Sizing Guide

Introduction

This topic helps "Prebuilt Dashboard for API" customers select the machine configuration (CPU, RAM, and disk space) required for Decision Insight to run the "Prebuilt Dashboard for API" application.

Several benchmarks were run based on different customer profiles to get a suitable configuration to enable the application to run with an acceptable performance.

The information provided in this document only give you an idea about the configuration you may need, but It does not guarantee you 100%.

Performance measurement

Performance is acceptable when you can navigate inside the dashboard with a good response time and when incoming transactions are absorbed without latency.

The performance of the application depends on several criteria:

  • Volume of transactions – Direct impact on load and performance. More transactions mean more data to process and persist. 

  • Peak throughput – You may have a reasonable average daily number of transactions, yet have a massive throughput during a set period of time. This huge throughput can impact the performance and make dashboards unavailable during the peak throughput.

  • Number of instances of analytics dimensions and computed attributes – The more dimension instances and computed attribute we have, more time we need to compute and update them.

Note: The number of live instances and their lifetime also impact performance in general. In the case of Prebuilt Dashboard for API, transactions have a very short lifetime so you should never have a large number of live instances.

Disk space measurement

The disk space needed depends on the following criteria:

  • Volume of transactions 
  • Number of instances of analytics dimensions and computed attributes
  • Data retention period for transaction
  • Data retention period for aggregated data
     

Customer profile

To get your customer profile, you need the following information:

  • Volume of transactions – Give an estimation of the number of transactions per second.
  • Computation score – The computation score for a customer profile is evaluated using the number of instances of the following dimensions:
Dimension Number of instances
Gateway
Remote Hosts
API
API Method
Client Application
Organisation
Protocols
Alert Type
Transaction Status
Gateway Type
User


Get your computation score

To evaluate your customer profile, calculate your computation score, that is the number of computations your application performs per second.

The number of computations impacting the performance depends on:

  • The number of instances of dimensions.
  • How they are used during data aggregation.
  • The number of computed attributes and their rhythm for each aggregate. These computed attributes are evaluated at different rhythms. Some are evaluated every 5 minutes and some others are computed every hour or day.  

Please note that the number of dimension instances that you fill in this table do not all have the same impact. For example, the API method or the Client Applications  dimensions may have a strong impact on your performance because they are used in an aggregation schema with three axis (example: API Method by Client Application by Transaction Status).

To take all these parameters into account, use the attached abacus excel file, Computation Score Abacus V2.xlsx, to get your computation score and evaluate your profile.  This computation score represents the complexity (the size) of your configuration. The score will help you determine the suitable configuration.

The picture below describes the abacus.


Abacus sheet


Computation sheet





  • Data retention period

    Type of data Retention period (in day)
    Transaction
    Aggregated data

Configuration by customer profile

RAM and CPU

The matrix below help you to determine the number of RAM and CPU needed for your profile.

Computation Score


Throughput (tps)

< 50 50 to 150 150 - 300 300 - 500


< 50


RAM: 4 GB

CPU: 2

RAM: 4 GB

CPU 4

RAM: 4 GB

CPU 4



50 - 100


RAM: 6 GB

CPU: 2






100 - 200



RAM: 8 GB

CPU: 4




200 - 500




RAM: 8 GB

CPU: 4



500 - 1000





RAM: 22 GB

CPU: 8

Note: The quantity of RAM in the table above corresponds to the quatity of RAM needed for Decision Insight. You must add 2 GBof RAM for overhead and 2 GB for the Oerating System.

Disk space

Tow types of data are stored on disk:

  • Transactions – The disk space required depends on the number of transactions by day and its data retention period (in days)
  • Aggregated data – The disk space required depends on the total number of computation by day (= Computation score * 86400) and its data retention period (in days)

You can apply the following formula to get the total disk space need:


Disk Space = (SizeOfTransaction * NbTransaction * TransactionRetentionPeriod) + (SizeOfAggregate * NbComputation * AggDataRetentionPeriod) + Constant

Note:

  • In Decision Insight, each time an aggregate is evaluated, Decision Insight generates a record. For example, for a rhythm of 5 minutes, we have a record every 5 minutes.
  • SizeOfTransaction: Average size of a transaction
  • NbTransaction: Number of transaction by day
  • TransactionRetentionPeriod: Number of days the customer want to keep the transaction in database
  • SizeOfAggregate: Average size of a record of an aggregated data 
  • NbComputation: Computed from the number of instances of the analytics dimension (using abacus above)
  • AggDataRetentionPeriod: Number of days the customer want to keep the aggregated data in database
  • Constant: ADI needs some more disk space for its its internal used


Related Links