Exalytics - Frequently Asked Questions

(FAQ)

1) What is Oracle Exalytics?

Oracle Exalytics is a new "BI machine" designed specifically for

Business Intelligence and Enterprise Performance Management.
With a massive 1TB RAM and 40 CPU cores, this machine can and
will take performance to a level never seen before.

The reason why performance is to extreme is that the Exalytics

architecture is designed to store your business intelligence data in
main memory, this means response time is dramatically faster with
no network latency or disk I/O.

The term "speed of thought" has already been used a lot with
Exalytics - this is referring to the fact that your reports run so
quickly that they refresh with new data even whilst you are typing in
the filter criteria.

2) What components come with Exalytics?

Exalytics is a pre-installed / pre-configured machine running on

Linux 64-bit. It comes with 3 main software components:

- Oracle BI Foundation 11g (OBIEE release 11.1.1.6)

- Oracle TimesTen (in-memory database)
- Oracle Essbase 11g (memory optimized)

3) Is there a demonstration of Exalytics anywhere?

Yes, see this recorded demonstration on

YouTube: http://www.youtube.com/watch?v=oLLic4ytFsw

4) What is the difference between Exalytics, Exadata and

Exalogic?

Fundamentally, the three "Exa" machines built by Oracle are

designed to overcome bottlenecks in either one of 3 areas: Memory,
Disk, CPU

Memor Exalytics is designed to massively boost performance of Business

y: Intelligence queries by processing vast amounts of data in memory

Exadata is designed to optimise the performance of large databases

Disk: where most of the query overhead is experienced whilst reading or
writing from disk

Exalogic is designed for large application servers (e.g. server farms,

CPU: data centres, virtualization) with massively parallel and scalable CPU
performance

5) I already have Oracle BI 11g, does Exalytics replace my

existing hardware?
Yes. Your existing Oracle BI 11g application would be migrated in its
entirety to the new Exalytics machine.

6) What is TimesTen?

TimesTen is Oracle's "in-memory" database - this means it stores all

its data in memory so there is no disk overhead. Historically
TimesTen has been used for applications which require lightning fast
real-time data management. For example: Banks would use
TimesTen to do fraud checks in a fraction of a second.

With the introduction of Exalytics, TimesTen has been enhanced to

support analytical processing for business intelligence queries. For
Exalytics, your Oracle BI 11g queries will be directed against the
TimesTen in-memory database cache rather than your underlying
data sources.

TimesTen also comes with "columnar compression" which means

that columns of data (e.g. Customer Name) are compressed at ratios
of upto 5x to help expand the amount of data that can be stored in
memory.

7) How does the TimesTen in-memory database get

populated?

The TimesTen in-memory database actually gets populated

dynamically by the Oracle BI 11g server - so you don't have to
manually write any scripts. The mechanism is actually an enhanced
form of the "Aggregate Persistence Wizard" which has been a
feature of Oracle BI since version 10g.

With Exalytics, a "storage advisor" can be used to automatically

analyse the usage of your BI deployment (i.e. the queries initiated
by your end users) and make recommendations as to what data
should be cached in TimesTen. You can then accept these
recommendations (if you wish to do so) or you can make your own
decisions.

The cache structures within the TimesTen database will reflect the
structure of the "Business Model" within your BI Repository (RPD).

So if you have a logical dimension table "Dim - Account" with three

columns "Account ID", "Account Name", "Account Type" then a table
will exist on the TimesTen in-memory database with a similar
structure. Note that if you choose to cache your data at the higher
"Account Type" level then obviously the in-memory database will
not contain any data for "Account Id" or "Account Name".

For your fact logical tables, the TimesTen in-memory database will
also contain tables to cache your logical facts. So there will be
columns to store "$ Revenue" and then surrogate keys to join to the
related dimension structures on the in-memory database (e.g. for
Dim - Account).

The actual physical tables created automatically on TimesTen will

have system generated names.

8) Can we cache further data into the in-memory database

whilst the system is running?

The answer from Oracle Product Development is "yes". So if a new

BI Dashboard is published during the week then you can
dynamically start caching the underlying data into the TimesTen in-
memory database without any downtime.

9) The in-memory database seems to be similar to BI Server

"caching", is this true?

No, not really! BI Server caching is for caching the end results of
"Logical SQL" queries. With Exalytics, the TimesTen in-memory
database caches data in structures that resemble the Business
Model in your RPD.

BI Server caching is optimized for lots of small results sets whereas

TimesTen in-memory database can handle upto 5TB of cached data
(making use of compression). BI Server caching is also a file-based
cache, whereas TimesTen data will reside totally in RAM.

There is no reason however why you should not also use BI Server
caching Exalytics - as it will cache the results of any ad hoc or other
queries which were previously not sourced directly from the
TimesTen in-memory database. The BI Server cache on an Exalytics
machine can be tuned to reside on a "RAM disk" so it also benefits
from in-memory capability.

10) My data is not stored in an Oracle database, can I still

use Exalytics?

Yes, definitely. Because the TimesTen in-memory database is

populated through the Oracle BI 11g Server, the underlying data can
be stored in just about any type of database.

11) It sounds fantastic, but does Exalytics have any

limitations?

So far the signs are extremely positive and Oracle have conducted
several successful pilots with "beta" customers. However, there are
a few limitations which are worth highlighting, the first of which is
common to all "aggregate layers" within a BI architecture:

1) To make use of the TimesTen in-memory database, your

calculations must be "additive" i.e. you must be able to accurately
derive the calculation results from the data stored in the TimesTen
cache. So you will need to take care when performing averages, for
example. COUNT DISTINCT operations are also non-additive.

