1. EXECUTIVE SUMMARY
TODAY’S BUSINESS MODELS
DEPEND ON DATA
OLDER DATABASES CAN’T KEEP UP
BUILDING THE FRAMEWORK
FOR DATA ON DEMAND
A MODERN CLOUD APPLICATION PLATFORM
CASE STUDIES: WHO NEEDS SPEED?
INSIDE:
VMware’s in-memory, distributed vFabric SQLFire
delivers unprecedented scalability, performance
Delivering data at the
speed of business
vFABRIC
SQLFire
IS NOW
vFABRIC
SQLFire
IS NOW
2. EXECUTIVE SUMMARY
JUST AS VIRTUALIZATION has driven huge improve-
ments in enterprise infrastructures during the past
few years, the application layer that sits above
these servers, storage and network components is
undergoing change as well. Today’s fast-paced busi-
ness climate is completely redefining the way users
interact with data, so applications look very different
than those of even five years ago. Now applications
could easily be mobile and/or Web-based, therefore
breaking down basic assumptions of location and
access points that were once considered givens. In
order for developers to keep up with new demands
and variables, they must be able to write and deploy
applications that suit the way employees, customers
and partners interact with data. As a result, a whole
new set of requirements is being placed on the data
management layer that supports these applications
as well.
There is an emerging modern midtier framework
that can deliver high performance and horizontal
scalability to the most demanding applications,
while offering developers a familiar interface to write
applications to. This framework distributes data
geographically so that it is closer to the point of
access, boosting performance while taking advan-
tage of the elasticity and affordability of cloud
models. And it provides all of the redundancy and
reliability of traditional relational database manage-
ment systems (RDBMSs) without the expense,
rigidity or proprietary hardware requirements.
VMware’s vFabric SQLFire is a modern data
system that supports the applications of today and
tomorrow while delivering the benefits of a distrib-
uted cloud architecture. It satisfies users’ need for
data anytime, anywhere while allowing developers
to leverage their SQL skills and experience and
answering administrators’ requirements for enter-
prise-grade, reliable data management. vFabric
SQLFire is the data portion of VMware’s vFabric
cloud application platform that creates an agile
computing environment for highly scalable, high-
performance computing.
1
VMware’s vFabric SQLFire
3. IN THE FAST-PACED BUSINESS WORLD, the unfet-
tered flow of information is critical. No longer do
geographic boundaries or traditional hours of work
matter—now business happens wherever, whenever.
To support modern business models that are heavily
dependent on data, information must also be avail-
able wherever, whenever—and it must be up-to-date,
on hand in an instant and always reliable.
A number of recent trends, both internal and
external, are forcing organizations to change the
way they make information available to employees,
customers, partners, suppliers and others. More than
ever, enterprises are allowing employees to leverage
handheld devices and mobile applications from the
field, at home or on the road in order to stay produc-
tive and collaborative. However, if these devices
don’t offer up-to-date access to critical corpo-
rate data, the productivity of mobile employees is
severely hampered. Not only do these employees
need access to the right information any time of
day, they also expect the same high-performance
response times from applications and websites that
they access from their mobile devices as they enjoy
when sitting in front of their PC connected to the
corporate network.
Some forward-looking enterprises are addressing
performance challenges by leveraging cloud models
to geographically distribute applications and data
so that these corporate resources are closer to the
point of access. Yet if the platforms supporting these
applications can’t easily transition to a cloud model,
developers and administrators struggle to accom-
modate this distributed approach. What’s more,
with the recent trend of organizations leveraging big
data to derive business insight and push innovation,
it’s not enough to simply collect huge volumes of
structured and unstructured data. Today businesses
need to be able to analyze that data in real time—
regardless of its volume, velocity and variety—and
act on it accordingly.
From the outside, enterprises are also feeling the
pressure to provide high-quality, high-performance
products and services in an increasingly competi-
tive environment. As customers push the limits of
e-commerce, expectations for website response and
transaction times grow ever higher; customers don’t
think twice about abandoning a slow-performing site
in favor of a competing one. Much like employees,
these customers want to be able to access and
interact with websites from the device of their choice,
be it a PC, tablet or smartphone.
The confluence of these external and internal factors
is rendering traditional approaches to data manage-
ment insufficient for enterprises that need to maintain
their edge in the ever-changing business climate.
