Evolution of the DBA to Data Platform Administrator/Specialist

Tony Rogerson
Microsoft Data Platform MVP
tonyrogerson@sqlserverfaq.com
@tonyrogerson

 Professional
◦ 29 years of Database experience – (6 on DB2, 1 on Oracle
and 23 on SQL Server)
◦ Freelance SQL Server and Data Platform specialist
◦ Fellow BCS, Masters in BI, PGCert in Data Science
◦ I also do F# (and the less relevant cousin C#)
 Community
◦ Founder member of UK SQL User Group,
SQLServerFAQ.com, DataIdol.com, DDD, SQLBits and SQL
Relay
◦ Microsoft SQL Server MVP since 1997, and now a Data
Platform MVP
◦ Technical blog:
http://sqlblogcasts.com/blogs/tonyrogerson (legacy)
http://dataidol.com/tonyrogerson (General DP blog)
http://sqlserverfaq.com/tonyrogerson (MS DP blog)

Group discussion – I can only discuss from
what I’ve seen myself over the past few years
and recent while looking for work

 What’s a Data Platform?
 Define the traditional Database Administrator
◦ Logical and Physical Modelling
◦ Data Governance
◦ HADR
 The importance of a play area
 The expanding skillset
◦ Beyond Relational – alternative Databases
◦ Polyglot Database Environment
◦ The Distributed Database and understanding CAP
◦ Alternate architectures - LAMBDA
◦ ETL
◦ Business Intelligence, Data Science, Data Platform Engineer
◦ What else? Audience please….

Types
Structured
Un-structured
Semi-structured
Applications
Fat client, Web
Intranet, Mobile
Storage
Database Type
SQL
NoSQL
NewSQL
Business Intelligence
Standard Reporting from
standard process metrics
from the Data Warehouse/
Reporting database
Business Analytics
Investigative Reporting
over past data.
Management Science
Data Science
Investigative {Data
Analytics, Business
Analytics}
over structured, semi,
unstructured data for
possible patterns – use of
Machine Learning and
Pattern Matching
algorithms.
Data Creators,
Data Contributors,
Data Consumers

Business
Intelligence
SSRS, Crystal,
Business Objects,
PowerPivot, Excel,
QlikView, Tableau,
Reporting apps….
Types
Structured – Normal Form, JSON, XML
Un-structured – {developers think all data is like this }
Semi-structured – JSON, XML, Key/Value Pair
Applications
C#, F#, Java etc.
[Data sourcing]
Storage
Database Type
SQL – Oracle, DB2, Sybase, SQL Server, MySQL etc.
NoSQL – CouchDB, Raven, Cassandra, Hadoop, MongoDB, Neo4j
NewSQL – Postgres-XL, Postgres-XC, Volt-DB, NuoDB
Business
Analytics
SAS, SPSS,
Statistica, MatLab
etc..
Data Science
BI + BA + ‘R’, Pyphon,
Machine Learning
packages, SQL, MapR,
Data Extraction, ML,
Visualisations, Story
Boarding
SQL, MapR, U-SQL..Data Creators,
Data Contributors,
Data Consumers

 SSIS
◦ pull RSS feed and store in SQL Server
◦ ODATA source example
 Azure File Share
◦ Storing archive data

Modelling
Data Governance
HADR
Releasing Stuff

 Data is an Asset – Security Guard
 Data Custodian – Compliance, ???
 Liaison between Business and Devs
 Liaison between Business and Infrastructure
 What else?

 Custodian of the Business Taxonomy
◦ Data Dictionary
 Logical / Physical
◦ Normal Form
◦ Logical Model (relationships) V Physical Model
(vender dependent schema)
 Relational V Dimensional
◦ Entity Relationship modelling (tables and
relationships between)
◦ Dimensional Modelling (facts and dimensions) –
models to usability and performance

 ICO Principals
 Data Protection Laws – Security, Retention
 Your responsibilities – vary within the Org

 High Availability
◦ Understanding Latency
◦ Mirroring
◦ Availability Groups
◦ Log Shipping (?)
 Disaster Recovery
◦ Practiced Procedures
◦ DR Resource misalignment
◦ Implementing contingency
◦ Dealing with Data corruption or Accidents (if I only
have AG’s – what’s the issue?)

