Advanced network testing issues

[Ian (PassMark)]

I would guess 1 - 3 months.

Ian (PassMark)

[ComarkSW]

Additional information regarding the network test.

In the past, I wrote a program that handled multi-homed systems by using the socket bind function for the clients. This allows binding clients to a specific network interface. If this step is not performed the connect call uses the wildcard (zeroes) for the local IP address.

[Comark Corp]

Do you have any idea when you might be patching this? It would seem that we are shipping more boards with dual NICs and this is becoming a critical problem for us.

Thanks again.

Bump.

Even a guess would be better than nothing.

Thanks.

[Comark Corp]

Sorry for the delay in responding. We plan to release V5.2 at the end of this week, 2 February. Unfortunately, it may not include the Advanced Networking work I mentioned. I expect this may in a patch release after v5.2.

Sorry.

Best regards,
Ian

Thanks Ian.

Noted. I did test 5.2 and found out that the problem still exists. Do you have any idea when you might be patching this? It would seem that we are shipping more boards with dual NICs and this is becoming a critical problem for us.

Thanks again.

[Ian (PassMark)]

Sorry for the delay in responding. We plan to release V5.2 at the end of this week, 2 February. Unfortunately, it may not include the Advanced Networking work I mentioned. I expect this may in a patch release after v5.2.

Sorry.

Best regards,
Ian

[Comark Corp]

Hello Ian,

All we are looking to do is to test multi-homed systems. There are configurations of ours that may include 6 or more ethernet interfaces.

As for version 5.2 - do you have any idea when this may come out? If we need to do ISO9001:2000 validation ourselves (see my other new thread regarding this that I posted today) I would not want to have to validate 5.1 (1014) now, only to have to do it all over again when 5.2 is released.

Thanks again.

[Ian (PassMark)]

I will start by saying that the Advanced Network test was initially written for a specific purpose. As a result, the advanced network test is actually more complicated than it probably needs to be. We plan to modify the Advanced Network test in V5.2 due around the end of this month to remove some of this complexity (that we believe is not being used by our customers and that is just confusing).

As part of this V5.2 work will will investigate your post further. My initial comments are:
1) Yes an error should be raised.
2) This should raise an error now, there may be a case when then does not occur - we'll investigate further.
3) We will investigate reducing the CPU utilization of Endpoint. The best configuration would depend on exactly what you are testing. If you are testing the 2 NICs on the 'BurnInTest' system (and not the intermediate network gear or endpoint systems) then a single Endpoint would suffice, but 2 may allow a higher loading level. If you are looking at testing a many servers (eg. 10 servers running BurnInTest's advanced network test), then I would probably advise that you don't do this (if you do want to do this let me know as this is a candidate for removal from the BurnInTest Advanced Network test).

Best regardes,
Ian

[Comark Corp]

Hello, we are looking into testing systems with 2 or more NIC interfaces, and have found the following possible issues with the testing:

(UUT with two ethernet interfaces and running Windows 2000 SP4)
1. Loading .bitcfg with two ethernet interfaces (but one connection is unplugged), Bit silently deselects the disconnected interface. This should be flagged as an error.
2. Disconnecting (and leaving it this way) the second ethernet interface (as detected by Bit) while testing is not tagged as an error.
3. Endpoint.exe is CPU intensive - is there a recommended server configuration for testing many systems concurrently?

We are using version 5.1 (1014)

Thanks for any help.