“We live in an increasingly mobile world, an increas-
ingly global marketplace, and employees are using
applications and creating the ability to share infor-
mation in real time, across continents,” says Jags
Ramnarayan, chief architect of data products with
VMware’s vFabric division. “Enterprises need to be
able to deliver the right information in a real-time,
global setting.”
2
Today’s business models
depend on data
More than ever, enterprises are
allowing employees to leverage
handheld devices and mobile
applications from the
field, at home or on the road in
order to stay productive and
collaborative.
4. 3 TRADITIONAL DATABASES that were developed in
the 1970s don’t make the cut in the current business
environment. For decades, enterprises have relied
on monolithic, proprietary and costly relational
databases to provide the foundation for their data
frameworks. However, these one-size-fits-all rela-
tional databases aren’t equipped to deal with the
unpredictable, shifting loads that have become the
norm. They weren’t designed to serve hundreds or
thousands of users at once and be able to provide
the low-latency rates that users expect. They
can’t manage the different types of data—website
comments, GPS coordinates, video clips, etc.—that
Older databases can’t keep up
enterprises need to be able to analyze in real time in
order to find answers to pressing business questions.
Nor can they support applications that are built and
deployed for a wide range of mobile devices.
Traditional databases end up leaving employees
disconnected from the data that they require to
make business-critical decisions, satisfy customers,
inform partners and push innovation. Attempting to
leverage decades-old data management technology
to solve today’s business problems can have serious
consequences on an enterprise’s ability to sustain
growth and competitiveness.
vFabric SQLFire vs. Traditional RDBMSs
VMware’s vFabric SQLFire offers a number of advantages over traditional
database management systems in the following areas:
SQLFire
memory-based, lower latency
scale linearly with
commodity hardware
virtualized, cloud-based,
distributed
consistent, real-time
view of data regardless
of where it’s stored
Traditional RDBMSs
disk-based, higher latency
scale up via expensive hardware,
adds complexity
centralized, proprietary,
monolithic
reside on a single server
ç Performance è
ç Scalabilityè
ç Designè
ç Visibilityè
5. 4
WHEN IT COMES TO CREATING the infrastructure
required to meet current application and website
performance demands, the database is typically
the most challenging element. A highly scalable
database is necessary to meet the needs of modern
applications. Enterprises require an agile, modern
cloud application platform that can provide:
Unparalleled performance of in-memory data
management. Unlike traditional database manage-
ment systems that store data in disk, in-memory
systems are optimized for low latency and high
performance, making data on demand a reality
regardless of how or from where it is accessed.
“Today, servers are more powerful, as are networks,
and memory is abundant. But one thing hasn’t
changed—physical disk speeds haven’t improved. So
there’s an order of magnitude difference between
how fast you can access memory vs. how fast you
can access disk,” says VMware’s Ramnarayan.
Real-time access to data. Superior performance
levels are also achieved by distributing data to
different locations across the globe for better prox-
imity. By enabling applications to execute stored
procedures in the database, data-intensive logic is
moved closer to where that data resides.
Continuous availability. Reliability is achieved by
replicating or partitioning database tables while
maintaining one or more redundant copies. Data
changes are replicated asynchronously as well as in
batch to one or more remote clusters for disaster
recovery. Active-active clustering across a WAN
allows application clients to read and write to either
cluster, and failover can happen to either cluster.
Horizontal scale. From the ground up, so that all
data management functions—data inserts, key
lookups, stored procedures and functions, joins,
transactions, etc.—benefit from better performance,
more capacity and higher availability.
Elasticity. Transactional websites that experience
unpredictable fluctuations in traffic need to be able
to accommodate these unexpected loads by scaling
out so that there’s no degradation in performance.
They must also scale back easily and automatically,
to reduce hardware costs and taxing resources
whenever possible.
Low-cost benefits of commodity hardware. A
modern cloud application platform that can pool
and share memory to manage data doesn’t require
expensive servers with extensive disk space, offering
significant cost savings.
Support recovery operations. With the option to
write data to disk in parallel across the cluster as a
function of the “shared nothing” architecture, fail-
ures in disk or cache don’t affect other instances of
the data.
Leverage existing knowledge. A modern cloud plat-
form must be able to incorporate existing on-staff
developer skills so that applications can be created
quickly and additional headcount is avoided.
