Arjan noted (and Steve Carbonari of Intel didn't object) that Infiniband is not on the Intel chipset disclosure documents for CY04. Arjan feels that 10GB ethernet and similar technologies will supersede Infiniband.
*** Bug 108541 has been marked as a duplicate of this bug. ***
IBM specific info: Priority: 1 Status: low-latency connection. RAC interconnect and middle-tier to DB connections are sensitive to latency and speed. Description: Include Infiniband drivers Business Justification: Provides a faster interconnect.
*** Bug 109127 has been marked as a duplicate of this bug. ***
*** Bug 113652 has been marked as a duplicate of this bug. ***
Sun wants Infiniband as a backplane interconnect rather than as a replacement for 10GB ethernet. The request for Infiniband doesn't seem to be going away from any of the OEMs....
Adding HP
significantly doubtful for RHEL4. There are like 3 codebases, all really really bad.
Adding SLAC
From PM: Based on Arjan's comments, this one looks like it's too ill-defined and the code is in too poor a shape for RHEL4 consideration. If it turns out that Topspin and its 10G patch are accepted upstream and it passes Community code inspection quality, then we should reconsifder. For now , setting status to CLOSE=DEFERRED. Suggest we consider again for RHEL5.
I believe Larry's comment is unacceptable in this case. In this case the business justification is so compelling in the time frame of RHEL4 that we need to take a proactive stance on the issue and work with the people at TopSpin and the other OpenIB partners to bring their code into shape. For this issue Red Hat must act not as an aggregator of already exant technology but a contributor to the technology to bring it in line with acceptable upstream coding standards. Virtually every big partner that we have is asking for this feature and we have lost significant market advantage if we try to put this feature out to RHEL5.
I think that all parties will agree that Infiniband is not ready yet. The big thing that we need to decide as a company is are we going to be a market leader deploying innovative technology in response to customer needs or are we going to be a follower and just gather together the already extant technology and resell it. There is a very notable advantage in being the first to market and this should not be discounted. At the moment SuSE wins every single time in those races.
Right now this is stuck for RHEL4 and will not be delivered. We need hardware. We need engineers to implement the IB upstream, and port it back to RHEL3. This is a good candidate for a separate project for Red Hat to take on, but it needs a Marketing Reqts Doc or some other justification as well as a project plan and schedule. Will mark this as a key feature for RHEL4 timeframe, perhaps as a layered add-on. Could this be released as a vendor-supported driver?
First of all we need the low level drivers that support the hardware. We would like IP networking over IB. We run NFS and other protocols over this network interface. We also need text based IB configuration tools sort of like ifconfig, ethtool and ethdiag. MPI over IB is really important. low level kernel data transfer capabilities are also necessary. On top of this the CFS people will build the interface that lustre uses. API that the batch job scheduler will use to initialize the MPI interfaces for the job.
Along with the development of the software stack, they want an integrated testing suite and regression tests. http://www.llnl.gov/asci/pathforward_trilab/iba_rfp/OSSODA_IBA_RFPSOW_DRAFT.pdf http://www.llnl.gov/asci/pathforward_trilab/iba_rfp/OSSODA_IBA_RFP_EVAL_PREP_DRAFT.pdf The way that they put it in the RFP on page 5 is: 1) complete InfiniBand access layer including host channel adapter (HCA) device drivers, user and kernel space Verbs interface, communication management, resource management; 2) subnet management and subnet administration; and 3) HPC upper layer protocols for MPI, IPoIB, SDP, and Sandia Portals.
For reference regarding SDP: To: Ben Woodard <woodard> Subject: Re: [Fwd: (UPDATED) Issue #36125 ([RHEL 4] Implement Infiniband Support)[FEG-RHEL4/Feature Escalations Group/LLNL (HPC)]] X-Message-Flag: Warning: May contain useful information References: <1089328133.16745.1512.camel.gov> From: Roland Dreier <roland> Date: Thu, 08 Jul 2004 21:42:08 -0700 In-Reply-To: <1089328133.16745.1512.camel.gov> (Ben Woodard's message of "Thu, 08 Jul 2004 16:08:53 -0700") Message-ID: <52vfgxbxlr.fsf> User-Agent: Gnus/5.1006 (Gnus v5.10.6) XEmacs/21.4 (Security Through Obscurity, linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-OriginalArrivalTime: 09 Jul 2004 04:42:10.0152 (UTC) FILETIME=[1896AE80:01C4656F] X-RedHat-Spam-Score: 0 X-Evolution-Source: imap://woodard.redhat.com/ Content-Transfer-Encoding: 8bit This is something I wrote up a while ago about SDP: SDP intellectual property Microsoft believes that they own certain intellectual property relating to the Sockets Direct Protocol (SDP)[2]. Therefore, we are including the following disclaimer required by Microsoft's license in SDP source that relates to the implementation of the protocol: "This source code may incorporate intellectual property owned by Microsoft Corporation. Our provision of this source code does not include any licenses or any other rights to you under any Microsoft intellectual property. If you would like a license from Microsoft (e.g., to rebrand, redistribute), you need to contact Microsoft directly." We realize that this is incompatible with open source licensing, and we are working to find a more satisfactory solution, but for the time being we are forced to comply with Microsoft's license. Please make sure you have fully understood the implications of Microsoft's claims before you redistribute any of the SDP source that contains the above disclaimer. [2] <http://www.microsoft.com/mscorp/ip/standards/>
PM is managing this from an overall program standpoint, in/out/when. Greg is project prime and will report on the Infiniband plans.
Greg, please update with: 1. What we are doing for RHEL4 or RHEL4 U1 (specify release) Please provide a synopsis of the functionality we are delivering as part of the RFP. 2. If we are delivering on RHEL3 U4, please provide a new FZ and provide the same sort of info. Thanks, this is a hot item for LLNL as you know and we have a SLA to meet to respond to feature queries. Also, many other vendors are interested in this.
Per Rob Perkins, plan is to support Infiniband via Vendor Provided "top spin" driver.
Not in U1. Being reviewed for a later update. Still not mature enough upstream. Unlikely to be in U2.
This is not in U2. We are pursuing initial IB support in the context of U3.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2006-0013.html
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHEA-2006-0136.html
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2006-0016.html
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2006-0132.html