Astrad Version 10 Poll Results
André Guimarães, 2012-01-11
Switzernet
This document compiles all voting information of the migrated servers DK1, FR3, FR13.
Results without the neutral votes
Detected Problems
Some problems were detected during the migrations:
- Some customers have NAT issues which Porta-SIP overcomes while Astrad does not. With Astrad v10 all outgoing calls for these users drop after 30 seconds and they are unable to receive calls. Some have been asked to disable this option in the router. Others (4 detected at the moment) use Netgear WNR2000 or Netgear WNR1000 which has known SIP ALG problems and in some firmwares is not possible to disable the option. These customers have been moved temporarily to the original dk1.youroute.net which has now the name sip100.youroute.net while a permanent fix in Astrads is not developed.
- During the Sunday of 11th December 2011 the Master stopped working. The problem started during the morning and it was solved only at 6PM. During this time Astrad authentication was made using DBA. Due to this all calls took at least 10 seconds to start (DBA authentication is only tried after the Radius timeout) and some services did not work as expected. Follow me for instance does not work completely and there are other known problems that were not solved in Astrad v10 while Master is down.
- Also when Master is down we have a Billing problem that makes the same call to be billed multiple times. Due to this some customers received warnings that their account balance limit add been surpassed thus being prevented from making calls. Due to propagation issues (it took two days for all Radius packets to be transmitted) the problem was only completely solved on the 14th.
- Some customers (at least two complained to support) were unable to make calls due to a strict match in the field “Authorization Digest” by the script /etc/astrad/script/agi-rad-auth.pl. Some phones don’t use the same capitalization and due to that the match failed and the customers heard the message “Invalid account”. This was corrected and added to Astrad v10+.
- Customers using Linksys SPA3XXX (one complained to migration support email) could not here ring tone after hearing the “Free call” message. These models were not compatible with the current Asterisk version Playback application without answering the line. A correction was added to Astrad v10 to disable the playback of this message to Linksys SPA3XXX owners.
- An issue with DBA prevented sometimes expired registrations from being removed. This expired information would also be in table multiple_ua which is replicated to Astrads. The size of this table would increase slowly and each time someone made a call to a phone in this table the CPU load of the Astrad would increase. Due to this sometimes the load of some servers reached 80% and caused bad quality problems. The problem is now solved in DBA v6 and later versions.
The three last problems were detected during the first week and first days of second week after DK1 migration and were all solved by the 8th December.
Server migration charts
DK1 Migration
This migration involved 237 customers. The customers were slowly migrated by DNS starting from 30/11/2011 as their DNS caches expired to an Astrad v10 named astrad17.switzernet.com. Some of them (5) have voted “No” before being migrated. The voting ended at 29th December 2011.
Day 2011/12/06 results (second day after sending voting email):
Day 2011/12/12 results:
Day 2011/12/19 results:
Day 2011/12/26 results:
Day 2012/01/03 results:
FR3 Migration
This migration involved 272 customers. The customers were slowly migrated by DNS starting from 05/12/2011 as their DNS caches expired to an Astrad v10 named astrad3.switzernet.com. The voting ended the 30th December 2011.
Day 2011/12/09 results:
Day 2011/12/16 results:
Day 2011/12/22 results:
Day 2011/12/29 results:
Day 2012/01/03 results:
FR13 Migration
This migration involved 510 customers. Due to the number of users involved this server’s customers were migrated to two different Astrad v10 servers: astrad14.switzernet.com and astrad15.switzernet.com. The customers were slowly migrated by DNS starting from 05/12/2011 as their DNS caches expired. The voting ended the 31st December 2011.
Day 2011/12/09 results:
Day 2011/12/16 results:
Day 2011/12/22 results:
Day 2011/12/27 results:
Day 2012/01/03 results:
Results through time
Notice the rapid increase in Negative votes during the period of 10 to 15. Afterwards all negative votes have stopped in all servers. Only DK1 server has more positive votes than neutral. All servers start and end with less negative votes than positive and neutral.
Other results
Voting percentage
In average, only 20% of customers have voted.
Almost half of the customers voted during the first day after the email was sent:
Final voting percentages for each server and total across all servers:
Results grouped by period
If we group the voting in periods, the first before the Master crash, the second including the Master crash and the 3 following days and a third with the remaining period we obtain some interesting results:
- More than 50% of users vote on the first days after receiving the email
- Half of the ‘No’s were cast in the first days after the mail being sent
- While Yes and Same have almost the same amount for each period the No percentage is increased in the period after the crash. This is especially noticeable in DK1 which had the No votes stabilized and FR3 which has half of the no’s during that period.
- Most ‘No’ votes were received in the first day of voting: 66%, 38.5% and 50% for DK1, FR3 and FR13 respectively.
- After the 14th the number of No votes stabilizes
Data from where the previous percentages were calculated.
Results without the neutral votes
Analyzing just the positive and negative votes we have in average almost 60% of positive votes. Except for FR13 results are around 60% for positive and 40% for negative votes. In FR13 the number of votes is almost the same for positive and negative.
Conclusions
In the last migrated servers and using the last voting format with 3 options, we obtained an average of 25% of negative votes while 35% of customers noticed an improvement. If we do not count with the neutral customers there are about 60% satisfied customers versus unsatisfied customers.
After an initial period where some problems where detected and that affected mostly DK1 (FR3 and FR13 were affected during only one or two days), the known bugs were solved. By the 9th all problems except for the NAT issue were solved.
The detected problems might explain most of the initial negative votes received as some of the affected customers could not register and other could not make or receive calls. Most of these customers have contact the support email we provided after each migration. Their problem was solved either by migrating to a temporary server or by making the necessary corrections in the Astrad and DBA code.
There seems to be a second period where the number of No’s has increased. This is the period after the Master crash in 11th December. During the crash period all customers where affected by degradation in speed (due to Radius authentication timeout) and some services did not work as expected as the DBA authentication code is not yet completed. Also in the following days there were problems with Billing which made some customers to be billed repeatedly by the same call and in some cases due to surpassing the balance limit prevent them from making calls.
In average only 25% of customers vote and 50% of all votes are made in the first day.
References
Voting results DK1:
http://ftp.switzernet.com/3/public/111229-astrad-stat/
Voting results FR3:
http://ftp.switzernet.com/3/public/111230-astrad-stat/
Voting results FR13: