Home » Use a Website Accelerator – testing automated web performance optimizer radware fastview

Use a Website Accelerator – testing automated web performance optimizer radware fastview

Summary:

Automated website acceleration products can speed up your site without developer intervention. We test the effectiveness of Radware’s Fastview appliance on a test site and find significant improvements in speed.

There are a number of automated ways to speed up your website. Web developers can deploy both software and hardware solutions to speed up their website automatically. By embedding best practices into code, vendors can speed up websites without the need to involve developer resources. This article evaluates Radware’s Fastview web acceleration hardware. Some other solutions are listed below.

Radware Fastview Web Acceleration Appliance

Based on Strangeloop’s Site Optimizer product acquired in 2013, Radware’s
Fastview optimizes web pages in real time to load faster, using best practice
performance rules. Fastview automates many of the front-end optimization techniques web site optimizers
have been doing for years, including:

Resource consolidation
Combine JavaScript and CSS files to reduce http requests
Payload reduction
Using GZIP, minification, and lossless image compression
Automatic CSS sprite creation
Chunked encoding
Servers can begin transmitting dynamically-generated content before knowing the total size of that content.
Domain sharding
Boost parallel connections
Caching
Proper caching reduces bandwidth usage and speeds repeat requests.
Predictive browser caching
Preloading likely resources for subsequent pages (adds to browser cache)
Landing page vs. repeat views
Inline resources versus use cached external resources.

Deployment Options

There are three deployment options for Fastview:

  1. Appliance – network appliance installed within the companies network
  2. Virtual – a virtual machine as software installed on the company’s server
  3. Cloud service – traffic is sent through a fastview instance external to company server

Testing Fastview Performance

We tested the Fastview hardware appliance on a client’s site with their permission (who shall remain nameless). The untreated page had 30 external JavaScript and CSS files before the page started to render (see Figure 1). The untreated page starts to render in 1.73 seconds, loads in 6.5 seconds, and fully loads in 8 seconds (see Figure 2).

waterfall-notreatment

Figure 1: Untreated Waterfall Highlights

timings-no-treatment

Figure 2: Untreated Page Timings

The Fastview treated page starts to render in 1.6 seconds, loads in 4.07 seconds, and fully loads in 9.3 seconds (see Figures 3/4). The start render time was 8.1% faster, and the load time was 37.2% faster.

waterfall-treatment

Figure 3: Fastview Treated Waterfall Highlights

timings-treatment

Figure 4: Fastview Treated Page Timings

Some Observations

Note the difference between document complete objects. There were 72.6% fewer objects loaded before the onload event in the treated page (117 vs. 32 objects). Also note the reduction in bytes downloaded before the onload event. The onload payload decreased by 39.2% from 1043K to 634K. Other measures also improved. DOM content loaded decreased from 3.19S to 1.77S a decrease of 44%. Google’s Speed Index decreased from 4.086 to 3.606S, a decrease of 11.7%. Recent Velocity conferences emphasize these latter measures over the onload event, due to the increased use of AJAX and JavaScript.

The fully loaded statistics increased by 19.1% for objects, 30.8% for total page size, and 16.3% for full loaded time. However, with this acceleration product the user experiences a page that displays content and loads faster, what happens after the onload event is less important for the user experience than what happens before.

Conclusion

Automated website acceleration software and hardware can make significant improvements in website speed. In this particular case we found an 8.1% improvement in start render time and a 37.2% improvement in onload time using Fastview. The number of objects loaded before start render and the onload event was significantly reduced, in this case by over 72%. Radware says that a 20-25% improvement in page loading speed is a typical floor, with more savings possible the more complex the website.

Further Reading

Leave a Comment