Building the framework
for data on demand
“Today, servers are more
powerful, as are networks,
and memory is abundant.
But one thing hasn’t changed—
physical disk speeds haven’t
improved.”
—Jags Ramnarayan
chief architect of data products
VMware’s vFabric division
6. 5
A modern cloud
application platform
VMWARE’S VFABRIC SQLFIRE, which is part of
the company’s vFabric Suite, is designed to run
business-critical and Web applications in virtual
and cloud environments. SQLFire is an in-memory,
distributed SQL database that delivers dynamic
scalability and high performance for modern, data-
intensive applications. The memory-optimized
architecture of vFabric SQLFire minimizes time
spent waiting for disk access, which is the main
performance bottleneck in traditional databases. It
achieves dramatic scaling by pooling memory, CPU
and network bandwidth across a cluster of machines
and can manage data across geographies.
VMware’s vFabric SQLFire can either replace or
augment disk-oriented database architectures
with data structures and indexes optimized for fast
main memory, with options for write-through or
write-behind to disk. SQLFire is ideal for primary
data stores that require high transaction rates,
continuous availability and simplified access by
database programming staff without specialized
coding assistance.
In addition to the fast performance and high reli-
ability that users experience with vFabric SQLFire
applications, the platform also offers these benefits:
q For developers—Database application devel-
opers can easily incorporate a memory-oriented
approach to data management with the familiar,
standard SQL interface that SQLFire offers; for
example, creating a simple table in SQLFire is
the same as creating one in any other standard
SQL database. This allows developers to leverage
existing skills and experience with SQL as well as
with other established standards such as JDBC
and ADO.NET. SQLFire also works well with a
large ecosystem of compatible products and
frameworks, such as object-relational mapping
tools, including Hibernate and NHibernate;
schema-editing and database management tools;
and Spring JDBC.
What’s more, applications that use the standard
SQL syntax supported by vFabric SQLFire can
easily migrate to and from other relational data-
bases, for flexibility and future-proofing as well
as unparalleled performance.
q For administrators—SQLFire ensures continuous
availability within or across data centers and
supports granular disaster recovery to the level
of individual tables. Data is replicated across
nodes for redundancy, so if one server goes down
users are automatically redirected to an available
server, ensuring there’s no interruption in service.
Use of nonproprietary hardware offers an
economical way to achieve high database
performance at extremely large scale.
In addition, vFabric SQLFire can be used as either
a primary database for new applications—which
is particularly useful to eliminate bottlenecks
in mobile and Web environments—or it can be
integrated with existing traditional databases or
analytics engines. In the second case, vFabric
SQLFire acts as a high-performance caching layer
between an application and an existing database.
SQLFire is ideal for primary
data stores that require high
transaction rates, continuous avail-
ability and simplified access by
database programming staff without
specialized coding assistance.
7. 6
Who Needs Speed?
Innovative enterprises in highly competitive
markets such as finance, travel, insurance and
telecommunications, as well as defense and intel-
ligence organizations, are turning to VMware’s
vFabric SQLFire to solve business problems and
gain a competitive edge.
Online travel sites, for example, base their financial
success on their ability to provide lightning-fast
results to visitors who come to their websites
looking to book travel. These visitors could be
accessing travel sites from a variety of endpoints,
including mobile devices. While presenting search
results for site visitors might mean having to pull
real-time data from a variety of different sources,
such as partnering airlines and hotels, all that the
visitors care about is getting a price quote quickly.
Being able to provide a full-featured site regard-
less of access method and return fast query results
is a business-critical issue, since a website’s ability
to provide price quotes has a direct impact on
the amount of revenue it generates. In this highly
competitive market there are few second chances;
visitors who have a poor experience at one travel
site aren’t likely to come back, opting instead to take
their business to a competitor.
“The data these sites rely upon needs to be relevant,
real-time and fast. If a site is sluggish, that company
isn’t going to be able to compete well. It’s important
to deliver the highest performance possible with
these applications,” says Blake Connell, product
marketing manager for VMware’s vFabric SQLFire.
“But it’s not just high performance that’s needed;
there’s also a real need for sustained, steady-state
performance regardless of spikes in activity, so that
site visitors keep coming back.”
