forums.uesp.net

@rpeh

Heh. You're right! Sorry, I should have said BST (British Summer Time), not GMT. In Winter, UK is on GMT and Montreal is on EST = GMT-5. In Summer with DST (Daylight Savings Time) of +1 hour applied, UK is on BST = GMT+1 and Montreal is on EDT = GMT-5+1. With or without the +1, the time difference is STILL -5 hours.

Anyway, on to problems. They don't occur every day but, when they do, the timing seems to tie in neatly with the pattern of typical UK/US daily Internet activities. Take a look at the table below, in which I've postulated a typical day (start work 8-9am, lunchtimes ranging from 11am-2pm, finish work 5-6pm, and leisure from 6pm-midnight). I've weighted the typical Internet activities and spread them across UK/US time zones. I've then added up the weights for each hour and the results reproduce almost EXACTLY the timing pattern I'm seeing!


As you can see, US activity overlaps the UK's from 1pm to 8am (19 hours) with US lunchtime/evening peak periods having greatest effect during UK times 6pm to 9pm (3 hours) and 11pm to 6am (7 hours) respectively. I've noticed that, when UESP timeouts DO occur, they typically start around 6pm and 11pm UK time. And, funnily enough, Weight Sum suddenly increases at those times, indicating high activity which I'm assuming is mainly from (on-line) Gaming!

Timeout: Page loading status stays at 0% progress until Firefox times out and indicates "unable to contact web server". Yesterday after 6pm, for example, UESP suddenly started acting up, taking up to 20 seconds for the web server to respond. Still within my 30 second Firefox timeout, though. Tonight, I haven't seen delays longer than about 10 seconds.

SQL Error: Message appears at top of UESP web pages. Unfortunately, it's not helpful in pinpointing the problem because "error" could mean anything, i.e. SQL Server timeout, SQL command syntax error, etc.

Three days ago, I upgraded from Firefox v3.0.x to v3.5.2 and browsing is now significantly faster, so maybe it's good riddance to timeouts? Then again, it may be too early to tell as yet.

Hope this helps.

I'll admit, I've never paid much attention to times or patterns, but I've gotten timeout errors and "Server didn't respond" errors several times. Sometimes I have to just leave and come back later because the site won't load. This has always occurred for me though on UESP and the forums, since I first joined. Only the SQL errors, which I rarely see, are new. I did get a SQL error yesterday when I tried to add a custom field to user profiles, however. Every time.

I spotted something of interest tonight. Take a look at this TraceRoute result from my home in the UK to the UESP Web Server in Montreal, Canada:-


I've re-run the above trace multiple times today from 4pm (UK time) onwards and the route to UESP has remained EXACTLY the same every time, always via the two NTT routers in London (Hops 5-6) despite them being frequently bottlenecked (high Bounce Times) and even timing out (* symbol). If these each only have ONE neighbouring router, NO alternate routes are possible, so the bottleneck can't be bypassed. This has the knock-on effect of UESP slow-downs and, if they're occasionally "gridlocked" for minutes at a time, UESP timeouts.

So UESP is not to blame. Instead, I can blame NTT (Nippon Telegraph & Telephone Corporation, apparently). Any chance of installing parallel routers to ease the load, NTT guys?