable
database, so Schema as a Service doesn’t rank as highly as PDBaaS for that reason.
• In addition, schema-based consolidation lacks isolations from both the database lifecycle
management and independence of patching and upgrades perspective.

Operational Isolation
• However, the area that makes Schema as a Service much more difficult
from an operational perspective is the issue of namespace collision
– Namespace collision occurs because a single database cannot contain multiple copies
of the same database object in a single schema.

Namespace Collision Example
• Namespace collision means there can only be one “EMPLOYEES” table
owned by the HR user at any one point in time.
– For Schema as a Service, this is not so much a concern at the database layer as it is at
the application layer.
• The implementation of Schema as a Service in Enterprise Manager requires you to provide a schema
prefix when you issue a create schema as a service instance request, and creates the new schema
with that prefix.
– For example, if you were using the HR schema from the PROD database to create a copy of the HR schema in
the TEST database, and provided the schema prefix MYHR, the TEST database would have a schema called
MYHR_HR owning a copy of the PROD HR database objects.
– From the database perspective, then, we can indeed create multiple copies of a
schema in a single database, and each schema will be named using the naming
convention SCHEMA_PREFIX_ORIGINAL_SCHEMA_NAME, removing the issue of
namespace collision.

• However, from an application perspective, there is clearly still an issue, as
any application based off the original HR schema will expect the objects to
be owned by HR. There are at least three ways to address this:
– Private synonyms – for each user in the database that will be accessing the HR
application, create a set of private synonyms for each object used in the HR
application.
• This would need to be performed for every object used in the HR application, and for every user that
would be accessing the application. Obviously this can be done in a scripted manner, but still
involves manual intervention by the database administrator.

– Public synonyms – one way to address the need for creating private synonyms for
every user as just shown is to use public synonyms instead. However, by their very
nature, only one public synonym can be created for an object owned by a specific
user, so this removes the ability to consolidate multiple schemas into a single
database, and therefore is not really a resolution we can use.
– Create a logon trigger for each user that will use the application to include a
statement of the form “ALTER SESSION SET CURRENT_SCHEMA=MYHR_HR”. Again,
this would require manual intervention after the schema has been created.
• Of course, one could also modify the application code to change every HR
schema reference to MYHR_HR, but that is rarely something that is easily
achieved.

• The end result of this is that from an application layer, namespace collisions
cause a lot of difficulty in the Schema as a Service paradigm.
• However, prior to the advent of PDBaaS, Schema as a Service was the
consolidation model that allowed greatest consolidation to occur, and a
number of customers have used one of the namespace collision resolutions
outlined above successfully in Production environments.

Resource Isolation
• Schema(s) created by Schema as a Service are just the same as
any other database schema.
– As a result, it is quite simple to use Database Resource Manager to create resource
consumer groups, map sessions to those groups, and then assign resources to those
groups based on resource plan directives.
– However, as by its very nature you do not know which database schemas will be
created in when using Schema as a Service, these sorts of methods are all
interventions by the DBA after the schema(s) creation.

Resource Isolation
• Methods that do not require this manual intervention after the
service has been created. These include:
– Sensible creation and selection of workloads – If more than one workload is created,
the SSA user can specify the workload size that will most meet their requirements.
– Proper definition of placement constraints at database pool level – when a database
pool is created, the Self Service administrator can set maximum ceilings for resource
utilization as placement constraints. These constraints can define:
• The maximum number of database services for each database
• The maximum CPU allocation for the service request
• The maximum memory allocation for the service request
– The service instance will then be provisioned on the member that best satisfies these
placement constraints.

Fault Isolation
• Fault isolation in a Schema as a Service request is normally provided at the
schema level, so application faults in one schema will not cause other
applications to fail.
– It is also possible that login storms or an improperly configured mid-tier will cause
impact to other applications.
– In addition, as more schemas are consolidated into a single database, the more
impact a fault at the database level will have. Of course, there are some faults (such
as dropping a table incorrectly) that can be resolved at the schema level, thus
isolating the fault from other schemas in the same database.

Fault Isolation
• Once a fault has been isolated and resolved, there are two parts of the
database architecture that allow fast recoverability and thus smaller MTTR
in any database, including one used with Schema as a Service.
– Flashback functionality, including both Flashback Drop and Flashback Table:
• Flashback Drop allows you to reverse the effects of dropping a table, including any dependent
objects such as triggers and indexes.
• Flashback Table allows you to undo the effects of accidentally removing (or indeed adding) some or
all of the contents of a table, without affecting other database objects. This allows you to recover
from logical data corruptions (such as adding or deleting rows from the table) much more quickly
than you might otherwise.
– Point-in-time recoverability –
• Can be performed at the individual tablespace level, so if you have multiple schemas affected by an
issue, you can issue parallel point-in-time recovery commands to improve MTTR.

Scalability
• All databases compete for the limited hardware resources (CPU, memory,
and I/O) within a database pool, so you should ensure that a database is
guaranteed sufficient resources and also doesn’t have a detrimental impact
on other databases within the pool.
• Variety of ways to ensure customers in a Schema as a Service environment
are getting the services or resources they are paying for. These include:
– Separation of resources at the pool level
– Quotas – define the amount of memory and storage, and the number of database,
schema service, and pluggable database service requests that can be allocated.
– Workloads – workloads, based on the CPU, memory, number of sessions and storage
requirements, can be defined that can be chosen by the Self Service user at request
time.

High Availability
• Not all consumers require the same level of availability in a cloud
environment.
– HA levels are normally defined at the database level, using technologies such as RAC,
Data Guard and so forth.
– End result is that careful planning during the database pool definition is needed to
ensure different schemas with different availability requirements do not end up in
the same database pool.
– Schema as a Service consolidation model provides the capability to execute up to
Platinum level HA (RAC with Standbys).

Delivering Schema as a Service

Delivering Schema as a Service

More Related Content

What's hot (20)

Similar to Delivering Schema as a Service (20)

Recently uploaded (20)

Delivering Schema as a Service

Editor's Notes