Shared Web Hosting Performance

[jammybiskit]

I've looked around for a post addressing this kind of issue but failed to find one and so I apologise if this is posted to the wrong forum - please move it if there's a more suitable place.

I am looking for advice/recommendations on shared web hosting providers.

We have a small retail website on a (popular) shared host. The site is mainly dynamic (using Perl/MySQL) but its home page is static. We get around 150-200 visits a day according to Google Analytics.

A while ago I began to get worried about the subjective response times we were seeing and I set up an experiment to monitor response time. We use two different hosting companies for two variants of the site (.com and .co.uk) and so I was able to write a program to access one from the other and record the results.

The monitoring program runs hourly and records the time taken to retrieve

1. the Google front page (for reference);
2. a static site page;
3. a typical site page (with a few SELECTs).

Only the html page text is retrieved - a real visitor to the site would see a slightly longer response than recorded as css file, images, etc. have to be transferred and rendered.

I've been monitoring for about a month now and the results are:

Worst Response Times
Google 1 second
static page 181 seconds (yes, 3 minutes)
dynamic page 80 seconds

Best Response Times
Google 0 second (i.e. 0-1)
static page 0 second
dynamic page 1 seconds (i.e. 1-2)

Average Times at 16:30UTC
Google 0.25 second
static page 0.89 second
dynamic page 5.36 seconds

The results show that it isn't a site code problem (if it was then the minimum times would be longer) but that the host performance degrades rapidly with load.

We feel that most users, especially broadband users, will not hang around for much more than 4 seconds and so we are losing traffic as well as user confidence.

Can anyone recommend a (shared) web hosting provider who is prepared to provide some kind of service level agreement not simply for uptime (they all claim nearly 100%) but for actual response time or some equivalent measure?

All other suggestions are welcome!

[J_Paul]

I would recommend going to http://rochenhost.com

I have been hosting with them for years. As a designer i'm able to set up multiple domains for my clients and the server is shared, but they give what they say they will give.

In the past 5 years i've only gone down once or twice and for minimal time.

Check them out.

Oh, in case you need to know.. my sites average about 2k a day for 2 of them... so i know how valuable response time can be!

[jammybiskit]

Thanks for that J.

I'm not quite sure what you mean by

Oh, in case you need to know.. my sites average about 2k a day for 2 of them... so i know how valuable response time can be!

Does the 2k refer to visitors, or dollars, ...?

I've done a little more monitoring of my own provider (who now claim to have fixed their performance problems).

I was able to add a friend's hosting provider to my monitoring list.

The results apply to monitoring since May 5 (when the problems were supposed to stop).

Percentage of queries taking more than 4 seconds
Google: 0%
My shared host: 7%
Another shared host: 0.4%

This is the kind of information I'm interested in - more than 4 seconds means probably losing a visitor.

[hostBrain]

I'm not sure if google is the best page to gauge results from, since the page is so "light".

You say you use 2 different hosting co's for each domain. Are the 2 domains the same site?
Did you do testing for both?

[jammybiskit]

I'm not sure if google is the best page to gauge results from, since the page is so "light".

Yes, you're right - it is short on bytes. It's really there to check that my monitor machine isn't suffering from some sort of interface problem when each test is done. The test URLs are read every hour. If it took, say, 30 seconds to read each test page then I would be suspicious that there was a problem with the monitoring machine or its connection to the net. (This hasn't happened yet).

You say you use 2 different hosting co's for each domain. Are the 2 domains the same site? Did you do testing for both?

I have been monitoring my own website's host and also the (different) host of another website (as well as Google for reference). The monitor reads a static page and a dynamic page from each host - to indicate both web server and database server performance.

The aim is to see just how often my shared host takes an excessive time to serve a page. By excessive I mean long enough that a typical visitor will give up and go somewhere else.

Sadly, this has been happening 7% of the time recently - so I expect I'm losing that kind of percentage of visitors.

The thing is, the uptime of my host has been 100% in the same period - so uptime doesn't tell you what you want to know about hosting performance.

I guess this is a loading issue - too many websites sharing the same hardware. The other host is doing a lot better - wish I'd chosen that one!

[Vectorman211]

It's very possible that your webhost is overloaded. Using Google.com as a reference is silly though. Google has the most advanced servers in the world and virtually unlimited bandwidth, moreover, their front page is insanely simple.

I use 1and1.com for web hosting. There are about a billion other webhosts out there. 1and1 is $3/mo with loads of features.

[webmaster@vzmedia.com]

I have found that there is some correlation between cost and performance but this is not a gauranteed thing. I guess if the host charges more, they can afford to operate better if they want. That is - low cost gaurantees limited performance, higher cost just makes it "possible" to have better performance but it depends on the hoster if they spend the money on performance or parties.

That said, I think you need to consider (at least) the following factors:
-Performance (mainly page-load speed)
-Reliability (mainly up-time)
-Support (what happens when there is a problem)
-Features (bells, whistles, as well as capacity- MB/GB available, email accounts etc)
-Ease-of-Use (this is really quality of features, important to non-tech clients)

It would be great to have a site that ranks and keeps tabs on these factors for multiple hosting services. (Evaluating or voting is often a problem, but that is another thread) I have looked, but haven't found anything yet. Is there such a "consumers-guide" site for hosting that anyone can refer us to?

Thanks,
John B.