WEATHER SIMULATION CAN OUTPACE MOORE’S LAW WITH AN AI BOOST

W H I T E PA P E R

THE TRADITIONAL DATABASE

MISMATCH: WHY TODAY’S COMPANIES
NEED INSTANT, CUSTOMER-CENTRIC
INFORMATION ACCESS
New business models require distributed SQL databases which optimize a multicloud
approach says MariaDB
THE TRADITIONAL DATABASE MISMATCH: WHY TODAY ’S COMPANIES NEED INSTANT, CUSTOMER-CENTRIC INFORMATION ACCESS

Business is changing. It is more real time, online and intimate

with the customer. The pace of innovation and the upheavals of
the last few years are driving companies, their competitors and
startups to adopt new business models.

A business must stay up 24/7 and prepare for both including private and hybrid configurations. Xpand supports
sudden peaks in demand as well as slow days. While doing a mix of both NoSQL and relational data and enables real-
this, it has to be able to adjust to changes in customer time operational analytics on transactional workloads. By
demand, market conditions, environmental disasters and using distributed SQL technologies, a company can do what
its supply chain. Adapting to these new demands and new is necessary to support new business models and market
business models requires up-to-date, accurate data, even requirements: scaling systems of record to the cloud. And by
when bad things happen. choosing Xpand specifically, companies can handle a mix of
workloads and real-time operational analytics.

This paper outlines the overall mismatch of existing

databases to new business models and requirements,
Distributed SQL databases enable
the suitability of distributed SQL databases, and the
companies to deploy new business specific advantages of MariaDB Xpand. It also outlines
models and meet the requirements situations for which distributed SQL databases are not
suitable, and related issues to consider when selecting a
of always-on service and dramatic
database or database provider. The paper provides a high-
fluctuations in traffic. level business purpose analysis with enough technical detail
to understand the purpose of distributed SQL and Xpand’s
technological benefits.

Neither the data technology of yesteryear, which was flexible

and an adequate system of record, nor the data technology WHY NEW BUSINESS MODELS DO NOT
of the recent past, which scaled well but was built for only WORK ON TRADITIONAL DATABASES
one purpose, will support these new models or the rate of
change. As a business adopts cloud computing technologies, Company executives know that a business must adapt or
it requires database technology that makes full use of the perish. Adaptation requires looking at the modern world
benefits of public, private and hybrid clouds. Meanwhile, and asking what customers want and how to provide it to
to support global operations and maintain a competitive them quicker, faster and more reliably. The answer to these
advantage, companies are adopting a multicloud strategy. questions is the foundation of how both existing and startup
To support these objectives and more, they need a database companies are transforming their industries. New business
that operates on multiple clouds, is as robust as a relational models rely on a few key principles: always-on, customer
database, as scalable as a NoSQL database, and provides engagement, customer intimacy and instant gratification.
availability for when things go wrong.
Always-on, from a data infrastructure perspective, means
Distributed SQL databases provide extreme levels of scale having fast, reliable systems near the customer and multiple
and resilience against multiple failures. MariaDB Xpand is redundancies. Keeping a service running often means using
a distributed SQL database that runs on multiple clouds, a public cloud or a combination of private infrastructure and

WHITEPAPER 2
THE TRADITIONAL DATABASE MISMATCH: WHY TODAY ’S COMPANIES NEED INSTANT, CUSTOMER-CENTRIC INFORMATION ACCESS

public infrastructure. For data infrastructure that means business relies upon must provide ACID transactions and
writing data to multiple “availability zones” in the cloud and they must be written synchronously. While this is not new, it
even replicating data across multiple regions. When things is something new to data infrastructure running at scale.
go wrong, as they inevitably do, the infrastructure should
automatically adapt.
MOVING TO THE CLOUD REQUIRES
Customer engagement is a two-way street. Customers look NEW THINKING
at more data before making a purchase, this holds true for
both business-to-consumer (B2C) and business-to-business Business leaders are driving cloud computing initiatives in
(B2B) customers. Meanwhile, companies communicate with order to provide greater reliability, scale and move capital
their customers more frequently. Except for the smallest expenses to operational expenses. However, simply installing
volume businesses, this is entirely data- and process-driven. something in the cloud provides none of those advantages.
For data infrastructure, this means being able to handle large While Amazon Web Services (AWS) and Google Cloud
amounts of traffic. At the database level it means read and Platform (GCP) are very reliable in developed countries, in a
write scale. global environment cloud service providers fail. Even if AWS
and GCP stay up, service failures in regions, availability zones
Customer intimacy is required for effective engagement. and other faults can disrupt critical business services. These
As customers engage with a company and its products, the kinds of outages are in the news regularly, especially during
company profiles their purchases, likes, dislikes, searches, peak times such as holiday seasons.
what they found, did not find, and what offers they accepted
or declined. This data must be captured in real time from Moving capex to opex means that instead of purchasing
clickstreams, purchase data, support centers and anywhere a bunch of infrastructure for peak load (i.e., Black Friday
else the customer is involved. Everything from the front page traffic), capacity is purchased as it is needed. This, however,
of the website to the sales call is personalized for one person means that there is a strong correlation between the cost
and their likes, wants and needs. For database infrastructure, of a company’s infrastructure and the performance of the
this means being able to handle large amounts of data. software running on it. Some software uses more resources
to perform the same number of operations. Purchasing
Instant gratification or real time has become a reality for decisions need to be made not just on infrastructure cost but
most consumer-facing companies and is increasingly a part how many operations the software services and performs on
of B2B firms’ business models. Whether it is made-to-order that infrastructure.
manufacturing, just-in-time inventory, or that the customer
can click on a pair of pants they want and possibly have them
delivered to their home on the same day – no one wants to Multicloud is not a “nice to have,”
wait. Businesses that have what their customers want and
it is the only strategy many companies
deliver it immediately thrive and those which have long
wait times and difficult processes are disrupted. Instant can pursue.
gratification or real-time business is the absence of latency.
At the database level, this means operations run in single
digit milliseconds (ms), or in the worst case under 60ms, for In the cloud, businesses are charged for ingress and egress.
both reads and writes. This means they are charged for not just storage and CPU
but every byte pushed in and out of the cloud. If more
Data integrity is still an absolute necessity for a system of services are used across the cloud talking to each other, each
record. When an answer is needed it must be authoritative byte pushed costs money. To control this, business leaders
no matter when it is read. This means that the database a must insist on a simpler architecture designed for more
general use.

