Software Requirements

The following software is needed to install Pure

Java

Pure requires a Java Developers Kit (JDK) installed on the server.

It is recommended to use a 64-bit version of Java as most Pure installations will require more than 4 GB memory.

Supported platforms

Tomcat Server

Pure requires a Tomcat servlet container installed on the server.

Supported platforms

Database

Pure stores all metadata in a relational database, so a database should be available for Pure during installation.

Pure supports Oracle, PostgreSQL, and SQLServer.

Supported platforms

Oracle

To use Oracle with Pure the following configuration is needed:

• Create table
• Create sequence
• Create trigger
• Create view

Granting rights to create view can be done by executing the following statement:

GRANT CREATE VIEW TO <insert_user>

On Oracle it is recommended that the character set used is able to contain special/national characters, AL32UTF8 or similar is recommended.

Please make sure that the NLS_* localisation parameters are set to appropriate values (non-default). Pure relies on the database sorting so if the NLS_SORT parameter is not specified explicitly the ordering will be erroneous. The NLS_SORT parameter is set in the SPFILE (also editable in the enterprise manager). Valid values can be found by running the following SQL: select * from V$NLS_VALID_VALUES where parameter = 'SORT';
Typical values for our current customers would be: WEST_EUROPEAN, DANISH, GERMAN or FINNISH.

PostgreSQL

To use PostgreSQL with Pure the following configuration is needed:

• Create table
• Create sequence

To get the best performance on PostgreSQL, set default_statistics_target to 10000 in the configuration file. This ensures that the query optimizer has access to detailed information about the data distribution in Pure.

The following minimum memory settings should also be used: temp_buffers = 32MB ; work_mem = 32MB; shared_buffers = 512MB.

It is recommended to use UTF-8 encoding, and the locale should be set according to how you want results to be sorted in Pure.

Ensure that the PostgresSQL database is created with the public schema.

SQLServer

To use SQLServer with Pure the following configuration is needed:

• DB owner

dbo is the default database schema

• Create view

Granting rights to create view can be done by executing the following statement:

GRANT CREATE VIEW TO <database_principal>

Specific SQLServer Notes

To avoid database locks on SQLServer row versioning-based transaction isolation should be enabled. This is done with the following SQL statement:

ALTER DATABASE <database name> SET READ_COMMITTED_SNAPSHOT ON WITH ROLLBACK IMMEDIATE;

See http://msdn.microsoft.com/en-us/library/ms345124.aspx for more information.

Ensuring the database has the correct collation - must be case insensitive and accent insensitive (i.e. end with _CI_AI)

-- creation of new database (in SQL_Latin1_General_CP1)
CREATE DATABASE <database name> COLLATE SQL_Latin1_General_CP1_CI_AI;
  
-- before migration of existing (to SQL_Latin1_General_CP1)
ALTER DATABASE <database name> COLLATE SQL_Latin1_General_CP1_CI_AI;

See http://msdn.microsoft.com/da-dk/library/ms184391.aspx for more information

Browser

To complete the Pure installation you will need to access Pure through a web browser that supports JavaScript and AJAX

Supported platforms

Pure Hardware Recommendations

CPU

Pure's performance is directly linked to CPU performance. Our current recommendation is at least dual Intel Westmere X5670 (2×6 cores + HT = 24 CPU’s for the OS, 2.93 GHz). We do not recommend SPARC.

Our experience with SPARC hardware is that it is optimised for running applications where the work can be cut up into relative small "work units", these work units can then be executed in parallel on the many cores provided by the SPARC CPU. In Pure our work unit is a page request by the end user, and unfortunately our experience is that this is too large to provide really good performance on SPARC CPUs.

Memory

We recommend a minimum of 8 GB memory for a server running Pure. We recommend 16 GB memory for large installations.

If you plan on running the database on the same server as Pure you should give it at least double the memory of a server running only Pure.

Disk Storage

We recommend minimum 70 GB disk storage for Pure's system data. The storage needed for full-text files depends very much on your organisation's full-text file deposit policy, so we recommend storing full-text data on an expandable file system; e.g. a SAN.

The database will in most cases use from 5 GB to 30 GB, mostly depending on how long you choose to keep the Pure job logs in the database.

Server Example

Example of a recommended server for running Pure. This will be able to run both the database and Pure on the same server. The example here is a Dell server, but any vendor will do.

Updated on 2020-07-23

Dell PowerEdge R640

2.5” Chassis with up to 8 Hard Drives and 3PCIe slots, 1 or 2 CPU

RAID 10

PERC H730P RAID Controller, 2Gb NV Cache, Minicard

Intel® Xeon® Silver 4210R 2.4G, 10C/20T, 9.6GT/s, 13.75M Cache, Turbo, HT (100W) DDR4-2400

Intel® Xeon® Silver 4210R 2.4G, 10C/20T, 9.6GT/s, 13.75M Cache, Turbo, HT (100W) DDR4-2400

4 x 16GB RDIMM, 2933MT/s, Dual Rank

4 x 480GB SSD SATA Mix Use 6Gbps 512 2.5in Hot-plug AG Drive, 3 DWPD, 2628 TBW

Virtualisation

Pure runs in virtualised environments. We use VMware, but other platforms are also supported.

The following quote from Atlassian sums up or experience very well:

In short, Pure will run in a virtualised environment, however, the setup and maintenance of such an environment is solely up to the customer (we cannot provide assistance in tweaking or tuning such an environment, as there are too many variables). If a customer experiences performance issues with Pure when running in a virtualised environment, then we recommend removing the virtualisation - if not, then any tweaking and optimisations in such an environment must be carried out by the customer on its own (we do not have the resources to assist in setting up or validating a virtualised environment).

Review

We are happy to review a brief description of any environment being considered, please just go through your normal Elsevier contact person.