All Posts

I just want to run those two standalone VMs(java and standby) that dont need the client software, but it seems like I still need to have the mail server up running to have the prime client join the mail server domain in order to run. My questions are 1. Do Java and standby VM have to reside on a Microsoft Windows member server joining the domain controller of the mail server? Can they be only a workgroup server? 2. Does the prime client have to be a member server of the mail server? 3. The VMMark benchmark guide kept mentioning '(refer to ''Hosts Files" on page xx) in many places, (e.g page 59 bottom line) what page number that page xx is refering to? Thanks.

I have run into a problem during my build of the web-server VM. Note, I am using the 20061219 version of VMmark & guide. I get errors during the "make" in step 7 of APC install process. I have version APC-3.0.12. \----- /usr/local/APC-3.0.12/apc_compile.c: In function 'my_prepare_op_array_for_execution': /usr/local/APC-3.0.12/apc_compile.c:1904: error: union has no member named 'jmp_addr' /usr/local/APC-3.0.12/apc_compile.c:1905: error: union has no member named 'jmp_addr' /usr/local/APC-3.0.12/apc_compile.c:1913: error: union has no member named 'jmp_addr' /usr/local/APC-3.0.12/apc_compile.c:1914: error: union has no member named 'jmp_addr' \----- I have reinstalled php (php-4.4.3) and APC-3.0.12 twice and get the same errors. What does this mean? What might I have done wrong? And, of course how do I fix it?

First. I am very very happy that this benchmark idea has finally come to fruition. As a newcomer to the process I wanted to make a few comments about the build/installation process. As was noted in the first posting to this forum the build process is rather lengthy and complicated. I fully recognize that complexity is a given, considering the various types of systems and workloads being put together. However, it seems to me that there are some places that we can reduce the effort a bit. In particular the steps of building the 7 systems (yes 7, 3 windows VMs + 3 linux VMs + 1 windows Client) per tile can be resequenced to allow more leveraging of your effort. Obviously using templates is already recommended in the documentation and is definitely highly desirable. For both sets of VMs (Windows and Linux) if you use a bit of thought and planning you can do just one OS install instead of 3. Create your VM, install the OS, install all the common software and do all of the common tweaking. Then clone that VM into the additional two VMs. Then you would do the specific workload installation steps in each VM before creating VM templates. This alone cuts out 1-2 days of effort and elapsed time if for nothing but just waiting for W2K3 or RHEL to finish the install process.

It would be nice to have VM appliances available for many of the workloads described in the test. This would reduce the possible variation in the results due to configuration differences within the VMs (patch levels, versions, etc). I realize this might not be possible with the Windows-based VMs, but how about the RHEL VMs? It is probably too much to ask that the vendors in question provide temporary benchmark licenses for such pre-configured appliances in order to provide standardized building blocks for the metrics. The investment in time to create these VMs is pretty significant and seems to require a bit of cross-platform expertise. Not that this is a bad thing, but it may limit the number of folks able to run the tests. That being said, I can tell that a lot of time and energy has gone into creating this benchmark -- and document. My two cents -- thrown in to seed the pot. Doug