I guess I just need to know why I never get “timed out.” It may take a while, but I don’t get “timed out.” And I’m connecting to the same server you are.
Just in case it matters, I use IE.
Yeah, yeah. I know. Stupid. But it works for me. I keep it clean.
I’m using the very latest version of IE6 on this box. Thus the reason you don’t get timed out has nothing to do with your choice of browser.
This is just a WAG about something I think I’ve noticed.
If I click on a thread in CS about the first ep of the season of Lost (a thread that almost always have a lot of traffic), I notice that there are new posts since last I checked and the last post is by Dex. The browser doesn’t start to load the thread, but instead (IE6) I get the progress meter chugging along at the bottom.
Almost always, I’ll end up getting a blank page “server not found”.
So I hit the back button and try again. This time it zips along and I get to see the last post. Which isn’t by Dex.
Conclusion: If I try to open a thread at the same moment the database is being updated with a new post, the hamsters can’t handle it.
I have used IE7, IE6 and Firefox, and I have had time-outs on each. But going on with The Gaspode’s comments, I have found closing the tab and opening a new one works rather well in speeding things up. I also started to compose my posts in notebook to save those ate by the gerbils.
I have observed the same thing. I type longer post in Word first anyway so I have a backup of post of the post even if IE explodes as it sometimes does.
Back again …
First thanks to those of you that complimented my post. A quick clarification, I don’t think we’re in any way entitled to see the costs and specs of the hosting. It’s simply that without that information we are reduced to guessing about what is and isn’t possible. We’re just trying to help. I’d love to just be able to see the bandwidth, min and max users / page views etc just to play with the figures a bit.
But anyway … timeouts. I’m sorry for those of you who already know this but I’m going back to basics and building up. It’s been a while since I’ve worked in the web/databases area so if I screw this up please correct me.
When you click on a thread / post / whatever on the SDMB what happens is your machine first contacts the server to establish a connection (like making a phone call as squeekster says). Once the connection is established your machine sends a request to the server “Gimme page X please” and the server sends back a response “Page X”. Now Page X may require some other parts (graphics etc) which your computer might need to request from the server, but we’ll assume here that you already have these cached locally so it’s a simple transaction. After the servers sent page X it should hang up.
So the transaction time to get a page is made up of several steps:
[ol]
[li]Time for your connect message to get to the server[/li][li]Time for the server find and allocate you a connection[/li][li]Time for the connection details to be sent back to you [/li][li]Time for your request (gimme page X) to travel to the server[/li][li]Time for the server to process the message and build page X[/li][li]Time for that response (page X itself) to travel from the server to your machine[/li][/ol]
At step 2 the server needs to have enough spare time to process your connection request. Otherwise you’ll get refused, or more likely put in a queue. Even if it has enough spare time there’s a limited number of connections that it can have open at once. Depending on when the connection is close and how long transactions take some connections can be open for a while. So either server load or a connection limit will define a limit of maximum simultaneous page requests.
At step 5 the server needs to build the requested page (perhaps posting a message first before building the thank you page). I suspect there will be some jiggery pokery done so that the server doesn’t have to go to the database for simple views on popular threads but a lot of the time the database gets accessed. And threads with heavy posting will likely be slower as their bits of the database get ‘locked’ during the updates thus delaying everyone elses viewing / updating.
Since this should all work on a first come first served basis you’d expect to see everyone who’s connecting at a certain time experiencing the same set of (intermittant) problems as the hamster struggle to build everyones page.
You’re only likely to see these during peak time, I browse during UK business hours (errr, yeah I should probably be doing something more useful, like working) and almost never have a problem. It’s noticably worse in the evening when the US wakes up.
If some people are browsing during peak hours and not seeing timeouts then it probably means that the server is generally responding in time but not quickly. This means that the transmission times (steps 1, 3, 4 and 6 above) come into play. Typically the round trip shouldn’t take long (less than a second easily, nearer a tenth of that) but that can be affected by lots of things in between.
Of course the transmission time will be dependant on the amount of data and the smallest ‘pipe’ that it has to pass through. The server could be hosted a major internet ISP with a great connection and it’ll matter not a jot if it’s connected to the ISPs outgoing connection by a slow internal network (I’d be shocked if that was the case) or if there is a similar weak link anywhere between it and you that can’t be bypassed.
So it could be a combination of several factors causing the apperance of lousy performance. No question making the server faster would help but if it’s a issue of how the traffic between the SDMB and your provider is transferred then it may not help much.
The final twist is it’ll depend on your browsing habits, some threads will be faster to access than others. There’s not enough information to make a guess which way but I’d say that people who are ‘View Latest Posts’ browsers will experience more or less problems than people who are ‘mooch through a forum’ browsers. It’d depend a lot on how things are configured.
All this leads us to not really knowing. It could be the server. It could be your internet connection. It’s probably a combination of both.
Eeek, I’ve rambled on again and made another huge post. Hopefully it’ll be useful.