There are many examples of an enterprise’s revenue
potential being tied to data performance. A tele-
communications company that found itself suffering
from low customer-satisfaction levels realized it
needed to revamp its customer-facing system in
order to respond better to inquiries and to also
enable the development of timely, personalized
offerings. The existing batch-processing system,
while rich in customer history and information, could
only process up to 60 transactions per second. The
prospect of rewriting the applications that were
dependent on this system proved too expensive,
and so the telecom company began seeking a way
to improve performance while still being able to
preserve the back-end system.
The company decided to use VMware’s vFabric
SQLFire as a front-end data management system in
order to develop new Web and mobile applications
that offered much better data responsiveness; these
new applications were able to reach 600 transac-
tions per second. It was by leveraging VMware tools
to extract data from applications and place it in a
container attached to SQLFire that the company was
able to achieve this significant increase in transac-
tion rates. This improvement had a positive impact on
satisfaction levels among existing customers, and the
company can now serve up personalized offerings to
attract new customers and improve the business.
CASE STUDY
8. 7 TODAY, fast access to information is more essen-
tial than ever, and corporate data has become the
cornerstone of many business models. Without
immediate, reliable access to that data, enterprises
run the risk of losing customers, hampering produc-
tivity and missing out on important business oppor-
tunities. Data management systems must be able to
deliver the right data to employees, customers and
partners anytime, anywhere in a way that’s reliable,
cost-effective and leverages existing resources.
VMware’s vFabric SQLFire has the power to deliver
these capabilities and more to today’s demanding
users, while easing back-end management and
application development. With its in-memory,
distributed design, the database delivers dynamic
scalability and high performance for today’s modern,
data-intensive applications. n
Conclusion
www.vmware.com
VMware’s vFabric SQLFire delivers the data management
platform required by today’s modern applications:
In-memory data management with optimized disk
persistence. vFabric SQLFire enables applications to
manage data entirely in memory by using partitioning
and synchronous replication to distribute the data across
numerous SQLFire members. vFabric SQLFire provides
an optimized disk persistence mechanism for long-term
storage. 2012 applications can use vFabric SQLFire to
actively cache table data from a traditional disk-based
RDBMS. A VMware study found vFabric SQLFire performs
30 times faster than the Oracle 11g database.
Continuous availability, elastically scaled, low latency.
A flexible architecture enables vFabric SQLFire to pool
memory and disk resources from hundreds of clustered
members. This clustered approach provides extremely
high throughput, predictable latency, dynamic and
linear scalability and continuous availability of data. By
co-locating application logic with data and executing
application logic in parallel, vFabric SQLFire substantially
increases application throughput.
High adaptability to existing applications. vFabric SQLFire
is implemented entirely in Java, and it can be embedded
directly within a Java application. vFabric SQLFire members
can also be deployed as standalone servers that partici-
pate in a cluster. Java applications can connect to a vFabric
SQLFire cluster using the provided JDBC drivers. Microsoft
.NET applications can connect using the provided ADO.
NET driver. The use of JDBC, ADO.NET and SQL means that
many existing database applications can be easily adapted
to use a SQLFire cluster.
vFabric SQLFire introduces several extensions to common
SQL data definition language (DDL) statements to manage
data partitioning, replication, synchronization with data
sources and other features. However, most common
queries and data manipulation language (DML) statements
are based on ANSI SQL-92, so experienced database appli-
cation developers can use their knowledge of SQL when
working with vFabric SQLFire.
Parallel, data-aware stored procedures. As with
common relational databases, vFabric SQLFire enables
applications to execute stored procedures in the data-
base to move data-intensive logic close to where that
data resides. vFabric SQLFire extends this functionality
by making application logic, as well as data, highly avail-
able. Procedure execution is routed transparently to the
appropriate data nodes; this approach increases applica-
tion throughput.
“Shared nothing” disk persistence. vFabric SQLFire
manages replicated and partitioned tables completely in
memory, or both in memory and on disk. In vFabric SQLFire,
each member node can cause data to persist in disk files
independently of other members. Failures in disks or cache
failures in one node do not affect other instances’ ability
to safely operate on their disk files. This “shared nothing”
persistence architecture allows applications to be config-
ured such that tables can be persisted on different nodes
across the cluster, reducing the impact of disk latencies.
Key features of VMware’s vFabric SQLFire