Shootout at Mortality Test Stats?

2WS-Steve · 10th May 2002, 03:30 AM

Hola folks; I was just hoping that we'd see a post (or maybe something on the main page) about the stats for the server-meltdown experiment. Things like total threads, page-views, posts, searches (if they can track that), highest server CPU temperature hit, etc.

**Morrus** · 10th May 2002, 03:39 AM

I have yet to talk to Adlon about it. All I know so far is that the server bore up, but really showed the strain. That said, it was dealing with about 800 new registrations in one day.

Adlon, at least, knows that he needs to upgrade his server sometime soon.

2WS-Steve · 10th May 2002, 04:26 AM

That was a great test. I thought it held up pretty well all things considered; people doing searches for the 1000 most popular words and all...

If that method does solve the problems it'd be a good thing. Mortality's radio interviews are great and with a bank of high-powered, handcrafted, gamer-maintained servers backing the site ENWorld could finalize it's plans for multiversal domination.

Quickbeam · 10th May 2002, 04:39 AM

It really was a great idea. There were a few kinks, and the site definitely began to sag under the weight of all our traffic, but overall a fine experiment!

**Adlon** · 10th May 2002, 05:01 AM

Heya guys. Good job today. You all kicked my sites ass, and even tho it buckled, it hung in there. A few sqo errors, and SLOW as heck, but, good job.

IT tells me AND Morrus a few things.

IT tells me that I need an upgrade for Mortality soon.
IT tells Morrus that he does not need a RAID array for his server.

Simply, a good strong server, with a good bit of RAM, and a strong forum package.

For your information, I will now divulge my server specs here:

CELERON 450 Mhz
128 MB PC-100 SDRam
15 Gig Fujitsu 5400 RPM HD
Any old Motherboard I had laying around.
10/100 MB NIC (Realtek 8139 series)
250W Power Supply

OS: Linux Slackware 7.1

Apache, mySQL, and PHP all the latest versions.

My bandwidth: Dual load balanced T1's, with a 10MB Connection to the LAN/T1's, and about 350 feet away, is the OC3 at the Local Telco. This is a Sprint backbone, peering with UUNET in Atlanta.

I must say, I'm proud of my little server. And, I have about 35 webhosting sites on it, 10 of which are busy, 3 of which are
E-Commerce sites. Plus, with all you guys there, you ALL had to register, and the registrations were essentially all at once.

Some of you had validation link errors, but, for the 500 + people who signed up today, I only got about 10 or so emails, and Morrus got about 20 or so, so, the forum package I'm using is not too shabby either, tho I am going to go to another system soon.

I will be getting with Morrus, and seeing if we can hash out a solution. The forum pacvkage I'm going to is viewable at:

Again, thank you all. You've shown me that my little box is tough for the 450 Mhz of RAW processing POWER

that it is.

You guys may feel free to visit again anytime, tho please, for awhile, NOT ALL AT ONCE

EricNoah · 10th May 2002, 05:27 AM

I also thought it was a very worthwhile test -- gives me confidence that a solid, reliable solution is out there somewhere.

Mark · 10th May 2002, 06:54 AM

I think we gave it a pretty good work out, for sure!

Sounds like it showed you what you guys needed to know.

Thanks for putting the plan together!

coyote6 · 10th May 2002, 07:04 AM

Mark, did you ever get through the 1000 words?

Omegium · 10th May 2002, 08:41 AM

Adlon, you should be proud for certain on your server. If I recall correctly, the servers the ENforum runs on are faster. What is the reason for that? Is it because you are running other forum software, or because you are running linux?

Liquide · 10th May 2002, 11:14 AM

Quote:

Originally posted by Omegium
Adlon, you should be proud for certain on your server. If I recall correctly, the servers the ENforum runs on are faster. What is the reason for that? Is it because you are running other forum software, or because you are running linux?

because the server the forums run on here have something around 2-3 request / second or maybe more if we count the banners and external systems.

Make that an even 4 requests per second to make sure, when the boards are at peak usage during the day.

Zappo · 10th May 2002, 11:58 AM

Quote:

Originally posted by Liquide
because the server the forums run on here have something around 2-3 request / second or maybe more if we count the banners and external systems.

Make that an even 4 requests per second to make sure, when the boards are at peak usage during the day.

You mean, counting the website and hosted sites too?

**Adlon** · 10th May 2002, 12:06 PM

Well, it seems that my DB took a dump on me, de to the mass registrations we've had yesterday.

So, the forum package I use itself was not necessarily under observation. The point of the test, was to look at a whole possible solution. Not just a server, or software.

I will get my DB unborked today. And I wonder how ANY package could handle that amount of mass registrations, truth be told.

We had approx 750 people regiser, most of which came in over 3 hours, and many were simultaneous.

I do believe, tho, that the test showed there IS a viable solution for the forums here.

After I get MY forums unborked, I'll get with Morrus.

**Adlon** · 10th May 2002, 12:53 PM

Mortality Update:

Seems we in our haste to recompile PHP before the onlaught, we left ONE critical setting OFF in the PHP.ini.

This critical setting, once turned on, fixed our forums instantly.

So, we can possibly guess that they would have performed better than they did yesterday, the delimiting factor still being the power of my server.

So, this news makes the test yesterday even more successful.

Quickbeam · 10th May 2002, 01:55 PM

Sorry to hear that an important setting was left OFF yesterday. But as everyone has noted, with the exception of a few activation bugs and slow movement in loading, your server held up well.

Thanks for letting us utilize your site as a testing grounds!!

Psychotic Dreamer · 10th May 2002, 02:15 PM

I was very impressed with the test yesterday. Near the end I started getting a few database errors, but a reload fixed everything each time I got one.

**Adlon** · 11th May 2002, 12:04 AM

Acknowledged.

One setting, and we may have had even better performance, certainly not any worse.

My initial thoughts are the fact that I run Linux. Now, any MS guys out there, I mean no ill will. Remember, Adlon uses MS stuff to make his day happen

But, I run Linux, and Apache web server. Having recompiled Apache 2, and the latest PHP, I see a performance increase based on our regular traffic.

So, I can say that there is definately an issue with an OS, and perhaps configuration, to give vBulletin credit. I've messed with vBulletin, but not much.

I'll email Morrus after the Radio Show tonite, and me and him need to chat.

reapersaurus · 11th May 2002, 07:09 AM

what was the impressions from the test?

As far as _I_ saw, the boards were incredibly slow.
So slow that it would really make browsing them a painful process, IF that performance was continued.

It sounds like things might go faster now, especially if it's upgraded from that dog 450 MHz processor and skimpy 128 MB of RAM

but remember, if the whole boards descended on that server, it would be much more bogged down than what it was that one day (would be my guess)

**Adlon** · 11th May 2002, 01:58 PM

During the test, being SLOW wasnt a bad thing. I expected my server to buckle, and bend. And I expected it to not break, and it didnt let me down.

Any solution for these forums would involve a much more muscular machine than mine, no doubt.

But remember, presently, if Morrus opens up the user limit, after about 200 users, the forums totally die, per Morrus.

The test was simply to look at the traffic in a different environment.