 Applying Database releases
◦ Which Databases? SQL / NoSQL etc.
 Supportability (level of reqd knowledge)
 Patching Servers

 You protect the Integrity and Availability of
the “Database Platform”
 Not limited to SQL Server
◦ NoSQL products
◦ Relational “SQL” products
◦ NewSQL

Play Areas
Knowing what to learn

 Align with your company
◦ Talk to developers, see what they are using, take a
lead with Data Technology – nurture their use of
Data.
◦ Data is an Asset, without data your company won’t
exist – make your company realise your importance
and you need to be right up there in the decision
making for technology direction
 Align with the industry
◦ Job boards, trends
 Be one (ok – a couple of) steps ahead!

 You can’t play in live!
 Decent laptop – 16GiB+ RAM, SSD / M2 Flash
 VirtualBox
◦ Multiple Windows Server, build a domain, build a
cluster etc.
◦ Multiple Linux
◦ Etc.

Beyond Relational – alternative
Databases
Polyglot Database Environment
The Distributed Database and
CAP
LAMBDA
ETL
MDM
Cloud

 Business environment is “Polyglot”
 Require understanding of
◦ NoSQL
◦ CAP Theorem
◦ LAMBDA (edge case)
◦ Big Data – what it really is
◦ CEP (is this a Database related tech?)
◦ ETL
◦ Data Science – what it really is
◦ BI
◦ Kimball, Inmon
◦ Data Vault

 Really means – No NF
 Key Value Stores (Riak, CouchDB)
 Column (Cassandra)
 Document (MongoDB)
 Graph (Neo4J)
 Object (Bit niche )
 Ironically – most have a SQL like interface
now or in development!

 Consistency
◦ All nodes show the same value
◦ Eventual Consistency
 Availability
◦ Node will return data
 Partition Tolerance
◦ Islands form when network fails – clients connect to
local nodes so when isolated you lose consistency.
 You can only have two of the 3 and never all
three.

1 2
3 4
5 6
Insert
Update
Delete
DatCtr A
Insert
Update
Delete
DatCtr B
Insert
Update
Delete
DatCtr C

 No – it’s not just Hadoop
 Velocity, Variety, Volume
 BD can be done in anything.
◦ Velocity – CEP, In-Memory, distributed computing
◦ Variety – varied types of data, structured / un.
◦ Volume – size of the data
 BD is not definitive – depends on your
budget, ability etc.

 Processing a data stream in flight
 Window over the stream and determine
trends
 Read the stream rather than poll the database

 If you aren’t using Machine Learning / Data
Mining algol’s you aren’t doing Data Science
 If you know what you are looking for – you
aren’t doing DS.
 DS isn’t just R, you can do DS in numerous
tools, R has a large library of packages to use
against your data
 DS is where you are looking for patterns in
your data and trying to understand them to
then formulate standard process flows to take
advantage.

 Scale out – distributed – data processing
architecture
 Batch, Speed, Service layers
 For low latency, high updates
 Robust

 Kimball
◦ Dimensional modelling with star schema
◦ Dimensions and Facts
◦ Bottom up – data marts to EDW
◦ Aspires to Single Version of the Truth
 Inmon
◦ Normal Form
◦ Can also use star schema
◦ Form the EDW and then use data marts
◦ Stronger approach to Single Version of the Truth

 Modelling method
 Pull all your uncleansed data and store it in
one place
 Buffer between Operational Databases and
the Conformed Data Warehouse

 Are you really on the Cloud or just managed
remotely located server environment?
 Real cloud has immediate elasticity, hides
infrastructure, easy to spawn up new
resource and near immediate.
 Market’d cloud is really managed servers – no
immediate elasticity, servers are provisioned
and that takes time.

 True cloud offers elasticity for Distributed
Database capabilities – proper scale out.
◦ Azure Elastic Database (Sharding)
◦ SQL 2016 Stretch Feature
 Remember CAP? Yep – you need to understand
that.
 On-Prem tends to be scale up, single box –
single database
 Cloud – some of your tasks will disappear
because it’s done for you. But your role is a Data
Centric role and not Infrastructure Centric.

Evolution of the DBA to Data Platform Administrator/Specialist

More Related Content

What's hot (20)

Similar to Evolution of the DBA to Data Platform Administrator/Specialist (20)

Recently uploaded (20)

Evolution of the DBA to Data Platform Administrator/Specialist

Editor's Notes