WHITEPAPER 3
THE TRADITIONAL DATABASE MISMATCH: WHY TODAY ’S COMPANIES NEED INSTANT, CUSTOMER-CENTRIC INFORMATION ACCESS

New business models have sharper peaks and valleys.

Companies should be able to maximize their capacity to meet
In order to reduce complexity from
demand in terms of both purchases and engagement, but
also shrink their capacity on a normal or slower business day.
a data infrastructure perspective,
This means data infrastructure that allows for immediate businesses must deploy fewer general-
horizontal scale-out, but also allows the company to scale purpose databases.
back when it is no longer needed.

MULTICLOUD IS THE ONLY

Finally, businesses have learned that when one vendor gets
VIABLE STRATEGY
too much power, they raise the price. This happens at a
market level, but also at a company-to-company level. The
Multicloud is not a “nice to have,” it is the only strategy many
Oracle salesperson of yesteryear knew a customer could
companies can pursue. Anything can fail, including an entire
not migrate off their software easily so they could negotiate
cloud service provider (CSP) in their most important region.
higher prices. There is no reason to believe tying a company
While CSPs are fairly reliable in the northern part of the
to a single cloud provider will have any different outcome.
western hemisphere, global businesses must think past that.
While list price always seems to be coming down, larger firms
It is prudent to maintain the ability to run on multiple cloud
often can negotiate a better volume price or level of service.
providers for both availability and disaster recovery reasons.
Maintaining the freedom to exist is key to these negotiations.

Acquisitions of other companies will mean acquiring services

Ultimately, many companies will pursue a multicloud strategy
that run on different clouds. Integration is faster when both
either deliberately, out of necessity or by accident. It is
companies have maintained the ability to run on different
best to pick software and services that support multiple
cloud providers. This means foregoing higher level services
clouds and are independent of any particular cloud provider.
from hyperscalers such as messaging, database services
A deliberate multicloud strategy ensures the company
and other services for which there is a standard or cloud
can easily adapt to globalization, outages, regulatory
provider-independent alternative.
environments, acquisitions and the market in general.

Buying software or software services often means running

in multiple clouds. Some SaaS services do not run on every
THE BENEFITS OF A SIMPLIFIED
cloud. Moreover, some necessary third-party enterprise
ARCHITECTURE
software requires specific databases or infrastructure.

Complexity is not just a pain, it is a cost, risk and productivity

Globalization requires a multicloud strategy. Some US
tax. Every byte in and out of a virtual private cloud or region
and European cloud providers do not even operate in the
is charged as “ingress” and “egress.” Every database deployed
Asia-Pacific region. Moreover, some cloud providers have
requires staff to learn and maintain it. Complexity is where
fewer resources and levels of service in various regions of
outages live, due to the linkages and pieces necessary to
the world. Various national and regional laws around data
keep each piece of software running and communicating
sovereignty may require running data services in-country
with each other. Developers who have to code to multiple
where a cloud provider does not serve.

WHITEPAPER 4
THE TRADITIONAL DATABASE MISMATCH: WHY TODAY ’S COMPANIES NEED INSTANT, CUSTOMER-CENTRIC INFORMATION ACCESS

databases and learn their particulars and how to integrate nodes and maintain redundant copies of data to ensure that
them will spend less time working on actual business logic when a node or entire availability zone goes down, the data
and applications. Finally, data is only consistent within is not lost and operations seamlessly continue. They provide
a transactional boundary – that generally means a the same level of consistency that traditional relational
single database. databases are known for.

