|
    |
     |
DiagnosticsThe future for this customer, is a continues surveillance the same problems, and the same bottlenecks - however now we will notify before they become a problem, and the users start to complain about long respond-times. Long respond-times - no errors on the server. 
A customer had problems with a SQLServer.A application is running on the SQLServer with 50 users from different locations. The users complain about long respond-times, and lost data. The supplier will not spend time on the problem, because the SQLServer is not running at full CPU load. (See the graph to the right). The it manager looks at the CPU load showing to the right, disc-activity, RAM usage, Event logs etc. but is unable to detect any problems. The graph show a total CPU load with an average of 50%. It drops around midday, and then increases until 16:00. First result with IN'Sync: 
IN'Sync added the SQLServer to monitor. We lack a base-line, and is limited in the amount of information's.Our first result after a week of data collection gave a single recommendation. The customer followed this advice, and the situation changed immediately - as shown in the left side. The graph only displays the SQLServeren load on the CPU, and we are gone from bellow 50% load, up to 90-95% - just for the SQLServer. At this stage, we have already improved the respond times for the users with 100%, and we eliminated the data loses. The end-result is a software supplier involved in the problems because we can show a pure CPU bottleneck from the SQLServer application. This is backed up but more measures including low disc activity, not a high ram usage, no hanging transactions etc. |