Tag Archives: Web

Internet Tech Web

Hosting services: my recommendations

After hearing on Twitter how an acquaintance’s new hosting provider went ‘mammaries skyward’ this week, much to their understandable annoyance, it occurred to me that I have some recommendations I can make on this subject. While I don’t host that many sites, I’ve been doing it for long enough and had experience of both personal and medium-traffic sites that I’ve experienced the highs and lows quite a few times already.

The Golden Rule: Support > Everything

When it comes to hosting, the most important thing to look for, beyond what all the statistics of how much space and bandwidth you get, beyond even quoted up-times, is the quality of the support service. The big question is: when things go wrong – and if you host long enough, eventually they will even in the best possible hosting environment – how quickly are problems resolved, and how responsive are the support engineers during the process. Literally nothing is more important than this, and unfortunately it’s the one thing that you’ll only really learn with experience, unless you’re hosting a site big enough that you can get a formal SLA. Assuming you’re not going big enough for that, the only way to judge this is by being with a provider for a while, or knowing someone who has been with them, or possibly looking at online review sites – although frankly these are often highly unreliable, polluted as they are with inaccuracies and omissions either because of ignorance (people who post glowing reviews after being with the site 2 weeks) and unfortunately by frequent shill reviews.

I’ll post a couple of hosts I’ve had good experience with over many years later in the post, through good times and bad.

Know Your Bandwidth

Personally, I instantly rule out any host that claims ‘unlimited bandwith’. This is a crock – there is no such thing, and to claim there is just means the host is already lying to you before you even start – they have to pay for their bandwidth, so they can’t possibly allow everyone truly unlimited bandwidth and stay in business. If you really need unlimited bandwidth, i.e. you have a high-traffic site with lots of media files, then you will quickly bump into the way that these sites offer this ‘unlimited’ deal – via throttling. You may not have an absolute physical cap on your bandwidth, but if the tap is locked off to a slow dribble beyond a certain usage, it’s really worthless. In practice, ‘unlimited bandwidth’ is just a marketing point that they hope will draw in people who will only actually use a tiny amount of bandwidth, but will somehow favour them because the offer looks good. Don’t be one of those dumb guys.

Really you need to establish your bandwidth requirements and head for a host that can fulfil them for a reasonable price. For example, ogre3d.org uses between 125GB and 250GB per month, which is a reasonable amount, compared to my personal site here which only needs 5-10GB per month.

If you have ‘spiky’ bandwidth, i.e. occasionally you need to be able to distribute large amounts of data, but it’s not a constant stream, it would be best to go for a lower monthly limit and host high-bandwidth items elsewhere. I often use Amazon S3 for this purpose which can be made to look like a sub-domain of your own site, and which charges for bandwidth at a very fine granularity so matches your demand closely – it’s more expensive than buying a monthly allowance if you use it a lot, but for on-demand spikes it works very well.

Shared, Dedicated or Virtual/Cloud?

I currently use two shared hosts and one dedicated host, to match the demands of each site. Personally, I’m still very skeptical about virtual private servers and cloud hosting, due to a bad experience I had a few years back when we tried running ogre3d.org on a VPS. We lasted not much more than a month before we moved the server to a dedicated machine because the VPS simply didn’t deliver on its promises – performance was unpredictable and to be honest you had the worst of both worlds – you had to admin your own server but you still didn’t have a 100% guarantee that no-one else would be screwing with something on the machine, or that the disk arrays wouldn’t be hammered by someone else (regardless of CPU assignment), or some other balancing issue. Virtualisation has evolved in the last few years so this may not be an accurate representation anymore, but personally I wouldn’t go for a VPS again any time soon, unless it was a machine I controlled entirely and partitioned myself into virtuals – at least with shared and dedicated servers you know exactly what you’re getting – either a low-maintenance but shared resource environment, or total control & power. VPS claims to offer a middle ground but in my experience it didn’t deliver.

So, who do I use?

For my shared hosting, I’ve been using Hosting Matters for about 10 years now. I went through a couple of other hosts before them and had terrible experiences, but since I switched to them I’ve been very happy. I can count the number of hours downtime (that I’ve been aware of) over those years on one hand, and whenever there’s an issue they’re incredibly fast to respond – they have both community forums and support tickets depending on the urgency. It’s also very reassuring to see the same names cropping up in the support responses over the years.

Their offerings are pretty standard, nothing that would make them jump off the page for anyone looking for a stellar feature list or super-cheap pricing. But they’re very reasonable, they’re honest about what they’re offering (like bandwidth), and as I said before, support > everything.

For dedicated hosting, since 2007 I’ve used Dedipower. They’re based in Reading, their support staff are all local and are on the end of a phone if you need them (no call centres). Having been through a UK dedicated server comparison twice in the last 3 years (once again just recently), Dedipower came out as the most competitive for the service they were willing to offer, and I’ve been happy with the support service. In once instance in fact, when I moved a sub-site off the server, they were quickly on the phone to me within 10 minutes to tell me it was ‘down’ – at which point I had to explain it was expected & apologise for not notifying them in advance. You really can’t complain about that.

