Any news on Xen?

@sarge:

Just wondering if there's a new 'ballpark' time estimate for deploying Xen for production use.
We've been waiting for PAE (> 4GB memory) support. Xen 3.0 is about to go into testing this week, meaning only bug-fixes, but unfortunately PAE isn't done and will be an exception. AFAIK, PAE-guest support is the only thing lacking.

When PAE is completed, there will be a test cycle which will continue in our private beta. When I feel PAE/Xen-3.0 is solid enough, I'll widen the beta testing to the public.

I'd put production Xen in the 4-6 month ballpark. Could be sooner, might not..

-Chris

Clearly the answer is to swap out all the Linode hardware with Opteron servers. No problem, right?

I'll take some of the old systems off your hands, Caker. :)

stumbles across this thread

yeah ive been monitoring thisthread quite a bit, kinda sucky that there isnt any new info.. prods

how is the beta going in terms of testing out performance issues etc? from my point of view stuff seems to run a lot faster on the xen box then the uml node box (am in beta test). i notice a lot better response times from horde2/imp3 in xen then uml, which is a major factor for me..

im hoping that the memory support stuff that we were waiting for appears sometime soon, so that the boxes can support the 6gig ram etc..

any more news caker?

cheers

NF

Is there any news on Xen?

Is it possible to get involved with the beta?

@egatenby:

Is there any news on Xen?
Not from my end. I don't want to make any promises or set an expectation as to when we'll have Xen ready for production.

I will say that PAE (able to use more than 4G RAM of the host) support has been working for me, and that was the only show stopper…

@egatenby:

Is it possible to get involved with the beta?
Yes, we'll do a private beta for a while and then open it up to existing Linode customers on a first-come first-gets to test basis. I'll announce that here.

-Chris

dibs! :P

lol

I hate to be distrusting and cynical (linode has never given me reason to be that - it is just the way I am), but the increase in performance from Xen will not just mean that you will be putting more instances on the same host?

@caker:

@egatenby:

Is there any news on Xen?
Not from my end. I don't want to make any promises or set an expectation as to when we'll have Xen ready for production.

I will say that PAE (able to use more than 4G RAM of the host) support has been working for me, and that was the only show stopper…

@egatenby:

Is it possible to get involved with the beta?
Yes, we'll do a private beta for a while and then open it up to existing Linode customers on a first-come first-gets to test basis. I'll announce that here.

-Chris

@peterhoeg:

I hate to be distrusting and cynical (linode has never given me reason to be that - it is just the way I am), but the increase in performance from Xen will not just mean that you will be putting more instances on the same host?
No. We've never oversold our machines. In fact, Xen makes it impossible to do so.

-Chris

> Xen makes it impossible to do so. How does that work?

> How does that work?

As I understand it, Xen requires that each domU have a certain amount of physical RAM. If you have 4 gigs and slice that up into 512M for dom0 and 28 domU's with 128M of RAM each, that's that.

With other technologies, I believe that you could have many more domU's because they will "steal" RAM that other domU's are not using at the moment. This is bad when, all of a sudden, that domU starts using the RAM you stole. Swapping would occur at the hypervisor level and everyone suffers.

Caker, is that anywhere close to correct?

• Brian

@bkirkbri:

> How does that work?

As I understand it, Xen requires that each domU have a certain amount of physical RAM. If you have 4 gigs and slice that up into 512M for dom0 and 28 domU's with 128M of RAM each, that's that.

With other technologies, I believe that you could have many more domU's because they will "steal" RAM that other domU's are not using at the moment. This is bad when, all of a sudden, that domU starts using the RAM you stole. Swapping would occur at the hypervisor level and everyone suffers.

Caker, is that anywhere close to correct?

• Brian

Without getting technical . . . yes. If you allocate $foo ram for 'user' Bob, then that specific chunk of $foo ram is dedicated to him, nobody else can have it. period.