Re: hosting the sage cell server

53 views
Skip to first unread message

William Stein

unread,
Apr 14, 2015, 8:18:11 PM4/14/15
to sage-devel, sage-notebook, sage-edu
On Tue, Apr 14, 2015 at 2:50 PM, William Stein <wst...@gmail.com> wrote:
> Hi Sage Developers,
>
> Is there anybody who would be willing to host the Sage cell server?
> http://sagecell.sagemath.org/
>
> This would take a dedicated machine or VM with at least 32GB RAM and
> at least 100GB fast hard drives.

This requires absolutely no maintenance apart from making sure that
the machine is functional and trusting Andrey Novoseltsev with admin
access.

-- William




--
William (http://wstein.org)

Sven Köppel

unread,
Apr 21, 2015, 8:49:15 AM4/21/15
to sage-n...@googlegroups.com, sage-devel, pokal-t...@elearning.physik.uni-frankfurt.de
Hi William,

we offer two machines located in our data center (Opteron 270 with 16GB RAM and .5TB HDD each) to host the Sage cell server.

Bye from Frankfurt/Germany,
Sven





--
William (http://wstein.org)

--
You received this message because you are subscribed to the Google Groups "sage-notebook" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-noteboo...@googlegroups.com.
Visit this group at http://groups.google.com/group/sage-notebook.
For more options, visit https://groups.google.com/d/optout.

kcrisman

unread,
Apr 23, 2015, 9:06:41 PM4/23/15
to sage-n...@googlegroups.com, pokal-t...@elearning.physik.uni-frankfurt.de, sage-...@googlegroups.com, Andrey Novoseltsev, William Stein

 
we offer two machines located in our data center (Opteron 270 with 16GB RAM and .5TB HDD each) to host the Sage cell server.

Bye from Frankfurt/Germany,
Sven

Danke!

William, Andrey - habt ihr das gesehen?  I mean, did you see this offer? 

Andrey Novoseltsev

unread,
Apr 24, 2015, 1:03:21 AM4/24/15
to kcrisman, sage-n...@googlegroups.com, pokal-t...@elearning.physik.uni-frankfurt.de, sage-devel, William Stein
Yes, I've sent a request for details to Sven.

For two months there will be (in a week, I hope) also a GCE instance
running, we'll see what happens after that.

Andrey

Sven Köppel

unread,
Apr 24, 2015, 6:40:10 AM4/24/15
to sage-n...@googlegroups.com, pokal-t...@elearning.physik.uni-frankfurt.de, sage-devel, Andrey Novoseltsev, William Stein
Hallo Karl-Dieter,

maybe I forgot to tell some details about the offer. I just sent Andrey an email containing some technical stuff (hosting and login details), but I would like to send some lines to the mailing lists considering our institute.

Namely, I am a theoretical physicist at Goethe University, Frankfurt. I maintain a positive relationship to the federal center for scientific computing, and we plan to host SAGE on a bigger scale. For us, self-hosting (in terms of a self-hosted cloud service) means independency of GCE et al. Anyway, after working on SAGE for two years with 4 students, I got the feeling computing power is clearly less a problem than having the manpower.

Actually I never realized you speak german, Karl-Dieter :-D  Do you know if there is any other SAGE community in Germany?

Bye for now,
Sven


--

pang

unread,
May 21, 2015, 5:15:20 AM5/21/15
to sage-n...@googlegroups.com, pokal-t...@elearning.physik.uni-frankfurt.de, sage-...@googlegroups.com, wst...@gmail.com, novo...@gmail.com
Sorry for the late reply, but after some talks, the department of mathematics of the Universidad Autónoma de Madrid can offer a core i3, with a 256GB SSD and 30GB of RAM.

Andrey Novoseltsev

unread,
May 21, 2015, 7:50:58 PM5/21/15
to sage-...@googlegroups.com, sage-n...@googlegroups.com, pokal-t...@elearning.physik.uni-frankfurt.de, novo...@gmail.com, wst...@gmail.com
On Thursday, 21 May 2015 03:15:22 UTC-6, pang wrote:
Sorry for the late reply, but after some talks, the department of mathematics of the Universidad Autónoma de Madrid can offer a core i3, with a 256GB SSD and 30GB of RAM.

