Data Modeling for Data Science

Overview of Data Modeling

Lecture 01

Henry Novianus Palit

[email protected]
 Data Modeling (1)
Data modeling is the process of producing a
descriptive diagram of relationships between various
types of information that are to be stored
The model can be thought of as a way of translating
the logic of accurately describing things in the real-
world and the relationships between them into rules
that can be followed and enforced by computer code
One of the goals of data modeling is to create the
most efficient method of storing information while
still providing for complete access and reporting

Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 2
 Data Modeling (2)
Data modeling is a crucial skill for every data scientist,
whether you are doing research design or architecting
a new data store for your company
Data modeling for data science requires the ability to
think clearly and systematically about
the key data points to be stored and retrieved, and
how they should be grouped and related
Data modelling is sometimes as much art as science
 Following the rules of normalization can be straightforward,
but knowing when to break them and what data to optimize
for later access takes perception beyond simply applying rules
Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 3
 Data Modeling (3)
Stages of Data Modeling (i.e., creating schemas):
Conceptual – imposes a theoretical order on data as it
exists in relationship to the entities being described, of the
real-world artifacts or concepts
Logical – imposes order by establishing discrete entities,
key values, and relationships in a logical structure
Physical – breaks the data down into the actual tables,
clusters, and indexes required for the data store
Visual representations of data models: entity-
relationship model/diagram, Bachman diagram,
object-role modeling, Zachman framework
Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 4
 RDBMS
The RDBMS store paradigm relies on the database
system to maintain internal consistency and
coherence of the data being held in it
Very large datasets have thrown something of a
wrench into the dominance of RDBMS, whether the
data being stored can easily be modeled relationally
or not
When millions or trillions of data points are being
stored, the price of this internal consistency can
bring performance grinding to a halt

Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 5
 NoSQL (1)
NoSQL databases such as MongoDB, Cassandra, and
HBase have been one of the most promising industry
answers to this problem
These use sometimes radically denormalized data stores
with the sole objective to improve performance
They rely on calling code and queries to handle the sort of
consistency, integrity, and concurrency, offering blinding
speed and scalability over ease-of-use
They adopt simplistic data stores such as:
Key-value stores
Document stores
Graphs
Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 6
 NoSQL (2)
Modeling these types of stores is a significant
departure from the RDBMS method
Data scientists may start from the result side of the
process, asking themselves, “What question am I trying to
answer?” instead of “What data do I need to store?”
They will completely disregard duplication of data and
have to plan to handle concurrency conflicts and other
integrity issues on the output end rather than in the design
itself
They might choose to aggregate data rather than breaking
it down discretely

Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 7
 NoSQL (3)
NoSQL data modeling uses advanced techniques:
atomic updates, dimensionality reduction, inverted
search patterns, tree aggregation
Understanding these techniques, and the capabilities
offered by NoSQL, allow data scientists to make the
best choice for what type of data store to use and
how to model it
In almost every case, data scientists in the real world
will end up using a combination of RDBMS and
NoSQL or other exotic data sources as a daily part of
their work
Source: https://www.datasciencegraduateprograms.com/data-modeling/
Data Modeling for Data Science 8
Data Modeling for Data Science
Foundations of Data Systems
Lecture 01