SD
You know… I really don’t get this. We pay our money to get a service, we’re aware of the quality of service when we pay our money. We get this service.
If you don’t like what you’re getting, don’t subscribe again. I don’t post a lot and could let my membership lapse and still read the board… which is what I do mostly anyway. But I remain a member… I’m sorry, but this does not give me any right to know why the Chicago Reader’s business model is set up the way it is, it does not give me the right to know why the company put us on the server that we’re on. As long as I get the service I paid for, I have no room to complain.
I do get that service and, therefore, have no complaints.
This is concern because you can on one hand simply rent a box and a net connection, ending up doing all the work to keep the site running yourself (ref. Jerry’s story), or you can rent a box from a company who takes the responsibility of keeping core application up and running. The latter costs more, but is in my experince cost-effective in the long run.
Shalmanese, nice find, I’ve been looking for something like that. Thanks.
I hate to rub salt into the eyes of other people, but making decent money on the net works with targeted trafffic only. Most general purpose boards have a hard time breaking even from ads alone. Affiliate marketing is an option, but the concept is usually tied to some sort of product recommendation (or price comparison engine). Yes, some money could be made but there’s an administrative side to it. I’m not even sure if AdSense, usually a safe no-hazzle route, would work on this board (if the powers allowed the spider in in the first place).
And also thanking Una Persson for providing her stats. This will help me out in a personal project. 
I ran this by a sysadmin I know. When making a connection, after a certain amount of time if the server at the other end just doesn’t communicate back, the ISP will break down the connection, and “hang up the phone” itself. This is because the ISP server can only have so many connections open at once. If the ISP didn’t do this, for someone on DSL or such that is always connected to the ISP, if the web server at the other end totally died the ISP could literally have that connection open forever waiting for a response, and wasting its resources.
The reason why some people time out faster is some ISPs will hold the connection open longer than others. Thus some people will time out before others.
It has been pointed out the Reader thinks of www.straightdope.com and the SDMB as being part of the same whole. Thus if the ads were put just on Cecil’s columns, this would bode well for the SDMB. Also, nobody is suggesting eliminating paid subscriptions and making the board free. Even if the SDMB brought in minimal click through ad revenue, this would be “icing on the cake.”
On DSL it’s also a matter of re-negotiating the handshake when the telephone box closes the connection or the router on the client side goes into sleep mode due to inactivity. Also, DNS resolution might be a problem on the client’s computer.
Personally, I have noe problems with the board, though a few more features would have been nice. The board can be a bit slow during lunch hours in America and likewise in the evening, but not so much that it bothers me. It’s miles ahead of what it used to be. I do have life offline ya’know (or at least I pretend that I do … ).
I agree. But I’m also dead sure that the administrators of this board had a long and thorough discussion about how to best create needed revenue before they decided on a pay to post model. I’m not going to second guess them witout any kind of info about traffic, demographics or principle. Whether that information will ever be provided is another matter.
By the way rfgdxm, care to finally tell us how you know that $200/month is enough to run this board when you don’t know how much bandwidth and how many page views the board has? I don’t know about the others, but I’m still waiting.
My guess is was more the Reader than the listed admins here who made the decisions.
http://boards.straightdope.com/sdmb/showpost.php?p=6499218&postcount=2
Note in this post TubaDiva states whe doesn’t even know who paid what, and thus what actual revenues are. And the admins say they don’t know the server specs, or hosting costs. The only people who really know are those running things at the Reader, and they aren’t talking.
Reasonable inference based on what is known. The number of users is known, and also the typical page size. It isn’t that there are huge binary files available to download that could mean possible huge GBs consumed by one user.
On this board you will have to do better than that. Put forward you estimate of bandwidth, follow up with a cite on how much that amount of bandwidth would normally cost, reasonable levels of support included. I’m not asking to mock you, I’m actually genuinely interested.
(btw, I haven’t read your other link yet)
Scrap that, your link points to page one of this thread. It doesn’t say what you claim it says at all. It’s a reply from an SDMB administrator saying she personally doesn’t know ‘who paid what’ because ‘it’s not my department’. That was a new low in this thread.
http://servers.aplus.net/premium.html
Dual Intel Xeon D 2.8 GHz, 800 MHz FSB with HyperThreading
1024 MB Dual Channel DDR400 ECC RAM (upgrades up to 4GB available)
120 GB hard drive (upgrades up to dual 250 GB available)
1000 GB Bandwidth included (Non-Cogent)
Free Managed Services
$199 a month
Would you consider this adequate for the needs of the SDMB? If not, cite why, and what specs would be required.
rfgdxm, that certainly seems to compare well with the hotdogger site’s server Shalmanese linked to in my amateur eye. From reading his link it looks like “Overclockers Australia” require a lot more power than the SDMB, as well. Wouldn’t be surprised considering their name, heh.
Oops, meant to include link to Shalmanese’s post, two pages back.