[Hosting] Schoolwebsite Doesn't Work in Snow!

[DavidYoung]

Hello,

Before the summer, we got Schoolwebsite.co.uk to design and host our new websites: www.coombegirlsschool.org, www.coombeboysschool.org, www.coombesixthform.org. The sites look nice and work quite well, but have always been pretty slow, sometimes taking over 60 seconds to load the homepage! The real problem, is that any time there has been snow, or the theat of snow anywhere in the country, the website has just been overloaded, even as I type this, I am getting an error 503 when I try to load it.

What I want to ask, is does anyone else have websites hosted with Schoolwebsite, and are having the same problem? Do schools who have their website hosted with other commercial companies (particularly companies that specialise in school websites such as e4Education or CleverBox) have similar problems? Are there schools on Schoolwebsite who don't have this problem?

We have raised it with our project manager a couple of times and had a bit of a mini rant at BETT, but this morning was the same story. The School Administrator and I are trying to get some data together to show them, so any responses are appreciated.

Thanks,
David

[Abaddon]

We had a similar issue some years ago, and subsequently moved to a different hosting company. A bit of digging turned up that the previous host had 867 (yes, I typed that right) websites running on it's webservers.. bit of adverse weather and all the hits took all of them pretyt much offline.

(this wasn't the site you mention, by the way - it was A.N.Other)

[glennda]

E4education did have a problem the first time we had snow - but by the second and third times they beefed up the infrastructure to cope better.

will Schoolwebsite allow you to move the sites they have created to a third party host? as the main problem these companies face is due to the fact they only host for education whereas a third party won't get the massive hit that these companies do.

[Mr.Ben]

PM Sent

[sonofsanta]

Had a similar problem at a different host - all schools hosted on one server, so same usage patterns, wherein 98% of the time there's minimal traffic and no urgency to uptime, but 2% of the time it gets hammered and everyone needs to know and its the one time you actually need the website up.

Given their company name I suspect it is an inherent weakness of their business; if they can't do anything to reduce/spread/expand-to-cope-with load then you'll need to move. FWIW, we worked with our hosts to display a simplified page between 6-8am when the relevant script is activated server side; with no menus/headlines to generate dynamically the load was much lower and the site stays stable.

Also, consider setting up a Twitter account so you still have some way of getting the message out - advertise it briefly on your site and word of mouth will soon spread.

[pete]

Move the site to a decent, commercial not-aimed-at-schools host.

[jdoyle]

visited your 6th Form site. Very nice but took a very long time to load.

we host ours internally. Response times are generally good. Took a big hit this morning (according to google analytics) but no problems encountered.

Downside to internal hosting is if we lose our SWGfL connection, which has happened a couple of times.

If I had a choice in our hosting, I'd be going the route @pete mentioned.

[Willott]

One company I've recently heard about and looked into is MediaTemple and their grid service - need to do further research, but seem fairly good so far. The other thing to ensure with web hosts (shared hosting types) is that they separate SQL off their web hosts. I know our LA currently have a setup where websites are hosted on the same box as SQL, so when any load is on servers, php pages reliant on SQL content are slow to load, or fail altogether - for snow days, I put up a temporary HTML page which loads with no issues, with a link through to the main site.

The issue with hosting web and sql on the same server is that the server is setup for "general" loads, ie mid amounts of RAM, okish hdds, setup in RAID5/6 or similar, so once you have more than a few sites, RAM is being eaten up and the disks aren't fast enough to service the SQL load. Dedicated SQL servers would have RAID10, 15k disks, and shed loads of RAM so can respond to queries quickly, and the web hosts then just get on with the job they're setup for... serving data over http/https (and ftp).

Cheers

Will

[DavidYoung]

Thanks for all the comments.

One thing we did notice, and I'm sure its completely coincidental, but all the websites shown as examples on their homepage were working while ours was offline, like I say, I'm sure there's nothing to read into that

It seems to be OK now, but we will find out next snow-day/potential snow-day!

David.

[GrumbleDook]

We have had similar issues at RBC level and have ended up investing in more kit and also providing alternate hosting for PHP heavy sites which Plesk didn't like. Seems to have gone well so far and many thanks to the folks who have done the work.

[Edu-IT]

Thanks for all the comments.

One thing we did notice, and I'm sure its completely coincidental, but all the websites shown as examples on their homepage were working while ours was offline, like I say, I'm sure there's nothing to read into that

It seems to be OK now, but we will find out next snow-day/potential snow-day!

David.

Can you not use something like Google Analytics / Pingdom to monitor the performance then slap it to them?

[crutchfieldj]

Sounds like their server's are overloaded- as stated above, over 800 websites on one server is definitely not recommended, even with normal usage patterns (i.e. no snow).

Switch to another host that doesn't overload it's servers.

You can also use loadimpact.com to simulate users and see how the site copes with the stress.

[schweb]

Hi David,

Our main point of contact for the school will have received further information on the outage you mention above but I thought I would take this opportunity to respond directly to the post. The cause of the issue was a hardware failure with the load balancer on one segment of our infrastructure (hence the reason that the outage only affected a small number of schools as you mention later). Whilst we cannot rule out completely that the failure itself was in part traffic related the infrastructure segment had been load tested to a much greater capacity many times before. We also succeeded in replacing the component in under 1 hour to resume service to those schools affected by the outage. Our hosting provider Rackspace UK have already put in measures to ensure if the same component fails in future, normal service is resumed in minutes rather than an hour. Incidentally, we run both a private and public cloud infrastructure which at peak times give us almost unlimited capacity in terms of dealing with traffic spikes. We certainly do not host 800+ websites on a single web server.

We would also not recommend a client load testing their own site. If this is necessary it is something that can be conducted on behalf of a customer, ideally on a test version of their site. It is also likely that the bandwidth of the user running the test would choke way before reaching even a reasonable simulation of a snow spike. The traffic would also probably be flagged as a DOS attack if the supplier were not aware of what you were doing.