I’m getting a little tired with dealing with shared hosting plans, their customer service, and slow server response times. I run a business, but I don’t quite have the pocketbook for a dedicated server (which is, what, in the hundreds of dollars a month)? I just learned about Linode and VPS hosting yesterday, and I’m intrigued.
I understand Linode takes a bit of set-up to get up-and-going and doesn’t have the traditional control panels web hosting services do, but I’m relatively okay with setting things up from the command line and stuff like that. It will take me a little work, but I’m fine with it.
Is this the right solution for me, as I am looking for a hosting services that will give me quick response time and good customer service. What impelled me to look into this is that yesterday, once again, my website was acting flaky and it all had to do with server response times. Even calling non-existant pages took 12 seconds or so at the worst, and the site even occasionally called up a blank page (and verified as down from “downforeveryoneorjustme.com” and server response time being flagged as a “high priority item” on Google PageSpeed analysis.)
I’m looking for something a little more stable and predictable. From what I’ve seen online, it seems like VPS and Linode is a good solution that’s affordable but not as overloaded as a shared server. And Linode appears to have good customer service reviews.
Am I understanding this correctly? Is this a possible good solution for me?
Static HTML on Apache?
Something centering around PHP?
Some kind of commercial bundle?
I’d submit that using a managed service is really nice vs being your own sysadmin if you don’t need all the customization.
Maybe you’re choosing providers that stink?
You can see my current website here. It’s through intothedarkroom.com and uses PHP, Flash, JS, all that kind of stuff. There a are non-Flash versions of the pages as well for devices that don’t support Flash. I’m not really concerned about the actual structure of the site of itself. What I’m concerned about is reliable and consistent response time. I mean, it really shouldn’t take a server 12 seconds to respond with a 404 should it? (That was one of my tests–to try to access a non-existant page and see how fast it responds.) On most sites, that’s near-instantaneous. That sort of lag affects the general response time from my page.
The website appears to be responsive at the moment. But that inconstancy is what is frustrating me. I know shared servers can cause variability in response time, but yesterday it got to the point that at 10:30 p.m. CDT, I was being served up with a blank page.
I’ve used the following hosting services: Aplus.net, Go Daddy, and FatCow. FatCow is the current one I’m having an issue with. They were great when I started with them a couple of years ago, but recently I’ve been noticing flakiness with the service, and their customer service has been as good as aplus.net’s, and GoDaddy’s, which is to say, leaving something to be desired.
I’m happy to maintain my own damned site if that means I could have a relatively consistent website. For example, right now, my Google PageSpeed rank reads 80/100, which is reasonable. Yesterday, it was as low as 9/100. Every time I check, the high priority item is “improve server response time.”
I run my personal sites on Amazon Web Services (ec2). Works well, easy to get going and very easy to scale up/down. If your site is Flash and image-heavy, you might get a bigger improvement in performance by moving those assets to a CDN (Amazon’s is Cloudfront, and I find it to work well and be affordable, but there are many others) than making the server itself faster.
Inconsistent times like that do tend to point to server load problems. But it’s not necessarily the shared server’s fault; it could be your code. In which case, moving to a VPS won’t necessarily help, especially at the low end.
Have you considered porting your site to a CMS? If you’re running WordPress, Joomla or similar, that gives you a number of options for managed hosting with CDN support built in.
Another option would be to use an optimizer/CDN service like CloudFlare on top of your existing hosting.
I asked about your tech so I could have a notion of whether or not you were doing something weird.
If you’d told me you needed ASP support, for instance, then you need a Windows server, and things change.
I can’t think of a good reason for a 404 to take 12 seconds to show up.
Have you asked the hosting provider for an explanation?
The server you’re on could be having an actual problem that they hadn’t noticed.
Based on what I read about the server requirements here, I don’t think you really need or should go to VPS or Amazon’s EC2.
I think you need to find a hosting provider that doesn’t have problems like your current one.
I’d take a look at using Pair.com were I in your shoes.
I use personally ChicagoVPS.net and Linode.com. We also have some Amazon EC2 instances at work.
As far as VPS vs managed, do you/will you enjoy setting up your server and possibly occasionally tinkering with it ? Can you handle following a tutorial on the web to set up IPtables firewall, add user accounts, disallow root access via SSH and set up automatic updating ? VPS hosting is far better than managed, and almost any hosting provider can beat what you have now (which, when I checked, was absolute garbage - 30 seconds to even load up to show that I needed flash, let alone to download it - that is in “constantly being DDOSed” territory). But it is more difficult and not as “fire and forget”. On the upside, one VPS can host pretty much as many sites as you want, so one $25/mo linode can be 50 websites if you want, and if they are all reasonable traffic all will be quite fast.