Henry Novianus Palit

[email protected]
 Data-Intensive Applications (1)
In the last decade we have seen many interesting
developments in databases, in distributed systems,
and in the ways we build applications on top of them
Various driving forces for these developments:
Internet companies such as Google, Yahoo!, Amazon, Facebook,
LinkedIn, Microsoft, and Twitter are handling huge volumes of data
and traffic, forcing them to create new tools that enable them to
efficiently handle such scale
Businesses need to be agile, test hypotheses cheaply, and respond
quickly to new market insights by keeping development cycles short
and data models flexible
Free and open source software has become very successful and is now
preferred to commercial or bespoke in-house software in many
environments
Data Modeling for Data Science 10
 Data-Intensive Applications (2)
Various driving forces … : (cont’d)
CPU clock speeds are barely increasing, but multi-core processors are
standard, and networks are getting faster; this means parallelism is only
going to increase
You can now build systems that are distributed across many machines
and even multiple geographic regions, thanks to infrastructure as a
service (IaaS) such as Amazon Web Services (AWS)
Many services are now expected to be highly available; extended
downtime due to outages or maintenance is becoming increasingly
unacceptable
An application is data-intensive if data is its primary
challenge – the quantity, the complexity, or the speed
at which it is changing – as opposed to compute-
intensive, where CPU cycles are the bottleneck
Data Modeling for Data Science 11
 Data-Intensive Applications (3)
A data-intensive application is typically built from
standard building blocks that provide commonly
needed functionality, e.g., many apps need to:
Store data so that they, or another app, can find it again
later (databases)
Remember the result of an expensive operation, to speed
up reads (caches)
Allow users to search data by keyword or filter it in various
ways (search indexes)
Send a message to another process, to be handled
asynchronously (stream processing)
Periodically crunch a large amount of accumulated data
(batch processing)
Data Modeling for Data Science 12
 Data-Intensive Applications (4)
These data systems are such a successful abstraction:
we use them all the time without thinking too much
But reality is not that simple:
There are many database systems with different
characteristics, because different apps have different
requirements
There are various approaches to caching, several ways of
building search indexes, and so on
When building an app, we still need to figure out which
tools and which approaches are the most appropriate for
the task at hand; and it can be hard to combine tools
when you need to do something that a single tool cannot
do alone
Data Modeling for Data Science 13
 Data Systems (1)
We typically think of databases, queues, caches, etc.
as being very different categories of tools; so why
should we lump them all together under an umbrella
term like data systems?
Many new tools for data storage and processing have
emerged in recent years; they are optimized for a
variety of different use cases and no longer neatly fit
into traditional categories
There are datastores that are also used as message queues
(Redis)
There are message queues with database-like durability
guarantees (Apache Kafka)
Data Modeling for Data Science 14
 Data Systems (2)
Increasingly many apps now have such demanding or
wide-ranging requirements that a single tool can no
longer meet all of its data processing & storage needs
Instead, the work is broken down into tasks that can
be performed efficiently on a single tool, and those
different tools are stitched together using app code
E.g., if you have an application-managed caching layer
(using Memcached or similar), or a full-text search server
(such as Elasticsearch or Solr) separate from your main
database, it is normally the app code’s responsibility to keep
those caches and indexes in sync with the main database

Data Modeling for Data Science 15

Data Systems (3)

