Site absurdly slow

Question

I don't know why, but the site loads extremely slowly for me, here in Israel. Nothing else has this problem, only physicsoverflow, so it can't be just me. Do we suddenly have a higher load on the servers? Is it possible to get better hosting?

Comments

I actually do not see bandwidth troubles from my position, ping times are around 75 msec. The number of daily visits is continuously growing (>500 per day since mid of July). The total amount of data transfered is around 500 MBytes per day since about one month. I also can't see any sign of attack in our log files. In my opinion the problem is not on our site.

Maybe I should develop a plugin to measure bandwidth directly from the site, so that we get some numbers. Does anybody know some open source code?

---

In Spain and Greece the site loads normally without any problems.

---

@RonMaimon No, in this case I would feel it also, I am administrator too. The difference in data volume transferred to administrators compared to ordinary users is only marginal. I have also contacted our provider (because it could be that they had a DOS attack) but they didn't log any issues. I am very sorry not being able to help you.

---

@ron but you're in Israel right now and I can imagine there's a lot of monitoring of internet traffic going on there because of the situation in Gaza.

---

I have set up a site monitoring at Pingdom since two days now. This site tries reqularly to access our site from different servers all over the world (this increases at least the number of visits slightly :). Pindom delivers then a report to me about the measured response times and I get alerted by email, if the site is inaccessible. This will not mean that I provide a 24/7 service from now on!

The response times are very different depending on where on the world you are. The measurements span from 300 msec to 6.5 sec.

Additionally, I have programmed and installed a logging procedure in the case of issues with the database. Every time you see the famous red text, I get a logfile on the server with a traceback of the code, so that I will (hopefully) be able to analyze the reason of the issue.

---

From Dilaton's comments, Dim10's, and my own experience, perhaps the problem only affects administrators. Could there some large amount of data transmitted for administrators which is not transmitted for ordinary users?

---

@RonMaimon I only felt it for a small amount of time, not now.

Answer 1

It seems that since Saturday at about 16:00 UCT, the issue with our hosting server has been solved. After several contacts and complaints, my measurements with Pingdom made enough pressure to move the support there to help us. Our site has now been moved to another server. Here for example the measured downtimes since Saturday:

Not one minute down since the change! The yellow marker is the time of move of our site. Also the response times have significantly decreased:

The average response time (measured from servers all over the world) is 0.624 sec for Sunday, with is very good. I hope we will stay stable from now one near these values. Let's see what happens on Monday. I continue to monitor these values and will get alerted, if we have again downtimes.

Comments

Great! When you say "moved to another server", do you mean that the problem only affected our server, and not the other servers on our host? Or have we changed hosts altogether? So does the admin email, etc. work as before? And is the payment plan the same as before?

---

@dimension10 No, we are on another server of the same company. They claim to have more than 1,000,000 customers in 149 countries. I assume that they have 'halls' full of servers, distributed over different countries. So we just got a better place now. Everything else stays as it was.

Answer 2

I am writing this answer to bump, because I am back in NY now, and it just happened again! Spinning site, "Database connect error" for a few minutes, and everything was unreachable for at least 3 minutes. This type of thing is more damaging than it looks, people can be dissuaded from contributing. I think there is something mysterious wrong with the hosting.

Comments

Yep, the same happens here in Kühlungsborn.

---

You are absolutely right, Now I have numbers from Pingdom. For instance during the last 24 hours, the server was in total down for 1h 24min, with a longest continuous period of 44 min yesterday. It seems that these issues occur every day with a peek between 07:00 and 10:00 UCT. Also the response time increases up to 10 to 15 sec during this period, while normally the average is about 1.5 sec.

I will present these numbers now to our provider One.com, with the urgent appeal to correct that immediately. Otherwise I will have to dismiss the contract. Does anybody have a proposition to get a hosting with a better performance, just to be prepared?