Traffic Simulator

TRAFFIC
SIMULATOR
Predictability and right business decision of the planned
software and hardware redesign
To make the right business decision and optimal price
allocation for the planned software and hardware redesign.

A few words about load-testing
Objectives of load-model appearance:
Load models are realized for IT - performance checking in terms of query speed
subject matter and procedures on the system level, as well as checking the
correctively of a particular operation performance, while one or few objectives are
achieved:
To decrease the
system drop
accidents
In case of the
changes in setting
and regimes of
information system
exploitation.
To increase
objectivity in
decisions of IT
design changes
In case of making a
choice of new
settings and regimes
of information system
exploitation.
Balanced
investments and
results
i.e. the amount
funding и obtaining
benefits during the
product decision
making.

Prerequisites for load model appearance:
Prerequisites for synthetic models design of load-testing conditionally can be
divided on two classes:
1. Risky IT-system changes
• Hardware changes or regimes of its exploitation ( for an example, migration
to new etc.).
• Reallocation of hardware resources (for an example, serve more then 1
system).
2. Questionable effectivity of changes:
• Delicate adjustments of hardware (for an example, DBMS settings and etc.)
• Making a choice/changes in hardware (for an example, choosing a new
vendor and/or new hardware configuration design).

A few words about load testing
Difficulties in realization:
№1. It is possible to make a load model. И it gives its own results in one or another results extend in
context of objectives achievement.
№2. The design of full functional load model is non-trivial. While it requires a serious financial funding.
№1. The analysis of information flow is non-trivial. The information flow in the system needs to be
analyze on the preparation stage, preferably in the moment pick-load. Without online monitoring service
and fixation of getting data, a task becomes non-trivial.
№2. The design of load-model is resource-intensiveness. If it was possible to conduct a full-size
analysis of the information flow, to divide it on components, the next step is to design the load-model: work-
out of scenarios and emulators complex, «robots», scripts and etc., that requires to qualified specialists
attraction.
№3. «Material» of conducted tests require of system analysis. The analysis of the gotten statistics
becomes the last step of the load-testing, in other words it requires to interpret the gotten information, to
make чтобы составить a valid conclusion in accordance with the conducted tests.
Confirmation
Facts

Full-functional load-testing technology, TRAFFIC SIMULATOR,
allows:
GYSTELL developed its own program product TRAFFIC SIMULATOR, as out-service of SQL
server, which allows:
1. To track the traffic of the productive system, while practically not loading it;
2. To reproduce written traffic, practically identical, as one-user regime, as taking into
account many-flows and competition, on another version of MS SQL server, another
hardware, with other settings;
3. To compare gotten data in terms of traffic accurate within the size, to get unpredictable
errors, uncommon system behavior. In the result, the possibility appears to conduct as
load as performance automated testing.
4. To compare the data in the origin and in the copy in the automated regime.

Comparison with self-writingsystems of load-testing:
Requirements Self-writing
system
GYSTELL Traffic
Simulator
Comments
Development of the
model, scenario, scripts
and etc.
Required Not required The price of the solution is
comparable with the price of full
programs STS
Application with any MS
SQL system
Not possible Possible in the scope of
functionality states of STS
Self-writing model is designed for a
particular system
Possibility of the
full- functional load-
testing
Not possible Possible in the scope of
Self-writing model is projected and
developed as known for one or few
information system functionality
parts
Possibility of the usage
of information system
with changed
functionality
Possible with
limitations
Possible in the scope of
In case of changes made in
information system functionality it
is always needed self-writing
model debugging

GYSTELL Traffic Recorder: responsible for the traffic tracking, going through the
DBMS server. Installation is as service on the DBMS server.
DB ServerApplication Server
GYSTELL Traffic Recorder
GYSTELL Traffic Simulator
GYSTELL Traffic Simulator – Traffic tracking:

Traffic tracking (management console):
In case of traffic tracking is turn on:Preparation for traffic tracking:

Monitoring complex:
GYSTELL Monitoring Service: is responsible for gathering and analysis of
parameters of productivity with working and testing environment. It is installed as
services on desktop or testing stand, as well as harvesting MS SQL tracks on
testing stand.
Server DB
Monitoring ServiceTraffic PlayerSQL formatted
track (file)

Traffic analysis:
GYSTELL Traffic Compiler: is responsible for the analysis and reотвечает за анализ и
working of harvesting data. It is installed on testing server, on which functional and load
testing will conducted. The functionality of given service includes:
• Replacement of testing require;
• Definition of dynamic parameters;
• Automated definition «floating», heterogenic parameters;
• Settings of excluding rules and replacement for floating parameters;
• Settings of rules for sections plugging on testing stand (taking into account there is no
possibility to set connection with DBMS protocols of security, which is analogical to
origin).
Traffic Compiler

Traffic analysis (print screen of management window):
The entering file is pointed out (gotten during the tracking) and exiting folder. The
separated file will be made with compiled traffic for each session.

Traffic reproduction:
GYSTELL Traffic Player: is responsible for reproduction and coordination
written and worked-out tracks and for between-flows interactions.

Reports block:
GYSTELL Reporting Service: is responsible for analysis of original tracks, harvested on
testing stand. Show the difference between records results, performance errors, data
difference. It shows the difference between accomplishment, time performance, time
CCR, volume of logical writings.
Configuration 1
Configuration 2
Configuration 3

Reports block (sample form 2):
Allows both to switch at DBMS from MS SQL 2008 to MS SQL 2014 during reproduction of recorded
traffic and to define which queries and how fast they will be performed.
TimeLess – shows, that
time was decreased
Time, in which query was
performed promptly
Speed-up query in percents

Allows to see which quires and on how much slower will be performed
TimeLarge – shows,
the performance
time increased
Time, during which the
query performance slow
down
Slow down performance in
percents

During traffic reproduction on new hardware configuration or new DBNS version, it is possible to
determine with queries will be performed not correctly or an error will appear.
Binary
Error

Monitoring tools:
During reproduction:During the record:
The place of drastic PLE drop was emphasized by color.
During reproduction the drastic PLE drop has been seen, increase in capacity of used memory.

Comparison of MS SQL 2012 and MS SQL 2014 performances:
with TRAFFIC SIMULATOR the traffic record of quires was conducted with information system
«1S:Enterprise» for 40 minutes.
Parameters:
Number of quires: 602147
Combined length of quires on MS SQL: 1279,2 seconds
Version DBMS for record: MS SQL Server 2012 Enterprise
MS SQL Server 2014 CTP2 with identical MS SQL Server 2012 settings were set on analogical hardware.
Then reproduction of traffic quires to MS SQL Server 2014 occurs, while information about the process
time length was the same.
Gotten result during reproduction:
Number reproduced and correctly performed quires: 602147
Combined length of performed queries on MS SQL: 1121,347 seconds
So, linear speeding up of quires in case of transfer from MS SQL Server 2012 to MS SQL Server
2014 composes more than 12%. To exclude such factors, as «heating up» of cache and others, the
traffic data was recorded 3 times for each version MS SQL Server, the results were identical to the above
mentioned.

Traffic Simulator

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Traffic Simulator

Similar to Traffic Simulator (20)

Recently uploaded

Recently uploaded (20)

Traffic Simulator