Data Modeling for Data Science 16

 Data Systems (4)
When you combine several tools in order to provide
a service, the service’s interface or API usually hides
those implementation details from clients; now you
have essentially created a new, special-purpose data
system from smaller, general-purpose components
If you are designed a data system or service, a lot of
tricky questions arise:
How do you ensure that the data remains correct and complete, even
when things go wrong internally?
How do you provide consistently good performance to clients, even
when parts of your system are degraded?
How do you scale to handle an increase in load?
What does a good API for the service look like?
Data Modeling for Data Science 17
 Data Systems (5)
Three concerns that are important in most software
systems:
Reliability – the system should continue to work correctly
(performing the correct function at the desired level of
performance) even in the face of adversity (hardware or
software faults, and even human error)
Scalability – as the system grows (in data volume, traffic
volume, or complexity), there should be reasonable ways
of dealing with that growth
Maintainability – over time, many different people will
work on the system (engineering and operations, both
maintaining current behavior and adapting the system to
new use cases), and they should all be able to work on it
productively
Data Modeling for Data Science 18
 Reliability (1)
Typical expectations of reliable software:
The application performs the function that user expected
It can tolerate the user making mistakes or using the
software in unexpected ways
Its performance is good enough for the required use case,
under the expected load and data volume
The system prevents any unauthorized access and abuse
So, reliability means roughly “continuing to work
correctly, even when things go wrong”
Things that can go wrong are called faults, and
systems that anticipate faults and can cope with
them are called fault-tolerant or resilient
Data Modeling for Data Science 19
 Reliability (2)
Note: a fault is not the same as a failure
A fault is usually when one component of the system
deviates from its spec
A failure is when the system as a whole stops providing the
required service to the user
It is impossible to reduce the probability of a fault to
zero; thus, it is usually best to design fault-tolerance
mechanisms that prevent faults from causing failures
Although we generally prefer tolerating faults over
preventing faults, there are cases where prevention
is better than cure (e.g., because no cure exists); for
example: security breaches cannot be undone
Data Modeling for Data Science 20
 Hardware Faults (1)
Examples: hard disks crash, RAM becomes faulty, the
power grid has a blackout, someone unplugs the
wrong network cable
Hard disks are reported as having a mean time to failure (MTTF) of
about 10–50 years; thus, on a storage cluster with 10,000 disks, we
should expect on average one disk to die per day
Our first response is usually to add redundancy to
the individual hardware components in order to
reduce the failure rate of the system
Disks may be set up in a RAID configuration
Servers may have dual power supplies and hot-swappable CPUs
Datacenters may have batteries and diesel generators for backup
power
Data Modeling for Data Science 21
 Hardware Faults (2)
As data volumes and apps’ computing demands have
increased, more apps have begun using larger
number of machines, which proportionally increase
the rate of hardware faults
Hence, there is a move toward systems that can
tolerate the loss of entire machines, by using
software fault-tolerance techniques in preference or
in addition to hardware redundancy
The advantage: a single-server system requires planned downtime if
you need to reboot the machine (e.g., to apply OS security patches),
whereas a system that can tolerate machine failure can be patched
one node at a time (i.e., rolling upgrade), without downtime of the
entire system
Data Modeling for Data Science 22
 Software Errors (1)
Another class of fault is a systematic error within the
system; such faults are harder to anticipate, and
because they are correlated across nodes, they tend
to cause many more system failures than uncorrelated
hardware faults
A software bug that causes every instance of an app server to crash
when given a particular bad input; e.g., the leap second on Jun 30,
2012, that caused many apps to hang simultaneously due to a bug in
the Linux kernel
A runway process that uses up some shared resources – CPU time,
memory, disk space, or network bandwidth
A service that the system depends on that slows down, becomes
unresponsive, or starts returning corrupted responses
Cascading failures, where a small fault in one component triggers a fault
in another component, which in turn triggers further faults
Data Modeling for Data Science 23
 Software Errors (2)
The bugs that cause these kinds of software faults
often lie dormant for a long time until they are
triggered by an unusual set of circumstances
There is no quick solution to the systematic faults in
software, but lots of small things can help:
Carefully thinking about assumptions and interactions in the system
Thorough testing
Process isolation
Allowing processes to crash and restart
Measuring, monitoring, and analyzing system behavior in production