Great! The more servers we have the better for the load we can handle and overal resilence in case of hardware or legal issues.

If anyone in North America can provide a server instead of GCE that also would be fantastic. The current setup on GCE will cost about $1000/year after my trial is over and at the moment can't handle graceful updates as I've discovered yesterday. Tweaking configuration and fixing some bugs may improve the situation and I certainly will try to do it, but the fastest solution is to have a serer with more RAM.

Andrey

William Stein

unread,
May 21, 2015, 9:54:51 PM5/21/15
to Andrey Novoseltsev, sage-devel, sage-notebook, pokal-t...@elearning.physik.uni-frankfurt.de
On Thu, May 21, 2015 at 4:50 PM, Andrey Novoseltsev <novo...@gmail.com> wrote:
> On Thursday, 21 May 2015 03:15:22 UTC-6, pang wrote:
>>
>> Sorry for the late reply, but after some talks, the department of
>> mathematics of the Universidad Autónoma de Madrid can offer a core i3, with
>> a 256GB SSD and 30GB of RAM.
>
>
> Great! The more servers we have the better for the load we can handle and
> overal resilence in case of hardware or legal issues.
>
> If anyone in North America can provide a server instead of GCE that also
> would be fantastic. The current setup on GCE will cost about $1000/year

Andrey, did you hear about the new "pre-emptible" pricing that came
out two days ago? You get VM's that are exactly like normal, except
(1) they get randomly -- but safe -- shut down, and (2) they cost
about 70% less. That's a massive discount, and given the nature of
the cell server a random shut down is probably fine. I just wanted
to point this out...

> after my trial is over and at the moment can't handle graceful updates as
> I've discovered yesterday. Tweaking configuration and fixing some bugs may
> improve the situation and I certainly will try to do it, but the fastest
> solution is to have a serer with more RAM.
>
> Andrey



--
William (http://wstein.org)

Andrey Novoseltsev

unread,
May 22, 2015, 2:45:23 PM5/22/15
to sage-...@googlegroups.com, sage-n...@googlegroups.com, novo...@gmail.com, pokal-t...@elearning.physik.uni-frankfurt.de
On Thursday, 21 May 2015 19:54:52 UTC-6, William wrote:
On Thu, May 21, 2015 at 4:50 PM, Andrey Novoseltsev <novo...@gmail.com> wrote:
> If anyone in North America can provide a server instead of GCE that also
> would be fantastic. The current setup on GCE will cost about $1000/year

Andrey, did you hear about the new "pre-emptible" pricing that came
out two days ago?  You get VM's that are exactly like normal, except
(1) they get randomly -- but safe -- shut down, and (2) they cost
about 70% less.  That's a massive discount, and given the nature of
the cell server a random shut down is probably fine.    I just wanted
to point this out...

Yes, I looked into it but I don't think it is a good option at the moment.

Random restarts were quite frustrating to me last spring since I often get interacts ready in the beginning of the class and then return to them from time to time. If they stop working in the middle, it wastes class time and I imagine many other instructors may get irritated as well. The situation last spring was that compute servers were restarted every 12-24 hours (automatically at random times - whenever things stopped working), so hitting a restart during class once a week was quite likely. I've managed to increase uptime to several weeks on old machines and intentional restarts for upgrades were mostly transparent to users. But there are still some bugs that lead to bigger memory usage than necessary and on 2core/13GB RAM GCE instance it does show - restarts are happening after a few days rather than weeks and last upgrade was quite opaque for 30 minutes or so.

Back to pre-emptible machines: assuming that random shutdowns are very rare, like once a month, it may be OK. But any random shutdown will be very visible to users or interacts/linked cells and every 24 hour migration will be visible because of the current IP-address sticking. Plus sometimes it may not be possible to create such instances at all, presumably during "busy times" when everybody is working actively including users of SageMathCell.

German servers on the other hand are running fine (they get about 1/3 of the load with current DNS setup) and if we can get another 2 in Spain, I would be in favour of just shutting down GCE for now. May affect latency from US, but that's how the rest of the world including Europe was living when all servers were in Seattle ;-)

Andrey
Reply all
Reply to author
Forward
0 new messages