2) In order to make use of the TimesTen in-memory database, you

need to bulk load data into it. So this means there will be an extra
task at the end of your ETL cycles to repopulate the entire TimesTen
cache. Exalytics comes with extremely fast network cards so data
transfer can be rapid, and there is even an Infiniband (40GB/s)
connection to connect to an Exadata machine. There is a "double
buffering" technique with high-availability architectures which could
potentially simulate incremental and trickle-feed loading.
UPDATED: Incremental loading for Exalytics is now possible
using Peak ETA! Please visit the Peak ETA Product Overview page
for details on how Peak ETA can significantly enhance the
capabilities provided by the Exalytics "Summary Advisor" by
incrementally loading TimesTen in-memory aggregates and in
parallel.

3) We have seen that Ragged, Skipped-Level and Parent-Child

hierarchies are not supported. So this means only level-based
hierarchies are supported, for the time being at least.

12) I don't currently use Essbase, so how would I use the

Essbase component on Exalytics?

The memory-optimised Essbase that ships with Exalytics is designed

to replace your existing Essbase cubes and boost the performance
of your EPM applications. If you don't currently use Essbase then
your initial focus should be on the Oracle BI 11g Foundation and
TimesTen components of Exalytics. However there is nothing to stop
you from using the Essbase functionality whenver it suits you!
13) I have more than 1TB of data, can I still use Exalytics?

First of all, it is worth pointing out that although the Exalytics

machine comes with 1TB of RAM, the reality is that around 400GB
will be available for your "in-memory" data. The reason is that the
TimesTen has two main memory configuration parameters:

- PermSize: The permanent memory region - this is where your data

is stored
- TempSize: The temporary memory region (used for sorts etc)

The recommended setting for both of these memory regions will be

400GB, consuming a total of 800GB RAM. This will leave
approximately 200GB for OBIEE and any other overheads.

However, despite the 400GB recommended limit, TimesTen also has

a compression feature so that you could potentially fit 5x more in to
your in-memory cache.

You should also remember that Exalytics is designed to be an

"aggregate" layer, so you shouldn't really be thinking about loading
ALL you data into the TimesTen in-memory cache. Instead, you
should aim to load in as much summarised data as possible.

So if you have >1TB of data then you can still use Exalytics!

NOTE: The memory-optimised Essbase component does not have a

compression feature.

14) Does Exalytics support real-time Business Intelligence?

Yes and no. Your Exalytics queries do not all have to come from the
TimesTen in-memory database, so you could report directly off your
live transactional database and perform real-time BI.

However, as things currently stand, the Exalytics TimesTen in-

memory database is not configured for "trickle-feed" incremental
loading to support near-real-time BI. So if you need to boost the
performance of your real-time database queries then Exalytics may
not be the answer for you.
As stated above, there is a "double buffering" technique with high-
availability architectures which could potentially simulate
incremental and trickle-feed loading.

15) Does this mean I no longer need a Data Warehouse?

Absolutely not! Data Warehouses are fundamental when it comes to

data consolodation, cleansing and scalability (for storing enormous
quantites of data in a format suitable for reporting purposes).

You should view Exalytics as an "aggregate" or "performance" layer

in your architecture - sitting on top of your Data Warehouse (such as
Exadata).

16) With Exalytics can I still use Maps, Mobile Analytics

(iPad), and Balance Scorecards?

Yes, the entire range of features available with the Oracle BI 11g
Foundation suite is supported.

17) How is Exalytics licensed?

As always, you should contact your Oracle Sales Rep to gain full
clairity on the license situation. But just as a guidance, licensing will
typically be done on a per user basis:

- Minimum 100 named users for both Times Ten & BI Foundation
Suite
- Existing customers of OBIEE or BI Foundation Suite can migrate
licenses to Exalytics

Exalytics supports clustered solutions for disaster recovery and high

availability:

- Requires a minimum of 2 hardware machines

- You need to also have a Weblogic Server Enterprise Edition full
license
The accurate pricing is available on Oracle's price list. To avoid
reporting incorrect pricing figures, we will not hard code any pricing
or links in this FAQ! However, you can search on Google for things
such as "exalytics license cost" to get the latest information.

18) I am already licensed for Oracle BI 11g, can I migrate my

existing licenses?

Yes, see question 17.

19) How has Essbase changed for Exalytics?

Essbase on Exalytics has a number of optimizations for in-memory

operation including improvements to overall storage layer
performance, enhancements to parallel operations, enhanced MDX
syntax and a high performance MDX query engine. Essbase on
Exalytics provides up to 16X faster query execution as well as up to
6X reduction in write-back and calculation operations, including
batch processes. These enhancements are particularly important for
advanced use cases such as planning and forecasting.

20) Once the box arrives, do I have to install Exalytics or is it

all pre-installed?

Once you have the box plugged in you will then need to download
and install the Exalytics software components: OBIEE, TimesTen and
Essbase (Optional).
There are two main reasons for this:

1) You are always downloading and installing the latest software

version
2) You are not tied down to a specific hostname or network
configuration

A utility is provided to fully install and configure OBIEE and

TimesTen together.
21) What happens if I load in more data than TimesTen's
memory capacity?

The short answer to this is that you can't! TimesTen is designed to

work with all data in memory. So if TimesTen has 400GB allocated to
your data, then you will not be able to load in more than this. If you
try then you'll get an error!

22) What happens if I reboot the machine, do I lose my in-

memory cache?

Although TimesTen is an in-memory database, the data is actually

also stored on disk. So if you reboot the Exalytics machine then you
won't lose your data! When you start up the TimesTen services after
the reboot it will load all the data from disk into main memory.

NOTE: to avoid data loss/inconsistency you should always shut down

TimesTen services gracefully before you reboot the machine.