![]() The PVS image is on a local disk to the streaming server.File-libs x86_64 5.30-3.ph2 photon-updates 4.80M 5031416 For all these tests the PVS accelerator function has been disabled. I'm guessing the difference between the first time the image is streamed after a PVS server restart and the subsequent target reboots is the image is now cached in the PVS servers memory. 32 Seconds on Target device reboots without restarting the PVS server. 69 Seconds on Target device reboots without restarting the PVS server.Ģnd stage boot time - 85 seconds on first boot after the PVS servers has been restarted. So over all my tests the times have been consistently around the following:ġst stage boot time - 100 seconds on first boot after the PVS servers has been restarted. The only noticeable improvement is when the PVS server and target are on the same host and its using the 10GB virtual network which reduces both the first and second stage boot times by 20 seconds, but that's to be expected anyway. XenServer is a few seconds quicker at each stage but not by much. Having the PVS streaming server on our VMware or XenServer environment makes little difference to the boot times. I can now confirm that we have all the suggested optimisations applied to both the streaming servers and target devices and none have changed the time it takes for the first boot stage. I spent yesterday running benchmarks with different PVS settings and I am coming to the conclusion that what ever is causing the delay is being caused by the PVS software and not anything else in our environment. Can I ask what MTU size you are currently using? Have you raised a support call with Citrix over this? Thanks for providing this information, it is very useful. Our network team are currently looking at the traffic from their end and we are looking at disabling NetQueue on the esx hosts but it would be good to have a rough target figure to aim for. We have the majority of the recommended settings configured on the PVS streaming and target servers, such as disable task offload and configure the arp cache etc. I know the boot time is dependent on many factors but generally how long should the first stage of the boot process take? Would 1 minute 40 be considered normal or is this slow? After this stage the actual boot time of the image is reasonably quick. This part is taking roughly 1 minute 40 seconds to 2 minutes to complete. We then see the windows logo and there is long pause before we get the spinning white dots followed by anther long pause before the connection information, image name and write cache settings are displayed. The initial connection, download of the bootstrap file and logon to the PVS servers is very quick. We are using BDM partitions as we use PXE already for other network services. PVS image is Server 2016, fairly large image (85GB). PVS target servers are XenServer 7.1 CU1. PVS streaming servers are hosted on VMware 6.0. Reading through various blogs and articles I am beginning to suspect we may have an issue with the first stage of the boot process. The PVS images are working well and we are at the stage of optimising the performance as much as we can. We are in the process of migrating from XenApp 6.5 to XenDesktop 7.15 and we will also be transitioning from conventional clones to PVS images.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |