Testing new future hosting facility

Jorge, can it be a database problem? The actual site is pretty fast with active threads but damn slow if you call up older content.

Since PHP creates the webpages from a database, this can be caused by damaged indexes.
 
The site is fast, but the real test comes when 100 or so users login during primetime.

The "short-lived" host for RFF was fast early in the morning, then degraded quickly.

Honestly, Dave, I am experiencing a memory leak, but I feel much better now.

You think HAL runs Linux-9000?
 
The new site looks ok now but...

With a heavy evening load (most visitors seem to be EU or USA) it will likely fall over too in a few days or weeks.

Stop thinking 'hosts' Jorge and start thinking 'dedicated'.

In bricks and mortar terms: You have outgrown your existing premises and need to move somewhere bigger.
 
cant I just point my browser to 66.249.3.74 ? I get to Dslrexchange when I do that but I'm not sure that's right. Why not try and schedule a time (say 19h GMT on saturday) and makea sticky post about it to get everyone on that server at the same time to stress-test it?


edit: here is a site to convert times in case we need it http://www.timeanddate.com/worldclock/fixedform.html
 
Last edited by a moderator:
Stephans idea is a good one, I intend to release some sort of a distributed denial of service attack from four locations ranging from 1MBit/s to 6Mbit/s and throw in a search or something from my browser on a fifth line. This should eqal the load typicaly produced by 20 or 30 users.
 
Hi Jorge

For me (in Australia, 256 ADSL) the test site seems to be running at about the same response time as the current RFF site. The test site was showing 6 users.
 
The site loaded up pretty quickly for me, but it's early in the morning in North America so that might be the reason. Much faster than RFF this morning, but there are only 3 users right now. And I have 100 Mb/s pipe.

It's 9am EST.
 
Last edited:
Clicking on the "Using older prime lenses on DSLR cameras" thread on the test site had a noticible delay in opening. Since this is an older thread, could it be a database problem as Socke pointed out?
 
I do find it running appreciably faster (and I made sure it was the test server). However, page loads were slower than I'd expect of any server I was going to use, with two users shown as accessing the site.

Bottleneck is defintely not at my end - I'm sitting at the bottom of a 4Mb/s cable link, in Edinburgh, and elsewhere I can get the full bandwidth through this morning.

I would tend to agree with Jon - for a site this size, if you're not thinking about a dedicated solution, then you should be. The boxes you've been using just aren't up to the job. I'm actually surprised one of the hosts you've tried haven't suggested it.

Is it time to have another word with the sponsors?
 
Last edited:
Nomade, which OS are you using? We can take this to PM if you prefer.

William
(did tech support for Unix, MacOS and Windows for just shy of 10 years...)
 
Not knowing your system architecture, and assuming it's all on a single server with RAID, the first thing I'd suggest is a database diagnostic analysis. See if there are bottlenecks on the database side. If you're running something like MySQL, you may need more threading. Same with Apache. Plenty of RAM on the server? SCSI drives? Hopefully a Unix operating system (Linux and FreeBSD good for Intel/AMD). Windows servers can get a bit sluggish.

Gene
 
Well, I just rebuilt all indicies. Obviously wont do much good till they are cached. Heads up, the original RFF was on a solaris 8 dedicated system with 1 meg up and down . Now its running on a shared server with redhat with considerably fasyer network connection. I will call the provider and see if they can do anything for me.
 
Ok, when we're all outing ourselves 🙂

I'm a Unix and database consultant, I like Informix most but can get around with Oracle, Postgress, MySQL and MaxDB. Even Microsofts SQL Server if need be and the solatium is high enough ;-)

Luckily we don't have to deal with streams and kernel parameters as much as in the old days when I learned to get performance out of a 66Mhz Pentium with 64MB RAM running SCO Open Server 4 and Informix 5.
 
You're clearly tuning things and you've done something right.

This site (at 16.00 GMT+1) today is enough to keep James T Kirk happy. Warp factor five or so.

Don't touch anything.
 
Back
Top Bottom