We have been using Slicehost's 256MB slice for a little over two months now for a PHP site and haven't had any issues; uptime has been excellent and support is great. The fact that they are one of the few providers to offer Arch Linux is another plus, too.<p>Another place I personally purchased a VPS from to test out is Linode (<a href="http://linode.com" rel="nofollow">http://linode.com</a>). I've only had it for a week so far, but so far no problems. If you're going to run a Rails app, you might want to look at them as for $20 they provide 360MB of RAM instead of 256 :) Their community seems on par with Slicehost. Their Xen is in beta testing, though, so for now their VPS' run UML.
I have nothing but good things to say about Slicehost. That said, 256 is a very small sandbox and Apache doesn't play nicely in small sandboxes without tweaking. Our site runs several PHP apps.<p>We hit the wall one day with a number of apache2 processes hitting the 60-80MB range. I could barely SSH in to fix things.<p>We ended up arriving empirically at the following (taken from /etc/apache2/apache2.conf), without stress testing, but watching patterns over several weeks.<p><pre><code> <IfModule mpm_prefork_module>
StartServers 2
MinSpareServers 3
MaxSpareServers 6
# MaxClients 150
MaxClients 10
# MaxRequestsPerChild 0
MaxRequestsPerChild 200
</IfModule>
</code></pre>
The key is to take MaxClients down to a small number, like 8-10, as each of these normally takes anywhere from 2-40MB. To make this work, you'll need to have the request count set reasonably small. We set ours to 200, so RAM consumption doesn't get too large before the process is cycled. People often suggest 4000 or some such number but the process just grew too large with that setting. You'll also note that we have tuned the startup and idle parameters to be very conservative.<p>This is probably a laughable configuration for many sites, but it works well for a light site like ours running a few PHP apps. I should note that we're also running MySQL and PostgreSQL. They don't have a big impact on RAM under light load.
I did pretty well on a 256 meg slice. I outgrew it after awhile, and I'm currently running 8 mongrels on a 512 slice, and I'm pretty much hitting the top.<p>RAM is by far the biggest bottleneck for me.<p>Overall, Slicehost is great. I just wish they'd include more RAM in their plans.
I have a 256 meg slice which has been rock solid and with no issues. I use Litespeed (<a href="http://www.litespeedtech.com/" rel="nofollow">http://www.litespeedtech.com/</a>) for the web server. It is excellent for Ruby on Rails hosting. It dynamically manages Ruby processes - if there is no traffic it shuts down the processes, if there is traffic it ramps up the number as necessary up to the max that you specify.<p>On my slice I have a few PHP sites, a few static sites and 4 RoR sites (most of which get very little traffic.) Overall about 30k page views a day.<p>Litespeed isn't open source, but it is free and it works great for my hosting.
Slicehost is amazing. I'm currently running 3 Wordpress blogs off of a 256MB slice. I would highly recommend either ditching Apache or going with a larger slice though. I did experience some memory issues when using Apache (although this could be cleared up with a bit more configuration on my end).<p>Regardless, I'm running Lighttpd (10 work handlers using the same amount of RAM as 1 Apache handler), proftpd, MySQL 5, PHP5 via FastCGI, and sendmail. I've never dipped below 80MB RAM left and load times are extremely quick.<p>I've offloaded my email services to Google Apps, which has it's advantageous and disadvantageous.
I've had both Slicehost and Linode, and Slicehost is much faster by far (Xen). I used the 256 slice to power a few of my Rails applications, but I upgraded to the 512 so I can add more.<p>The best experience I had with Slicehost was the upgrade from 256 to 512. It was a couple clicks on their admin panel, 30 minutes of downtime, and then a quick SSH to verify everything was good.
This question pops up a lot on Slicehost's Forum. Search for "views 256" @ <a href="http://forum.slicehost.com/search.php" rel="nofollow">http://forum.slicehost.com/search.php</a>. You will see all the many people running from blogs to rails on 256mb without a problem.
I just posted about doing some basic load testing on my 256 slice here:<p><a href="http://blog.notedpath.com/2008/02/10/using-jmeter-for-load-testing/" rel="nofollow">http://blog.notedpath.com/2008/02/10/using-jmeter-for-load-t...</a>
Actually, we've been having some issues with our 256 slice. We're running our svn and trac off it, and trac runs incredibly slow. Our app has poor data transfer rates when we send anything of a significant size over the wire.
on slicehost 256mb ram, running erlang + yaws . nice documentation, setup. but off late have have to do couple of server restarts , coz of what seems like to be unscheduled reboots . need to investigate. so helps a lot to have a service like monitorus etc that pings u via mail when the server down,etc.
Yes, my experience so far has been positive. I got Ruby on Rails up and running fairly easily; it's documented here:<p><a href="http://www.perplexedlabs.com/2008/02/04/building-a-rails-capable-slice-from-scratch/" rel="nofollow">http://www.perplexedlabs.com/2008/02/04/building-a-rails-cap...</a>
<a href="http://searchyc.com" rel="nofollow">http://searchyc.com</a> is hosted on linode for a couple of months now and it's been great. we initially wanted to go with slicehost but last year their wait list was so atrociously long that we gave up on ever seeing their invite.<p>for the same $19.99 linode has more bandwidth and memory and xen support. really, what we've realized is actually most expensive is the cpu cycles since our app would crash consistently because the shared resources peaked when we started stress testing. since then with better resource management, better apache fiddling, and a couple rewrites we're handling a peak loads of more than 2000 people and an average load of 500 people.<p>linode's documentation is somewhat lacking, i much prefer slicehost's wiki and forum for support but other than that we have a cluster of 4 or 5 linodes running and it's been great.