You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 3 Next »

This document describes the results of performance tests for WebSpellChecker Web API. The performance and load were tested and accessed depending on the following setup:

  • Certain number of users accessing the server simultaneously (10/20/50/100 users);
  • Cache enabled for spell checking purposes; 
  • Certain hardware and software used (EC2 m5.large instance with 2 CPU and 8 GB RAM); 
  • Number of words to be checked (10K words or 6K characters);
  • Number of spelling and grammar problems in the text (50 grammar problems | 200 misspellings);
  • Type of the language used for check (17 default languages). 

Testing Goal and Idea

Our main goal was to observe the response time of text processing and CPU utilization on the server in case when 10/20/50/100 users send simultaneous requests on various languages to the server with WebSpellChecker v5.5.9.

Environment and Testing Tool

Testing Process

We have run our tests continuously increasing the number of users accessing it for each of the languages in the default language group (17 languages). The cache setting was enabled for all set of tests. The tests took place in the following order:

  1. 10K words (6K characters) with 50 grammar problems 200 misspellings;
  2. 10K words with 50 grammar problems only;
  3. 10K words with 200 misspellings only.

We used the concept of tokens which is a complete sentence to be spell and grammar checked. In the test setup, we had 10 tokens. 

The measured was the response time and CPU utilization.

Observations and Findings

Our observations are presented in tables and charts below for WebSpellChecker Server v5.5.9.

  • The response time and CPU untilization increase as more simultaneous users are added.
  • These two metrics depend a lot on the number of words in the dictionary for spelling check and number of rules for grammar check.
  • The average time for processing of 1K words containing words in mixed case and misspelled words only grows 15-25%, 
  • If users need to proofread 1K words containing only grammar problems, and there are no misspelled words in these sentences, it will take more time, namely 27,777 seconds in version 5.5.4x compared with 46,551 seconds in version 5.5.5x. For details, see section 1K words with 50 grammar problems only depending on the number of users below.

1st Test Case for 10 Simultaneous Users

Chart below represents test results comparison for WSC v.5.3 and WSC v.5.4 and 10 simultaneous users with disabled cache.

Recommendations 

Here are the outcomes and aftermath as well as our advice on hardware and software requirements and notes on performance issues which users may encounter:

  • General performance of our spell check engine has increased, but grammar engine performance is not as high as expected. For details, see the charts showing test results depending on the number of users. For example, 1K words (6K chars) with 50 grammar problems 200 misspellings and other graphs in this section.
  • One m5 instance can process 150-200 simultaneous users, or simultaneous threads, without any issues, but when the number of users increases to 200+, it entails 100% CPU load and a significant increase of response time. Our recommendation for the case when more users are added and CPU load constantly reaches 100% on the machine:
    • upgrade instance type and add more CPUs to it;
    • add one more machine to distribute the traffic (requests) between two or more machines, for example, using load balance and auto-scalling.

When cache is enabled, tests run much faster, and the results are almost identical for different cases due to the processed texts are the same. This case needs refinement of the text uniqueness or some text randomising for each request being sent. Our recommendation for this case is the following: specify the desired value of the CacheSize parameter in AppServerX.xml file to increase the speed of requests processing.


  • No labels