Data Modeling for Data Science 24

 Human Errors (1)
Even when they have the best intentions, humans are
known to be unreliable; e.g., a study of large internet
services found that configuration errors by operators
were the leading cause of outages, whereas hardware
faults played a role in only 10–25% of outages
The best systems combine several approaches:
Design systems in a way that minimizes opportunities for errors, e.g.,
well-designed abstractions, APIs, and admin interfaces make it easy to
do “the right thing” and discourage “the wrong thing” (However, if the
interfaces are too restrictive, people will work around them, negating
their benefit, so this is a tricky balance to get right)
Decouple the places where people make the most mistakes from the
places where they can cause failures; in particular, provide fully featured
non-production sandbox environments where people can explore and
experiment safety, using real data, without affecting real users
Data Modeling for Data Science 25
 Human Errors (2)
The best systems combine … : (cont’d)
Test thoroughly at all levels from unit tests to whole-system integration
tests and manual tests; automated testing is widely used, well
understood, and especially valuable for covering corner cases that rarely
arise in normal operation
Allow quick and easy recovery from human errors, to minimize the
impact in the case of a failure; e.g., make it fast to roll back config
changes, roll out new code gradually (so that any unexpected bugs
affect only a small subset of users), and provide tools to recompute data
(in case it turns out that the old computation was incorrect)
Set up detailed and clear monitoring, such as performance metrics and
error rates (referred to as telemetry in other engineering disciplines);
monitoring can show us early warning signals and allow us to check
whether any assumptions or constraints are being violated, while
metrics can be invaluable in diagnosing issue when a problem occurs
Implement good management practices and training – a complex and
important aspect
Data Modeling for Data Science 26
 Scalability
Even if a system is working reliably today, that does
not mean it will necessarily work reliably in the
future; one common reason for degradation is
increased load
The system has grown from 10,000 concurrent users to 100,000 or
even to 1 million concurrent users, or
It is processing much larger volumes of data than it did before
Scalability is the term to describe a system’s ability to
cope with increase load; discussing scalability means
considering questions like
If the system grows in a particular way, what are our options for coping
with the growth?
How can we add computing resources to handle the additional load?
Data Modeling for Data Science 27
 Describing Load (1)
We need to succinctly describe the current load on the
system, only then can we discuss growth questions
Load can be described with a few numbers which we
call load parameters; the best choice of parameters
depends on the architecture of your system
Request per second to a web server
The ratio of reads to writes in a database
The number of simultaneously active users in a chat room
The hit rate on a cache
Perhaps the average case is what matters for you, or
perhaps your bottleneck is dominated by a small
number of extreme cases
Data Modeling for Data Science 28
 Describing Load (2)
Let’s consider Twitter as an example, using data
published in November 2012
Two of Twitter’s main operations are:
Post tweet – a user can publish a new message to their followers
(4.6k requests/sec on average, over 12k requests/sec at peak)
Home timeline – a user can view tweets posted by the people they
follow (300k requests/sec)
Simply handling 12,000 writes per second would be
fairly easy; however, Twitter’s scaling challenge is
not primarily due to tweet volume, but due to fan-
out (i.e., each user follows many people, and each
user is followed by many people)
Data Modeling for Data Science 29
 Describing Load (3)
Two ways of implementing these 2 operations:
Posting a tweet simply inserts the new tweet into a global
collection of tweets. When a user requests their home timeline,
look up all the people they follow, find all the tweets for each of
those users, and merge them (sorted by time).
SELECT tweets.*, users.* FROM tweets
JOIN users ON tweets.sender_id = users.id
JOIN follows ON follows.followee_id = users.id
WHERE follows.follower_id = current_user
Maintain a cache for each user’s home timeline – like a mailbox
of tweets for each recipient user. When a user posts a tweet,
look up all the people who follow that user, and insert the new
tweet into each of their home timeline caches. The request to
read the home timeline is then cheap, because its result has
been computed ahead of time.
Data Modeling for Data Science 30
 Describing Load (4)

The first version of Twitter struggled to keep up with the load of home
timeline queries, so the company switched to the second version.

Data Modeling for Data Science 31

 Describing Load (5)

