|
How can I control when detections occur? |
| Practically speaking, there is no real way that configuring the detection
interval can guarantee the time of day that a client checks in. With the default of 22 hours, the client will get a random checkin every 17-22 hours. That effectively slips the client approx 6 hours each day, and by the end of four days, the client has checked in at five different times of the day. For example: Day 1 - the client detects at noon, the next detection is set for approx 6am the next morning. Day 2 - the client detects at 6am, the next detection is set for approx midnight that night. Day 3 - the client detects at midnight, the next detection is set for approx 6pm that same day Day 3 - the client detects at 6pm, the next detection is set for approx noon the next day. Day 4 - the client detects at noon With a twelve hour detection cycle, the client will be checking in twice a day, but the time of checkins will float backwards by a couple of hours each day. With a six hour detection cycle, the client will be checking in four times a day, and the time of checkins will float backwards by about an hour each day. Regardless of the detection interval configured, the client will cycle back through the hours about every 5-6 days. It seems to me the way to determine the minimum affordable detection interval is to divide the total number of clients by 24 hours, and compare that to the number of clients the server can support per hour. Using Microsoft's guidance of 500 systems on WMSDE, that computes to about 20 systems hitting the server every hour around the clock. If the server can handle 40 sessions per hour, the detection interval could be reduced to 12 hours without impacting the load on the server. If the server can handle 80 sessions per hour, 6 hours would be a functional detection interval. On the other hand, what's the real expected advantage of a shorter detection cycle if updates are typically only approved once every thirty days. Is it worth the extra load on the server and network just to ensure a six hour response to update approvals when they come out. My thoughts are that the optimal detection interval is probably around 12 hours. This would pretty much ensure that a client executes a detection on the same day as the updates are approved, and within the time frame of the scheduled install. For example: Patch Tuesday updates are approved on Wednesday morning before noon. A twelve hour cycle gets all clients checked in sometime between noon and midnight. A 3am installation gets all clients updated overnight Wednesday. The default detection interval might require a second day to get all of the clients updated, for those clients who's scheduled installation time (3am) occurs before the client performs the next detection cycle after the updates are approved on Tuesday morning. |