In order to reduce complexity from a data infrastructure Elasticity in the architecture means that nodes can be
perspective, businesses must deploy fewer general-purpose added and removed without data loss. This allows businesses
databases. The so-called polyglot persistence strategy of the to enable peak performance on heavy business days but scale
past using various “made for purpose” single-point solutions back capacity when traffic is lower. This gives a company
is untenable for future applications. Databases must the right performance for the right price while saving on
continue to scale and provide reliability in the cloud, but also cloud costs.
recover the full functionality of the relational databases.
Not all distributed SQL databases are alike. They differ in
performance characteristics, compatibility with existing
HOW DISTRIBUTED SQL SOLVES datastores and other capabilities. Some distributed SQL
PROBLEMS databases are better at autobalancing data when there
are “hotspots’’ or nodes that get used more than others.
Businesses need to handle larger datasets, with more read Moreover, having a distributed SQL database may not be
and write throughput, at low latencies, while staying up 24/7. all of the infrastructure needed. For instance, what happens
They require the simplicity of a standard query language during failures? Does the client have to handle existing
and the ability to handle both relational and NoSQL (JSON) transactions? For these issues, other software
data. Distributed SQL databases spread data across multiple and infrastructure may be needed.

Right-size capacity

WHITEPAPER 5
THE TRADITIONAL DATABASE MISMATCH: WHY TODAY ’S COMPANIES NEED INSTANT, CUSTOMER-CENTRIC INFORMATION ACCESS

WHEN DON’T YOU NEED MARIADB XPAND FOR PERFORMANCE,

DISTRIBUTED SQL COST, MULTICLOUD AND FLEXIBILITY

Not all applications or datasets need distributed SQL. Xpand is MariaDB’s distributed SQL database. It is available
Consider a client-server database in the cloud (such as with both on-premises and cloud deployments of MariaDB.
MariaDB Enterprise Server) if: Xpand not only provides the elastic read-write scale,
absolute availability and support for multiple clouds but
also real-time, ad hoc and operational analytics. Finally, in a
recent benchmark, MariaDB Xpand on SkySQL dramatically
outscaled CockroachDB’s comparable offering using
· The size of the database fits cleanly on one server.
the same configuration, indicating it has a better price/
performance value as well.
· Write scale is not needed and read scale is
sufficient.
The flexibility to handle multiple types of models and
workloads with the best price/performance is MariaDB
· Failover to a redundant node is sufficient and
Xpand’s strong point. Xpand’s columnar index capability
multiple availability zone failure is not a major
gives it superior support for real-time, ad hoc, personal and
concern.
operational analytics. Developers can just query without
having to create the perfect index. This gives the business
· Elastic scale is not needed and traffic is
greater flexibility and does not require a separate database
predictable.
for intraday analytics such as “how many did we sell today?”
It also enables new use cases such as real-time pricing
· An existing application depends on the
adjustment or various Internet of Things (IoT) read/write
performance characteristics (lower latency) of a
with real-time analytics. In many ways, Xpand is the first fully
traditional client-server database.
functional, general use distributed database in the cloud.

Cost comparison to CockroachDB

WHITEPAPER 6
THE TRADITIONAL DATABASE MISMATCH: WHY TODAY ’S COMPANIES NEED INSTANT, CUSTOMER-CENTRIC INFORMATION ACCESS

When using Xpand, additional infrastructure is deployed performing poorly? MariaDB offers SkyDBAs as an
(MariaDB MaxScale) which automatically handles errors and always-on human-provided service to augment the
replays in-flight transactions on other nodes. This capability capabilities of development and system
ensures that developers do not have to write complicated administration teams.
error-handling code and that node or zone failures are
handled correctly.
HYBRID/EDGE DATABASE MODELS

SUPPORTED SELF-SERVICE There are use cases that do not perform well in the cloud
such as manufacturing systems. There are also highly
Aside from moving capex to opex, a benefit of the cloud regulated industries that are not permitted to store various
is that teams can just provision what they need, where data on third-party servers. Many use cases benefit from
they need it and when they need it while cutting down on being able to process some data at the “edge” of the network
operational costs. This is the so-called “self-service” nature or in a private nearby datacenter while offloading some
of modern IT. However, “self-service” has its limitations. processing to public clouds. Secondly, some businesses use
Most cloud service providers do not offer the kind of help the public cloud as a disaster recovery site rather than for
that a database administrator usually performs. Did the day-to-day operations. For these reasons and more, MariaDB
team create a schema that will perform well? Are backups Enterprise is available both as a standalone product and in
scheduled at the most optimal time? Why is this one query combination with SkySQL.

Try Xpand on SkySQL

Learn more about MariaDB’s SkyDBA
services here.

Teams that want to get started with cloud-scale

infrastructure that stays up and is flexible enough
for modern everyday applications can try Xpand
on SkySQL here. MariaDB offers tutorials,
community forums and a Slack channel to help
them get started.

WHITEPAPER 7