The second version of Twitter works better because the average rate of
published tweets is almost two orders of magnitude lower than the rate of
home timeline reads, and so in this case it’s preferable to do more work at
write time and less at read time. The downside is that posting a tweet now
requires a lot of extra work. Some users have > 30M followers. Doing this
in a timely manner – in 5 secs (Twitter’s target) – is a significant challenge.
Data Modeling for Data Science 32
 Describing Load (6)
In the example of Twitter, the distribution of followers
per user (maybe weighted by how often those users
tweet) is a key load parameter for discussing
scalability, since it determines the fan-out load
The final twist: Twitter is moving to a hybrid of both
approaches
Most users’ tweets continue to be fanned out to home
timelines at the time when they are posted
A small number of users with a very large number of
followers (i.e., celebrities) are excepted; tweets from any
celebrities that a user may follow are fetched separately
and merged with that user’s home timeline when it is read
Data Modeling for Data Science 33
 Describing Performance (1)
Once you have described the load on your system, you
can investigate what happens when the load increases:
When you increase a load parameter and keep the system
resources (CPU, memory, network bandwidth, etc.)
unchanged, how is the performance of your system affected?
When you increase a load parameter, how much do you need
to increase the resources if you want to keep performance
unchanged?
In a batch processing system such as Hadoop, we
usually care about throughput – the number of records
we can process per second, or the total time it takes to
run a job on a dataset of a certain size
Data Modeling for Data Science 34
 Describing Performance (2)
In online systems, what’s usually more important is
the service’s response time – the time between a
client sending a request and receiving a response
In practice, in a system handling a variety of requests,
the response time can vary a lot (we need to think of
it as a distribution of values that you can measure)
Most requests are reasonably fast, but there are occasional
outliers that take much longer
In order to figure out how bad your outliers are, you can
look at higher percentiles: 95th, 99th, and 99.9th percentiles
are common (abbreviated p95, p99, and p999), e.g., if the
95th percentile response time is 1.5 secs, that means 5 out
of 100 requests take 1.5 secs or more
Data Modeling for Data Science 35
 Describing Performance (3)
In practice, … (cont’d)
High percentiles of response times, a.k.a. tail latencies, are
important because they directly affect users’ experience of
the service, e.g., the customers with the slowest requests
are often the most valuable customers
On the other hand, reducing response times at very high
percentiles (e.g., 99.99th) is difficult because they are easily
affected by random events outside of your control, and the
benefits are diminishing
Percentiles are often used in service level objectives
(SLOs) and service level agreements (SLAs), contracts
that define the expected performance and availability
of a service
Data Modeling for Data Science 36
 Describing Performance (4)
For example, an SLA may state:
The service is considered to be up if it has a median
response time < 200 ms and a 99th percentile < 1 s
It may be required to be up at least 99.9% of the time
 These metrics set expectations for the clients of the service
and allow customers to demand a refund if the SLA isn’t met
Queueing delays often account for a large part of the
response time at high percentiles
As a server can only process a small number of things in
parallel, it only takes a small number of slow requests to
hold up the processing of subsequent requests (i.e., head-of-
line blocking)
Data Modeling for Data Science 37
 Describing Performance (5)
Queueing delays often … (cont’d)
Even if those subsequent requests are fast to process on
the server, the client will see a slow overall response time
due to the time waiting for the prior request to complete
When generating load artificially to test the scalability of a
system, the load-generating client needs to keep sending
requests independently of the response time, and doesn’t
wait for the previous request to complete before sending
the next one (i.e., artificially causing the queues short in
the test, which skews the measurements)

Data Modeling for Data Science 38

Approaches for Coping with Load (1)
An architecture that is appropriate for one level of
load is unlikely to cope with 10 times that load; on a
fast-growing service, it is therefore likely that you will
need to rethink your architecture on every order of
magnitude load increase (or perhaps even more)
People often talk of a dichotomy between scaling up
(vertical scaling, moving to a more powerful machine)
and scaling out (horizontal scaling, distributing the
load across multiple smaller machines, a.k.a. shared-
nothing architecture)
In reality, good architectures usually involve a pragmatic mixture of
approaches: using several fairly powerful machines can be simpler and
cheaper than a large number of small virtual machines
Data Modeling for Data Science 39
Approaches for Coping with Load (2)
Some systems are elastic (i.e., they can automatically
add computing resources when they detect a load
increase), whereas other systems are scaled
manually (i.e., a human analyzes the capacity and
decides to add more machines to the system)
Taking stateful data systems from a single node to a
distributed setup can introduce a lot of additional
complexity; common wisdom until recently was to
keep your database on a single node (scale up) until
scaling cost or high-availability requirements forced
you to make it distributed

Data Modeling for Data Science 40

Approaches for Coping with Load (3)
The architecture of systems that operate at large scale
is usually highly specific to the app – no such thing as
a generic, one-size-fits-all scalable architecture (a.k.a.
magic scaling sauce)
The problem may be the volume of reads, the volume of writes, the
volume of data to store, the complexity of the data, the response time
requirements, the access patterns, or (usually) some mixture of all of
these plus many more issues
An architecture that scales well for a particular app is
built around assumptions of which operations will be
common and which will be rare – the load parameters
Scalable architectures are usually built from general-
purpose building blocks, arranged in familiar patterns
Data Modeling for Data Science 41
 Maintainability (1)
It is well known that the majority of the cost of
software is not in its initial development, but in its
ongoing maintenance – fixing bugs, keeping its
systems operational, investigating failures, adapting
it to new platforms, modifying it for new use cases,
repaying technical debt, and adding new features
Yet, many people working on software systems
dislike maintenance of so-called legacy systems –
perhaps it involves fixing other people’s mistakes, or
working with platforms that are now outdated, or
systems that were forced to do things they were
never intended for
Data Modeling for Data Science 42
 Maintainability (2)
We can and should design software in such a way that
it will hopefully minimize pain during maintenance,
and thus avoid creating legacy software ourselves
We will pay particular attention to three design
principles for software systems:
Operability – make it easy for operations teams to keep the system
running smoothly
Simplicity – make it easy for new engineers to understand the system,
by removing as much complexity as possible from the system
Evolvability – make it easy for engineers to make changes to the
system in the future, adapting it for unanticipated use cases as
requirements change (a.k.a. extensibility, modifiability, or plasticity)