I hope that’s useful to someone. In case I need to point this out, I’m not getting paid or receiving discounts to promote either of these hosts, they’re just the two I’ve been most happy with over the ~10 years I’ve been hosting sites. YMMV but they’ve worked well for me :)

Internet OGRE Web

OGRE OS & Browser Stats

I’ve had requests to post the OS & browser stats for the OGRE site, which I didn’t include in my previous demographics post, so here we go.

A caveat to start off with – as a programmer-oriented site our users are obviously a little different in their choice of tech to the population at large!

Operating Systems

Not really a surprise there, Windows dominates the landscape, with Linux and OS X pulling up the rear. Personally on the desktop I’m a Windows and OS X user so my visits are contributing to those rows. There’s a decent showing for the mobile platforms too, iPhone and iPod particularly, a fair few on Android and even some early iPad hits.

Windows Breakdown

It’s worth drilling down into the huge Windows stat to see what versions people are using:

XP still rules the roost then, and thank goodness Vista is sinking rapidly to the bottom like the cast-iron turd sculpture it is – Windows 7 has already more than doubled its share. And it’s amusing to see that a couple of crazy b?st#rds are still running ME and NT.

Mac OS X Breakdown

I’d heard that supporting pre-Leopard versions of Mac OS X was increasingly becoming pointless since almost everyone had upgraded, and these stats bear that out – 96% of users are running 10.5 (Leopard) or better.

Even more interesting was the sole visitor running 10.7! Was that an error, or did we have someone from Apple on an unreleased future version visiting the site? And what’s that ’68K’ entry about – someone running an Apple ][ emulator or something?

The Linux breakdown wasn’t interesting (99% ‘not set’, the rest just kernel versions), so there’s nothing to post for that – don’t email me about  missing out Linux please ;)

Browsers

Firefox is the clear winner here at a huge 50% – this definitely reflects our developer-focussed audience. In fact, I used to be an avid Firefox user until quite recently, when the new version of Chrome added the equivalent to the extensions I relied on in Firefox, at which point I switched because Chrome is more efficient with memory in particular. IE’s share at 18% is definitely not representative of the general user population, but then our users tend to be a bit better informed than that :)

So there you go, request fulfilled – hope it was interesting.

Internet OGRE Web

OGRE web demographics, revisited

Almost exactly three years ago, I posted an analysis of the traffic on ogre3d.org and the rough country breakdown of our users, which is always fascinating to me. I hadn’t actually been collecting web stats on the site for about a year (the previous set-up was lost when I had to recreate the server in a hurry, and somehow reinstating it never seemed to rise to the top of my TODO list), but a month ago I finally got around to adding Google Analytics to the site. The results have been very interesting, particularly when compared to 3 years ago, so I thought I’d share some factoids with you.

Visitors still increasing

In 2007, log analysis indicated we were getting a little over 100,000 unique visitors per month; obviously this is not 100% accurate due to shared & dynamic IP addresses, people logging in from multiple sites, etc, but it’s a reasonable order of magnitude indicator. Analytics is typically more conservative in its figures, since it excludes bots better as well as non-Javascript browsers, but still in the past month (actually only 28 days) we’ve had over 120,000 unique visitors to the site – and 1.2M page views – so we’ve sustained and slightly improved our user traffic. And all without any Slashdot posts ;) Also, this doesn’t count visits to Sourceforge, BitBucket or ome static generated HTML like our online documentation pages.

The Meteoric Rise of China

In 2007, China ranked 15th in our league table of visitors. Three years on and they’ve risen to the number two spot, comfortably surpassing Germany – at first I wondered if that was down to users there using fewer proxies, but since the figures for other countries have remained fairly stable I think the majority of this is genuinely a vast increase in the number of Chinese visitors to OGRE’s site. Here’s the top 10 countries (figures are for the number of visits):

The range of countries is demonstrated by how many are in the grey ‘others’ section (38.73%). Except for the massive change in China’s share, most of the other countries have stayed approximately in their relative positions & shares of the user population since 2007.

Region View – Europe still dominates, Asia challenges the Americas

The country view is, however, quite misleading if your aim is to decide where to locate a web server for example, because it naturally biases the figures towards large unified countries (like the USA and China), and doesn’t really show a true regional picture. For that, we have to examine the numbers (again, number of visits) by continent:

Now, even though the continent view includes Russia in Europe which screws up the locality principle a bit, even if you exclude that Europe dominates our community, with close to 1 in every 2 visitors to the site being from Europe. The Americas  and Asia share most of the rest almost equally now, which is a change from 2007 when the Americas were more dominant, and everyone else shares the scraps (3.5%). The Americas figure is made up of about 86% North America and 14% South America, and Asia is predominantly (60%) the Eastern Asia countries (mostly China, but South Korea holds its own too), with South-East and Southern Asia sharing the rest – particular hotspots there are India, Indonesia, Malasia, Turkey and Vietnam.

