Workload Qualification - Oracle Exadata Recipes: A Problem-Solution Approach - page 77

Table 4-1. ( continued )

Software Feature What to Look for in Current Database Exadata Fit Assessment Comments

Smart Flash Cache Variable data access patterns on

databases with active data sets larger
than a very large buffer cache can
support, combined with a requirement
to avoid unnecessary disk I/O or desire
to bypass SSD/PCI flooding by non-
critical operations (such as backups)

Smart Flash Cache is a feature that can only help
and never hurt, but many competitors offer SSD/
Flash solutions at lower costs.

Smart Flash Logging Very active bursts of redo log writer

activity that physical disks and disk
cache can't consistently handle

Smart Flash Logging is a recent Exadata
enhancement that provides physical disk redo
write relief in times of high LGWR activity; alone,
probably not a driver for Exadata.

Storage Indexes Frequent large sequential scans on

well-ordered, un-indexed data. You can
obtain this by examining data access
patterns from AWR/ADDM and having a
deep understanding of your data.

Storage indexes are another software feature that
never hurts and always helps. Storage indexes are
not a means to provide performance assurance.
Storage index entities are transient in nature and
outside control of administrator or developer and
thus shouldn't be a driving factor for Exadata.

I/O Resource
Management

You have a desire to consolidate
your database platform and ensure
I/O performance isolation between
competing resources.

Depending on your budget, IORM could be a key
Exadata enabler in your business. IORM provides
more granular, Oracle-aware I/O prioritization
and SLA isolation than other hardware vendors
can offer.

How It Works
With Exadata's software functionality in the storage servers, combined with its balanced, high-performing hardware,
Oracle has designed the platform to reduce or eliminate disk I/O as a bottleneck. As such, when performing a
workload fit assessment for Exadata, you should focus on identifying areas where disk I/O latency and/or throughput
is a bottleneck, and further understand whether Exadata's software functionality is well suited to address, reduce, or
eliminate the bottleneck.
This recipe is clearly geared toward Exadata Smart Scan. Most companies that purchase or are evaluating Exadata
do so for performance reasons. Smart Scan provides Exadata's most important performance-related advantage
and the software that allows Exadata to be more than “just a bunch of fast hardware.” Exadata's other software and
hardware capabilities certainly provide advantages over many competing solutions or products, but in the real world
these rarely provide the overall business value Smart Scan does.
One of the points in the Smart Scan examples above should help illustrate that the “Exadata fit discussion” is
not necessarily a matter of “data warehouse vs. OLTP vs. mixed workload.” These workload classifications have
industry-defined definitions that are relatively well accepted, but in the real world the application and database
design are the factors that truly identify your data access patterns. Your data access patterns are the key to unlock
Exadata software features.
In contrast with most of the other recipes in this topic, the main emphasis here isn't technical in nature, and
it's really not even directly related to Oracle Exadata; it's about your guiding processes to provide the right technical
solution to a business problem. You first need to understand where your business problem lies. Do you have
application performance issues that are directly limiting your ability to generate revenue? Do you have performance
problems that are negatively impacting end-user experience? What is the cost, in dollars, to the business, based on
these problems? How much are you willing to spend to resolve these issues? With these questions answered, you can