Data Modeling for Data Science 43

Operability: Making Life Easy for Operations (1)
Good operations can often work around the
limitations of bad (or incomplete) software, but good
software cannot run reliably with bad operations
While some aspects of operations can and should be
automated, it is still up to humans to set up that
automation in the first place and to make sure it’s working
correctly
Operations teams are vital to keeping a software
system running smoothly
Good operability means making routine tasks easy,
allowing the operations team to focus their effort on
high-value activities
Data Modeling for Data Science 44
Operability: Making Life Easy for Operations (2)
Data systems can do various things to make routine
tasks easy, including:
Providing visibility into the runtime behavior and internals of the
system, with good monitoring
Providing good support for automation and integration with
standard tools
Avoiding dependency on individual machines (allowing
machines to be taken down for maintenance while the system
as a whole continues running uninterrupted)
Providing good documentation and an easy-to-understand
operational model
Self-healing where appropriate, but also giving administrators
manual control over the system state when needed
Exhibiting predictable behavior, minimizing surprises
Data Modeling for Data Science 45
Simplicity: Managing Complexity (1)
Small software projects can have delightfully simple
and expressive code, but as projects get larger, they
often become very complex and difficult to understand
This complexity slows down everyone who needs to work on
the system, further increasing the cost of maintenance
A software project mired in complexity is sometimes
described as a big ball of mud
Various possible symptoms of complexity: explosion of
the state space, tight coupling of modules, tangled
dependencies, inconsistent naming and terminology,
hacks aimed at solving performance problems, special-
casing to work around issues elsewhere, etc.
Data Modeling for Data Science 46
Simplicity: Managing Complexity (2)
When complexity makes maintenance hard, budgets
and schedules are often overrun
There is also a greater risk of introducing bugs when
making a change in complex software: when the
system is harder for developers to understand and
reason about, hidden assumptions, unintended
consequences, and unexpected interactions are
more easily overlooked
Conversely, reducing complexity greatly improves the
maintainability of software, and thus simplicity
should be a key goal for the system we build
Data Modeling for Data Science 47
Simplicity: Managing Complexity (3)
Making a system simpler does not necessarily mean
reducing its functionality; it can also mean removing
accidental complexity (i.e., not inherent in the
problems that the software solves but arises only
from the implementation)
One of the best tools for removing accidental
complexity is abstraction
A good abstraction can hide a great deal of implementation
detail behind a clean, simple-to-understand façade
A good abstraction can also be used for a wide range of
different apps (not only is this reuse more efficient, but it
also leads to higher-quality software)
Data Modeling for Data Science 48
Evolvability: Making Change Easy (1)
It’s extremely unlikely that your system’s
requirements will remain unchanged forever
They are much more likely to be in constant flux: you learn new
facts, previously unanticipated use cases emerge, business
priorities change, users request new features, new platforms
replace old platforms, legal or regulatory requirements change,
growth of the system forces architectural changes, etc.
Agile working patterns provide a framework for
adapting to change
The Agile community has developed technical tools and patterns
that are helpful when developing software in a frequently
changing environment, such as test-driven development (TDD)
and refactoring
Data Modeling for Data Science 49
Evolvability: Making Change Easy (2)
We search for ways of increasing agility on the level of
a larger data system, perhaps consisting of several
different apps or services with different characteristics
The ease with which you can modify a data system,
and adapt it to changing requirements, is closely
linked to its simplicity and its abstractions: simple and
easy-to-understand systems are usually easier to
modify than complex ones
However, since this is such an important idea, we will
use a different word to refer to agility on a data
system level: evolvability
Data Modeling for Data Science 50