CyberKnowledge Technology Blog

Everything tech – reviews, tips, software, news

All tech all the time

Welcome to CyberKnowledge technology blog.

Uptime isn’t everything: Analyzing page load time

Posted by Alex Bailey On September - 17 - 2014

When picking out a web host there is more to it than uptime and how much bandwidth and disk space they offer. The biggest factor is often times the hardest to discern before actually registering an account. It’s so important that it will even play a large role in how well your websites will rank in Google. That’s right, it’s your website’s response time. However, there are a couple things to do to get an idea of how well your new server will respond. There’s also some things you can do to protect yourself if you got it wrong. Let’s take a look at those things, as well as analyze some data at just how important your website’s load time actually is to search engines.
For the past six years I’ve been using Hostgator shared hosting to put my 20 or so websites on the internet. The customer service was reasonable and the uptime was fairly good as well. However, during my last few months there I started to notice a lot of jitter. Some page loads would take two seconds, while just one reload later it would take 6-8 seconds or even more. It experienced a lot of random high load times, even at low traffic times. Finally I had enough and made the switch to a different hosting provider.

The Hardware

One of the major things I was looking for in my new web host was SSD hard drives. It’s 2014 and when I came to Hostgator way back in 2008, SSD was not really a thing back then. An SSD is going to pull files from your database far faster than a HDD, and be able to handle a far higher load.
I looked at many hosts but eventually fell towards Asmallorange. Their shared hosting packages were fully driven SSD raid 10 arrays (both data and database on SSDs). Compare that to Hostgator who only have their MySQL databases on SSD, and they are on Raid 1. The difference between raid 1 and raid 10 is going to be night and day. Raid 1 is a simple mirroring of two or more drives. There’s no performance gain at all. Raid 10 stripes data across multiple drives, while also mirroring those two drives. This means that the server will simultaneously write data to two drives at once and then mirror that data across two more drives. This theoretically allows for much faster write times than standard mirrored raid.
The next thing I compared was processor speed. HostGator shared servers are on AMD Opteron 6376 processors while Asmallorange are on Xeon E5 latest generation. Again, benchmark tests between the two are a night and day difference. Even a low model E5 is going to outperform an Opteron. The distinction between these two companies just became even more obvious.

The Results

The end results between hosts are amazing, and Google agrees. My average HTTP response time went from 3,000 ms to 700 ms across all my websites. Page load time halved or more across all my websites. And Google experienced very similar results.
Page response time
The above image describes exactly what I’m talking about. Before the drop off is what my response time was on Hostgator. It was random, it was high, it was medium, it was all over the place. One thing it wasn’t was low. It never made it under two seconds. Immediately after I switched it leveled off and remained extremely consistent – between 700-800ms. That’s a far cry from 2000-8000 (high results not shown because Pingdom calculates averages between attempts).
So now the SEO in you is probably wondering how this effected my Google crawling. I have to say, the results have been fantastic. Since the switch, downloaded data has been up, while time to download that data has been down.
Bot's time spent downloading pages
As you can see, the average time spent downloading a page before the switch was jerky and high, just like the response time. Google prefers sites that respond faster, so that jerkiness definitely hurt my SERPS.
Amount of data downloaded per day
And finally, this graph tells it all. The amount of data Google pulled from my site in a day skyrocketed. The number of pages Google crawled on my site also skyrocketed. This tells me that my website is now responding in an optimal amount of time for Google. Not only did Google spend more time on my website, but it also crawled more pages and downloaded more data. The results were so dramatic that I’m kicking myself for not doing this sooner.

Making sure your host preforms

While there’s no definite way to make sure your host will have good page load times, there are a few things you can do to get an idea. The first and most obvious is to make sure your websites will be SSD driven on optimal hardware.  The second is to simply check the page load time of your host’s main site. Did it load under two seconds? If they can’t make their own homepage load in under two seconds, don’t expect your website to. However, that’s not full proof. Often times they will stack the deck and give their homepage an advantage. Hostgator’s main website loads in around one second, while my website’s hosted with them will load in three seconds. Remember, their site is a dedicated server, likely running on the best hardware they have to offer.
Really the only thing you can do is make sure your host offers a refund policy of at least 30 days. During this time, you can use a website monitoring service like Pingdom to track your server’s speed. It will let you know about downtime and also give you a graph of your website’s response time. Don’t like what you’re seeing? Cash in that refund.  Pingdom has amazing tools for webasters, particularly their uptime monitoring tool.  This service sends a ping to your website every minute to check the response time, and see if it’s up or down.  Downtime isn’t a major concern for those of you who have websites in the cloud, but for traditional hosting, downtime does happen.

Final Thoughts

Unfortunately it’s pretty hard to confirm your new host’s response time without actually purchasing first. However, there are a few things you can look at to make sure your websites will have the best chance at low load times. Confirming your web host utilizes SSD with a striped raid is definitely a good metric for judging a host. Don’t fall for the unlimited bandwidth and unlimited disk space game. Often times companies offering unlimited everything sacrifice quality for quantity of users they are signing up. All of the major web hosts like DreamHost, BlueHost, and the rest have these offers. Well guess what? You can’t offer unlimited disk space for $3 a month and still provide a good service. It just doesn’t work like that. The age old saying “you get what you pay for” still applies to the technology industry.


Server

One Response to “Uptime isn’t everything: Analyzing page load time”

  1. Barry says:

    I can’t agree 100% that website uptime doesn’t matter. Some of the site’s downtime can be invetable and watching the website uptime is really crucial. As for page load speed, surely it is important, but I would put it in the second place. When you work with a site, you should be ready to work with everything. Monitoring services are critical and inseparable parts of the process. You named Pingdom. The tool is not bad, but it is expensive. There are cheaper variants like Anturis or even free ones like Nagios.

Leave a Reply