Understanding the GPU tool
Overview
A GPU (Grid Performance Unit) is equal to:
- 1000 GPUs is equivalent to running a continuous 10% load on one processor for the month.
- You can use up to 1.38 GPU's per hour and not go over the 2,000 GPU limit.
-
1 GPU = 7.24% of 1 CPU for 1 hour
NOTE: The above numbers are based on a 30-day calendar. These numbers will vary based on actual calendar days for each given month in your billing cycle.
Access the GPU report
- Log into your AccountCenter.
- Click on the Admin button for your (gs) account.
- Click on the GPU Report icon.
Details
Although the number of variables that can cause GPU usage are complex, experience has shown that major GPU drivers are usually caused from a traffic spike or server-side scripting.
Example 1
A typical very high-traffic site that is averaging just under the 2,000 GPU limit would look like this.
Example 2
GPU increase caused by a traffic spike can be verified by comparing your GPU graph to your Urchin graph. In the screen shots below you can clearly see that near the end of the month traffic to the site increased in parallel with the GPU usage and lasted about 5 days. This would be indicative of the site possibly being "dugg" or "slashdotted."
Optimizing your scripts can be a difficult task. The following are some ideas on where to look for help.
- If you are using third-party software, look into their documentation or their community forum.
- Try searching online using phrases like "php optimization", "optimizing wordpress", etc.
- Try disabling plug-ins or add-ons to your site. After a few days, you might see a reduction in GPU usage.
Example 3
The following examples illustrate high GPU usage along with some helpful balloon tips:
This index.php file is using over 87% of all GPUs: Consider optimizing and/or reducing the content of your homepage as this will most likely be your most frequented page.
Popular CMS software, e.g. WordPress, will dynamically auto-generate missing files such as favicon.ico, robot.txt, and any other missing files. This means the server needs to generate the file "on the fly" every single time the file is requested. It is much more efficient to create the file permanently. Serving a static file is an ideal way to decrease your GPU usage.
The aforementioned examples are just some of the common areas we see. Using the GPU report can tell you even more about your GPU usage. Please look for any highlighted areas and hover over our caution icons for further information. Please consult the following related Knowledge Base articles for more clarification if needed.