Dear DAT members, Since Yahoo no longer archives attachments, please, use file area to upload attachments. I will start by uploading the next DAT errata list...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Aug 1, 2003 7:55 pm
2509
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Aug 4, 2003 6:58 pm
2510
I have some thoughts on possible evolutions of the DAT object model to support some iWARP features while maintaining IB compatibility. An essay can on the...
We will have a DAT Collaborative conference call this Wednesday August 6, 1:00-2:00pm EDT (10:00am-11:00am PDT). Moderator: Arkady Kanevsky Phone: 888-827-8686...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Aug 5, 2003 6:53 pm
2513
I am happy to report that Siliquent Technologies has joined DAT Collaborative. Welcome to DAT train. Arkady Kanevsky, Chair of DAT Collaborative Arkady...
This posting proposes a construct that could allow for support of receive buffer sharing of the type proposed via various mechanisms for iWARP. The key new...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Aug 6, 2003 6:15 pm
2516
Errata 105 and 108 have been approved. Errata 111 is scheduled for next meeting vote (OS Wait). There are no other header erratas on the list for now. Arkady ...
I am happy to report that Chelsio Communications has joined DAT Collaborative. Welcome to DAT train. Arkady Kanevsky, Chair of DAT Collaborative Arkady...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Aug 9, 2003 4:19 pm
2520
Caitlin, Receive buffer sharing seems like a useful scalability aid, but not sure how this is implemented. Am I correct in assuming the scope of a buffer...
... The scope would be a single address space, and a single protection domain. If you differentiate between Local and Remote Protection Domains the scope would...
Dunno if this has been discussed already, but: The RDMAC Verbs spec limits the RDMA READ work request sink address to 1 scatter gather entry. IE a single STag,...
... A stretching point might be more accurate. A Provider *could* simply break up the SGL into smaller RDMA Reads. The tricky part is pacing them in the event...
The limitation is that the RDMA READ must be restricted to a virtually contiguous buffer. The buffer itself may represent multiple physically contiguous...
I understand that an STag/TO/Length can actually describe a scatter-gather list of physical pinned pages. But the DAPL API allows a _list_ of STag/TO/Lengths...
... Probing a bit deeper, I am beginning to agree that this is a serious conflict between IB semantics and the RDMAC semantics. The root difference in iWARP...
The RDMAC iWARP verbs allow for the resources associated with RDMA Reads to be either per-QP or shared. If my reading of the document is correct, this is...
... I believe option a) is the path of least resistance. However, we need to hear from the DAPL user community on this issue: Do existing and soon-to-be DAPL...
Haven't seen any responses from the DAPL user community to Steve's question below. It poses an interesting application design issue, i.e. do application's...
... I realized that I omitted "c", which would be c) Add a provider attribute allowing the consumer to determine how many RDMA Read credits will be required to...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Aug 26, 2003 3:57 pm
2534
Steve, Currently DAT specifies that posting Send, RDMA Read, Write and RMR Bind will fail synchronously. I agree that this not really needed. We can leave it...