Hi,
I completely understand that you need some extra info, I was just
checking if there was room for improvement on your side.
My application and deltacloud are in the same server, the vsphere server
is another machine but located on the same network segment.
I use different strategies depending on what I want to show. This tool
I'm working on creates a set of 48 machines. When I want to show details
on one machine I access the instance details, I followed the same
strategy when I wanted to show a list with 48 machines but I realized
this was too slow so instead I get the whole list and I filter it
locally on my code. This is slightly faster than getting the details for
the 48 machines one by one but still too slow, we are talking of about
10 secs for listing 48 machines when using the deltaclouds web interface.
Right now I'm thinking of using some AJAX on my app in order to smooth
it a little bit but I'm afraid this won't be enough when the system is
on production and the list I'm getting from vpshere grows to the hundreds.
I haven't compared this response times with querying directly to the
vsphere server (that's why I chose deltacloud in the first place) but
I'll be happy to do it if you can give me some directions on how to do it.
Xavier
On 10/05/2012 08:06 PM, David Lutterkort wrote:
Hi Xavier,
On Fri, 2012-10-05 at 10:19 +0200, Xavier Naveira wrote:
I hope that I'm not completely offtopic here but I'm implementing a
django application that uses deltacloud to retrieve data from a vSphere
5.1 installation. Everything is working fine but I find that the time it
takes to get info about the instances and the instance list itself is
too long to be used in as "online polling" from my web application.
Maybe it's a bit naive of me to ask but, is there any plans on improving
performance of this queries? Maybe it depends on the VMWare hosts poor
performace... Is there any tip on caching or paralleling or something
else achievable in python to improve performance on my side?
what operations exactly are slow ? Are you listing all VM's or just
looking up the details of specific VM's ? Do you have any comparisons
between how long these operations take via Deltacloud and via the
vSphere API directly ?
Also, how do you have the various pieces deployed ? To avoid additional
network overhead, your Deltacloud server should either be running close
to your vSphere installation or close to your app in terms of network
topology.
It might well be that there are things we can do in the vSphere driver
to speed things up, but we'd need some more details.
David