As far as support for a VPS you have to calibrate your expectations - it is not their job to help you setup the slice, or deal with it if you use a crappy password and get hacked. Linode.com is a pretty good provider. $25/mo is the minimum you should spend there ($20/m for a 1 gig slice + $5/mo for backups). Less people per node than
There are many cheaper providers, I have two slices with ChicagoVPS as well that I use for less critical sites that I pay $40/YEAR for (http://slickdeals.net/f/5983904-New-ChicagoVPS-3072MB-RAM-120GB-Disk-3TB-Bandwidth-2x-IP-address-55-year-or-2GB-40-year-4-Location-Choices).
Either of those are going to be enough to run dozens or even hundreds of Apache/PHP websites if the load isn’t high.
You need to know a little bit to compare the choices there -ChicagoVPS from a numbers perspective is going to look a lot better than Linode, but they are actually pretty comparable.
OpenVZ virtualization (like the cheaper chicagoVPS) allows more overselling so there will be more users on your node. Xen (like on Linode) makes it harder to oversell/overprovision.
OpenVZ virtualization is less flexible, since all clients have to run the same kernel - which means that some advanced features some people might want to do can’t be done on OpenVZ but can be done on Xen.
Linode has much better management tools and interface for everything.
I find the amazon management interface confusing as hell, and the pricing is difficult to figure out in advance since you essentially pay for what you use. On the plus side you get a free micro instance for the first year, which is probably enough to run your website.
This is a really good point.
I have plenty of background in all of the above config jobs, but I would rather chew my arm off than intentionally subject myself to doing them needlessly.
I would also really like to avoid having to worry about kernel updates and security patches… even if my distro has very good management tools.
There are a number of things you just don’t get good management tools for with EC2. Amazon expects you to be your own support or pay for a premium plan. They also expect you to develop your own management interface to deal with having to change servers, spin up new ones, etc.
I use it myself, but for a very specialized application that I can’t put on VPS or managed hosting.
I strongly disagree here. The micro instance will fall over for a bit if a search engine crawler hits it.
The small instance, the next one up, is enough for a few clients at a time. You’re still not getting a whole CPU IIRC.
Everything above the small instance should handle a healthy number of clients.
The price for a small is around $43 per month.
A medium, which is where I’d take this in case my site got relatively busy, would run $87 per month.
With the limited requirements that the OP has to meet in order to able to run his ‘app’, he’d be silly to go EC2.
Shows what I know - All I knew was that a micro had enough resources to run a site (at low load), didn’t know they limited the CPU that badly. Anyway, definitely agree that Amazon is a poor choice unless you are trying to manage many, many slices in the enterprise (and even then you can usually beat their price).
I have, and this is what is so frustrating. Almost every tech just answers “optimize your website, run it through websiteanalyizer.com (or something like that), blahblahblah.” And I have to explain over and over that I’m not talking about images taking a long time to load, I’m talking about the first connection between my browser and the server taking up to 12 seconds or so. I told them that this even happens with no-content pages where I get a 404 and on plain text pages, where there is nothing to be optimized. I told them the analyzers I’ve used (Google PageSpeed and some other one) all point at server response time as being the culprit. In fact, it’s the only red-flag item on Google PageSpeed. I even had one of the online guys get a blank page and confirmation from downforeveryoneorjustme.com that the website was flaky and going on and offline for no reason.
Of course, when the next tech checked it out, it was working fine, so there was no problem to talk about. Ugh. Until about two hours later, and back to the same circular chain of emails where I explain again it’s server response time and point to 404s taking up to 12 seconds to load, and then the site acting normal again when the next guy checks. I thought I was making progress when I found the online guy to confirm my experience, but he just bumped the ticket up the next level, by when the problem resolved itself. Argh.
Best use case for Amazon is spinning up a few dozen instances for a few hours a week to handle peak loads and not having to pay for them during off hours.
That is the only case where EC2 has an unbeatable price.
Document that it keeps happening.
Make short videos, slap them on youtube with the ISP name on them, reilly them it needs to not happen or they are losing you.
Demand escalation.
If not worth your time, bite the bullet and switch ISP for the hosting.