City Clusters

One thing I like about Analytics is the ability to drill down into countries and look at the local clustering. There are the expected clusters around cities – in the USA, the top 2 cities are unsurprisingly New York and Los Angeles, although Columbus OH takes the number 3 spot, and in the UK the clustering around London is massive – but they typically represent only about 25-30% of the audience, with the rest being scattered pretty much uniformly across most areas of the country in question. It’s fun to be able to point at almost anywhere in Europe, North America and the southern and eastern parts of Asia and to have a pretty good chance of being quite near to someone who has used the OGRE site.

The Monday morning OGRE fix

With OGRE obviously used by a lot of people in their spare time, you might expect that the weekends would be the busiest times for the site, but the opposite is in fact true – Mondays are consistently the busiest days (particularly 6-9am PST), with Saturday being the least busy. Whether this is because people are working with OGRE, or just cheekily surfing in their work time rather than face the Monday workload, is hard to verify!

Final Thought

I get a kick out of looking at these stats so I hope you find them interesting too. It’s really cool to think that there are only a very small number of countries (such as North Korea and Laos) from which we don’t get any (non-proxy) visitors from in a typical month, and it’s very interesting to see how the visitor base is gradually spreading out and diversifying, something which I’m sure every site witnesses but it’s interesting to see it in your own data. The question is – will China keep the current trajectory? At this rate, they’ll take the number one spot from the USA in only a few more years and put Asia second in the regional rankings!

OGRE Web

Misery loves company

So, now that I have basic service back at my newly reinstalled server, it’s time to start re-downloading clean packages for the software that makes up ogre3d.org. Imagine my surprise when I went to phpBB.com and saw this:

Looks like I’m not the only one having a bad week :?

Current priority is getting the forums back online, watch that space.

Tech Web

Dodging bullets

So, as we all know the whole MicroHoo! idea has been called off now, unless you believe the conspiracy theorists who believe this is all still part of Count Ballmer’s plan to devalue Yahoo! (as some of its shareholders go through a set of inevitable legal tantrums) and make it easier to pick up later. I’m not so sure about that myself – after all didn’t the rotund billionnaire say he wasn’t going to raise the original offer for Yahoo!, before doing exactly that? Doesn’t really sound like a bluff, unless you factor in that he knew they were going to reject that too, thus increasing the chances of annoying their shareholders. But then we start getting into the ‘he knew that they knew that he knew’ territory and it all gets dreadfully confusing. Whatever the case, the whole deal has always sounded like a recipe for total disaster to me. So much so, I was kinda hoping it would go ahead just so that I take some perverse pleasure in watching the train wreck later.

No-one I’ve spoken to ‘gets’ the Yahoo! bid at all, seemingly a constant across the entire spectrum of opinion about both companies. I can understand that Microsoft would love to get hold of Yahoo!’s ad customers, and to a lesser extent all the freeloading users required to eyeball said ads (and the services required to keep them sweet), but they can’t have seriously thought that it would be a smooth transfer. The philosophies of the companies couldn’t be much further apart, with Yahoo! very much invested in open source technology and service models, and spritually the culture of the company is very much of a younger, consumer-oriented, more agile and open thinking sort, very different from the business-focussed, shrink-wrapped and closely integrated, keep-it-in-the-company sort of vibe that Microsoft tends to exude. I’m sure Microsoft must have earmarked a bunch of money to use in encouraging key people to stay, but honestly in my experience the very best people aren’t swayed that much by that kind of offer. I think had the deal gone ahead, MicroHoo! would have haemorraged much of the best Yahoo! talent to other Valley companies like Google (or to new start-ups) in the first few weeks, whatever reassurances might be given.

And what about the technology? Yahoo! is again the antithesis of Microsoft here, running their core business on open source stacks. Quite whether the acquisition would have eventually led to that being replaced with equivalent Microsoft technology I don’t know, but a switch would seem like a pointless effort – regardless of your technology preference, the ‘if it ain’t broke, don’t fix it’ rule is universal. But at the same time an entire growth arm of the business running ‘competitors’ products would be somewhat jarring I would have thought, so perhaps their principles would have had them invest in that transition anyway – none of which sounds like a particularly efficient investment.

So actually, I think Microsoft has had a lucky escape here – I think they would have been hugely distracted (not to mention considerably poorer) trying to make a Yahoo! acquisition work. Sure it might have advanced them to some degree, but it surely would have been a messy, inefficient fight which would have given their main competitors – ok competitor – much amusement. The big question is whether they can achieve more on their own. The word is that Steve Ballmer is obsessed with beating Google (you would have thought he’d be content with all the billions he already has, but I guess there’s no pleasing some people), so it will be interesting to see what Plan B is. I hope they’ve screwed the chairs down in his office.