If you are considering Perfion, two questions usually come up early: how quickly does the system respond when working with your products, and which performance tier should you choose? This article aims to visualize how Perfion scales horizontally by adding more items to the database while keeping the product configuration the same.

We measured two usage scenarios, on product databases of 25,000 and 100,000 items and 159 features.

The two databases are identical in configuration, and they only differ on the number of items in them (horizontal scaling).

Neither of the two databases have any self-remote or normal remote features and the tests are conducted with only one concurrent user using the system.

Tested three performance tiers

Perfion is available on numerous tiers. They differ in the amount of computing resources allocated to the Perfion database.

We have tested the 2 database sizes with the three most used performance tiers:

Throughout this article we use the names SR1, DR1, and DR2 to keep the comparison easy to follow.

What we measured

We focused on two typical actions that a product manager performs.

1. Searching for all items in grid

A feature search for all items, with the page size set to 1,000 items and every product feature included in the result. This is the most demanding read in everyday use — it loads not only the product rows but every value for every feature on them.

The Items view in Perfion with a 1,000-item page and the full feature set displayed.

The features included in the search are all the features defined on a Product in Perfion. For the test database, a total of 67 features were configured, as seen below:

The product feature configuration used for the test — every feature shown here was part of each search.

2. Opening an item for editing

Opening the item editor on a single product. This is the most common interactive action in Perfion: clicking an item to view or change its details. The measured time covers everything from the click to the fully loaded editor with all feature groups (Main Information, Marketing, Technical Specifications, Media, Workflow, and eCommerce) ready to use.

The item editor with all feature groups fully loaded.

How the numbers were collected

On each tier (SR1, DR1, DR2):

• The all item search was run 25 times in succession; each elapsed time was recorded.
• The item editor was opened 25 times on the same item; each elapsed time was recorded.
• The averages of those 25 runs are what the charts below show.

Results — 25,000 items

With a product portfolio of 25.000 items, Search all item searches finish in 2.4 to 3.1 seconds depending on tier, and the item editor opens in well under a quarter of a second on every tier.

Average response times (ms) at 25,000 items, by tier. Lower is better.

Results — 100,000 items

With a product portfolio of 100.000 items, Search all item searches grow to 4.5–5.9. The item editor stays just as fast as on the smaller dataset, opening in 130 to 180 milliseconds. Opening and editing individual products feels the same whether you have 25,000 or 100,000 items.

Average response times (ms) at 100,000 items, by tier. Lower is better.

Disclaimer

Actual performance on live installations will vary on many other parameters than the number of SKU’s in the database.

Examples of such parameters are:

• Number of concurrent users – many people working simultaneously will increase the general workload on the system.
• Automations – triggers, imports, or other scheduled actions can put extra load on the database
• Configuration – data from external sources, or complex hierarchies and business rules can require extra calculations before results are displayed

Regarding the configuration, and how to ensure it is done with proper attention to performance considerations, please refer to the performance guide:

Perfion Performance Recommendations