Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


On demand cloud servers?
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

On demand cloud servers?

CloudxtnyHostCloudxtnyHost Member, Host Rep

Anyone using them what do you think? Do you rent the server in isolation or just have the credit to as many as you like.

What about the technology that will be used, openstack, cloudstack.apache.org, do you even care?

Interested to hear from anyone using it or offering the service.

Comments

  • I'm curious to hear people's thoughts on what stack to use; I don't feel the other thread has focused much on that.

    I've only dabbled a little so far, but my use case consists of bursts of very high CPU, parallelizable across maybe a dozen VMs, too infrequent to warrant dedis. I'm not (so far) into needing HA failover, advanced networking, etc. But API access from ansible is a must for me (and both OpenStack and CloudStack have that).

    In my very limited experience, Apache CloudStack is pretty straightforward and just fine for managing compute, OpenStack is complex and has some funky interactions between different modules but seems to be the future, and OnApp is functional but sometimes a bit flaky.

    Within the LET community, I feel LunaNode is a great example of a good OpenStack implementation, and iwStack has a decent CloudStack offering (at amazing prices). I would be delighted to see more competition in this space.

  • The stack and api probably matter much less than the hardware available, which of course eats more money than anything. If Seanho needs a dozen CPU's and the host doesn't have them in stock, the software doesn't matter. With a big enough inventory you can do stuff like AWS spot instances but that's probably out of reach of LET hosts.

    My workload is similar to Seanho's but I've been finding it very nice to have a cheap relatively fast dedi (Hetzner i7-3770 at about 30 USD/mo) as a long term rental, since it has 2.5TB of raid-1 storage as well as about 9k passmark of always-available cpu power, probably about what you'd get with 8 or so cloud "cores". It's no big deal to let something run overnight or for 24h and that handles lots of computation tasks. If it's a week long I just start it going and look in on it now and then. I'm generally busy enough to not be blocked by that.

    So adding parallel hardware gets interesting if the computation is more than a maybe week long, which means I probably want at least 20 more cores, maybe 50 or more, since if I'm paying by the core-hour I might as well try to get the thing done in a day or less. And most hourly compute resources are at least 2x as expensive per core-hour as my i7, so if I need more than 2 weeks worth of hourly compute, I can spend less and get 1 month of another dedi.

    Maybe for more typical internet server uses the above logic doesn't apply, but for cpu hogs (workloads like the above) you really need a lot of gear to be in the game, imho. I can't really speak for the person who wants a huge internet presence for some special event but that I'd guess that person wants monstrous bandwidth AND maybe a lot of computation.

    Also it's nice if you can rent out single tenant or bare metal computers rather than VPS, so that the full box's resources are available to the user. Scaleway has bare metal, SpeedyKVM V-dedi is a VPS version, Vultr dedicated (super expensive) is another, etc.

    Finally I sometimes want to spin up a temporary small server to test something or to deliver a clean environment to another person. That's a fairly easy workload (the equivalent of a $5/mo DO/Vultr instance is plenty) and hourly is nice, but daily or even weekly would be fine.

    Thanked by 1seanho
  • Thanks @willie for your experience; that's really helpful! I'll start looking into dedis more seriously.

Sign In or Register to comment.