Otherwise you have to restart BTS or at least break the RSL connection to apply the change.
On 30 May 2015, at 21:43, Alexander Chemeris alexander.chemeris@gmail.com wrote:
Otherwise you have to restart BTS or at least break the RSL connection to apply the change.
Yes we certainly should apply modifications at runtime but probably not by adding auto-update to each command. When I change LAC/CI/ms_max_power I don’t want to update the SI three times.
We really need to have a “commit”/“apply” kind of command that applies OML and RSL modifications.
holger
On Jun 2, 2015 9:26 AM, "Holger Freyther" holger@freyther.de wrote:
On 30 May 2015, at 21:43, Alexander Chemeris <
alexander.chemeris@gmail.com> wrote:
Otherwise you have to restart BTS or at least break the RSL connection to apply the change.
Yes we certainly should apply modifications at runtime but probably not
by adding
auto-update to each command. When I change LAC/CI/ms_max_power I don’t
want
to update the SI three times.
We really need to have a “commit”/“apply” kind of command that applies
OML and
RSL modifications.
I agree that transactional functionality would be nice and I had been thinking about this. But it's a big change which should be well planned and requires considerable effort to go through all commands and split configuration from application. One issue is that we'll need to create "shadow" registers for the non applied settings. E.g. in case of power control, the setting was actually applied at the BSC part of the control (it was using the setting variable directly), but was not propagated to the BTS party, which was really confusing from a user perspective.
In short - I agree that such feature would be nice, but I don't think this is a showstopper for this patch, because it just makes things more consistent.
Btw, control interface also applies the change immediately. Having vty and control to diverge is quite confusing.
-- Regards, Alexander Chemeris CEO Fairwaves, Inc. https://fairwaves.co
On 15 Jun 2015, at 21:01, Alexander Chemeris alexander.chemeris@gmail.com wrote:
Good Morning Alexander,
I agree that transactional functionality would be nice and I had been thinking about this. But it's a big change which should be well planned and requires considerable effort to go through all commands and split configuration from application. One issue is that we'll need to create "shadow" registers for the non applied settings. E.g. in case of power control, the setting was actually applied at the BSC part of the control (it was using the setting variable directly), but was not propagated to the BTS party, which was really confusing from a user perspective.
In short - I agree that such feature would be nice, but I don't think this is a showstopper for this patch, because it just makes things more consistent.
we will not be able to find an agreement here. The current behavior is not consistent (we have attributes that apply immediately, after a BTS restart or even only after the restart of the system). Moving one VTY command from one class to another doesn’t improve the consistency in any way. The same inconsistency remains in the system, we face the same issue in documenting the inconsistencies.
So when/how do we want to apply changes? Should all changes be abpplied after an “apply” command? How is it done by other vendors that offer a VTY like interface?
holger
The current behavior is not consistent (we have attributes that apply immediately, after a BTS restart or even only after the restart of the system).
...
So when/how do we want to apply changes? Should all changes be abpplied after an “apply” command? How is it done by other vendors that offer a VTY like interface?
I think it is sometimes tied into the admin lock/unlock functionality. So for example, a cell maybe 'locked' and the transmitter disabled, at which point parameters can be changed as required. Then when the cell is unlocked, the new parameter set is then applied before the transmitter is re-enabled.
For co-dependent parameters there may be a validation step prior to unlock to ensure that the whole configuration is consistent and valid, though I can't think of many such parameters in this situation.
An alternative method is to 'lock' a cell by leaving it transmitting but marking it as barred in the SI.
Such a locking scheme does raise some questions as to the granularity i.e. do we lock at a BTS, BSC, MSC level or each one. And therefore the vty needs to know which parameters can be changed according to which network entities are locked - though the command hierarchy could lend some description to that.
Kind Regards,
Mike
Hi Holger,
On Wed, Jun 17, 2015 at 2:04 AM, Holger Freyther holger@freyther.de wrote:
On 15 Jun 2015, at 21:01, Alexander Chemeris alexander.chemeris@gmail.com wrote: I agree that transactional functionality would be nice and I had been thinking about this. But it's a big change which should be well planned and requires considerable effort to go through all commands and split configuration from application. One issue is that we'll need to create "shadow" registers for the non applied settings. E.g. in case of power control, the setting was actually applied at the BSC part of the control (it was using the setting variable directly), but was not propagated to the BTS party, which was really confusing from a user perspective.
In short - I agree that such feature would be nice, but I don't think this is a showstopper for this patch, because it just makes things more consistent.
we will not be able to find an agreement here. The current behavior is not consistent (we have attributes that apply immediately, after a BTS restart or even only after the restart of the system). Moving one VTY command from one class to another doesn’t improve the consistency in any way. The same inconsistency remains in the system, we face the same issue in documenting the inconsistencies.
Please note that the current behavior of this command is that it already applies the change immediately by setting bts->ms_max_power. This value is then used at rsl_rx_chan_rqd() to initialize lchan->ms_power. But it does not apply it to SIs sent by the same BTS, which makes the effect of the command inconsistent. The commit I suggest just makes the behavior of the command consistent.
Please also note that it makes complete sense to apply this value immediately when you're doing live cell tuning.
On 24 Jun 2015, at 00:27, Alexander Chemeris alexander.chemeris@gmail.com wrote:
Dear Alexander,
Please note that the current behavior of this command is that it already applies the change immediately by setting bts->ms_max_power. This value is then used at rsl_rx_chan_rqd() to initialize lchan->ms_power. But it does not apply it to SIs sent by the same BTS, which makes the effect of the command inconsistent. The commit I suggest just makes the behavior of the command consistent.
we have completed the circle now and I don’t want to run another round. The ms_max_power aspect is annoying but it shows that we need to progress and look at the bigger picture. Are you willing to do that?
holger
Hi Holger,
On Wed, Jun 24, 2015 at 10:50 AM, Holger Freyther holger@freyther.de wrote:
On 24 Jun 2015, at 00:27, Alexander Chemeris alexander.chemeris@gmail.com wrote: Please note that the current behavior of this command is that it already applies the change immediately by setting bts->ms_max_power. This value is then used at rsl_rx_chan_rqd() to initialize lchan->ms_power. But it does not apply it to SIs sent by the same BTS, which makes the effect of the command inconsistent. The commit I suggest just makes the behavior of the command consistent.
we have completed the circle now and I don’t want to run another round. The ms_max_power aspect is annoying but it shows that we need to progress and look at the bigger picture. Are you willing to do that?
In the spirit of scratching the itch, I do need the ms_max_power patch, as it fixes an obvious issue and I want to share it with the community. I'm fine maintaining it шт a branch if it's not good enough for the master.
OTOH I don't have an itch with the lack of “commit”/“apply” functionality, so it's hard for me to discuss it - I don't know requirements and use cases. I see a theoretical value of it, but relying on theoretical values is a sure way to over-engineer in my experience. If you have a real need for it, then you're probably much more qualified to look at it.
I hope that's fair enough.
-
Alexander Chemeris -
Holger Freyther -
Mike McTernan (wavemobile)