Re: [tlug] VPS providers in Japan

[Mark Karpeles <mark@example.com>]

Thanks to everyone for a lot of precious feedback.

On Wed, Aug 20, 2014 at 1:49 AM, Daniel A. Ramaley <daniel.ramaley@example.com> wrote:

That was my first question. I have a couple Linodes, one of which is in
the Tokyo datacenter. From a quick look at the forever.net page, the
significant differences are that forever.net is about 1/3 the price of
Linode, and that it uses KVM instead of Xen. I don't know about AWS as
i've not used that service.

The fact that we are using KVM instead of Xen is an important point I believe. We do not plan on providing Xen instances, for different reasons (technology has improved a lot and we do not see the issues we had before with memory management, but at the same time KVM based machines have also improved a lot and loss in term of performances due to hardware virtualization is minimal) but mostly because we want to handle all servers the same way.

Forever.net's concept would be to handle VPS and dedicated servers the same way, in order to have the same interface on both cases. Some differences are unavoidable (reboot time for a dedicated server is usually 3~4 minutes, while a VPS will finish booting our rescue image in 15 seconds).

On Tue, Aug 19, 2014 at 10:24 PM, Scott Robbins <scottro@example.com> wrote:

You mention various Linux flavors, will there be FreeBSD instances
available?  My current place is actually (vaguely) considering doing
something in Asia, but we're very FreeBSD centric.

We've had some demand for FreeBSD too, and we are working on this. Main issue there is that we need a FreeBSD rescue image that will need to be maintained separately from our current rescue image. Note that we are planning to allow upload of ISO files at the same time as we open remote console for both VPS and dedicated servers, that will allow advanced users to install virtually anything.

On Tue, Aug 19, 2014 at 9:26 PM, Josh Glover <jmglov@example.com> wrote:

> I.e. your big advantage over
> pay-as-you-go cloud is you are cheaper, so you could actually show the
> equivalent EC2 costs for the same powered server. (Most people I know
> who use ec2 are using it 24/7.)

Make sure you do an apples-to-apples comparison between your VPS
instances and EC2 reserved instances:

http://aws.amazon.com/ec2/purchasing-options/reserved-instances/

The advantage there of your solution is that you don't have to reserve
instances for 12 or 36 months. :)

Something else I saw with Amazon (I do not know if it applies to reserved instances) is that sometimes Amazon will send you an email telling you that the host for your EC2 instance will be going down ("scheduled retirement"), in which case your EC2 instance goes down too unless you reboot it before, or migrate to a new one. More details on http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/monitoring-instances-status-check_sched.html quoting: "An instance is scheduled to be stopped or retired when AWS detects irreparable failure of the underlying hardware hosting your instance."

The system we created is cluster based, allowing us to live-migrate instances from an host to another transparently, which means that when maintenance is required on a given host or in case of hardware failure, instances running there can be moved somewhere else without impacting customers (maybe a couple seconds lag during the migration).

Of course some cases of hardware failure may cause machine to go down completely (we have redundant network, power, and use what's best in term of hardware so usually we should have some kind of warning before anything major) but even in that case, instances can be launched again immediately thanks to the cluster replication.

Mark