Daley and I are at odds about many tech topics. His points are always reasonable, even if I don't agree or share the experience. So it goes. I think his field of expertise is in IT, but I can't be sure. I just build shit.
Yup... heavy background in IT/IS, networking, (and VoIP at this point)... but I also do some web development and general server ape services as needed. There's muh background. I personally don't mind the disagreement, it keeps things interesting... and I think we'd have probably hit things off better if I didn't mistake him for a marketing shill with his first posts to the forums here.
I am not sure why Daley said that 12 database requests and 4.5MB of data per second is an issue. Most requests are dominated by seek time (that is, the amount of time it takes the hard drives to physically find the data), and 4.5MB of data per second on even a 1 gig backplane is trivial... and these days, even cheap servers have multiple 10 gig ports, each of which gives about 500x the bandwidth 4.5MB/sec would need. Which is why I can transfer large files at speeds limited by my network connection, not theirs - they have way more bandwidth than you would need.
You are absolutely correct in this assessed statement, and there should be
zero problem handling 12 database requests and 4.5MB of data a second by
any reasonably equipped server. These are reasonable numbers, even potentially for SQL server loads... the problem are the statistics that accompanied those throughput numbers:
one out of every 78 SQL connections are timing out, one out of every six queries served are slow. The failure and slowdown rates indicate a taxed server that cannot handle the loads it is being given. The site in question had a small Wordpress install with this (far smaller than Technical Meshugana). Low traffic, but the SQL server issues were causing regular problems loading the site. It got worse when I introduced something with
a lot more content and a database larger than 4MB.
It gets worse when you compound the fact that most of those other 2000 some-odd websites are frequently using database resources as well, and appear to primarily be SEO spam and malware honeypots designed to try and drive traffic... layer in an increasing stream of abusive and malicious botnet traffic on nearly every domain - especially on CMS installs looking for weaknesses (and so few people knowing how to eliminate that traffic), and it gets a bit ugly on anything but raw file data throughput... but even that, I've found they have a nasty habit of throttling traffic exceeding 1GB in a 24 hour period. This also indicates a management issue, however, and caring more about paying customers than
what is being hosted.
As a point of raw numerical comparison to the Bluehost logs in question from what I linked, Bluehost's SQL failure rate was 1.28% of all queries (which alone doesn't mean much, malformed queries get tossed into this number), and 16.667% of all SQL queries were slow (one out of every six), which is a sign of server load issues. The slow queries are the truly problematic ones, especially when they're so slow they time-out and fail, and it's important to remember that stock Wordpress (IIRC) runs at least 18 queries per loaded page. Their SQL servers on their shared hosting also ran through approximately 16.3 GiB of total traffic and ~43,150 queries an hour. This SQL server was being restarted about once every 2-3 days. I found similar statistics on other Bluehost hosted Wordpress installs on different shared hosting servers.
Currently with Technical Meshugana on the shared hosting plan I'm using with ASO (their Small shared hosting plan - $5/month plus further discounts on annual purchases and coupon specials), the relevant SQL stats are as follows: failure rate of 0.44% of all queries (429,100), and only 869 slow queries
total, which works out to something like 0.000009% of ~97 million total connections. ASO's SQL server these numbers are from push 15.6 GiB of total traffic and ~49,990 queries per hour, and as of this message has an 80 day uptime which coincides with a maintenance event.
You'll note, similar throughput on shared servers between the two providers, but the difference is staggering on the slow queries end for SQL performance between the two. You can have as fat a pipe as you want, but when dealing with scripted content, the SQL server makes and breaks load times. I'm paying less than $5/month to ASO for a plan that has plenty of bandwidth and storage space, and the SQL server performance (not to mention the technical support) is like night and day compared to Bluehost, which costs a fair bit more for that promise of "unlimited". Management goes a long ways, as Fantabulous has pointed out... and it bears out in the numbers between the providers discussed by myself.