headers
Internet-Drafts | 23 Oct 2006 00:50
Picon
Favicon

I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-08.txt

A New Internet-Draft is available from the on-line Internet-Drafts 
directories.
This draft is a work item of the IP over InfiniBand Working Group of the IETF.

	Title		: Definitions of Managed Objects for InfiniBand Channel Adapters (CA)
	Author(s)	: S. Harnedy
	Filename	: draft-ietf-ipoib-channel-adapter-mib-08.txt
	Pages		: 22
	Date		: 2006-10-22
	
InfiniBand Architecture (IBA) specifies a high speed, channel based,  
   switched fabric architecture that delivers scalable performance in  
   data centers. 

   This memo defines a portion of the Management Information Base (MIB)  
   for use with network management protocols in the Internet community.   
   In particular, it defines objects for managing InfiniBand Channel  
   Adapters (CA).

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-ipoib-channel-adapter-mib-08.txt

To remove yourself from the I-D Announcement list, send a message to 
i-d-announce-request <at> ietf.org with the word unsubscribe in the body of 
the message. 
You can also visit https://www1.ietf.org/mailman/listinfo/I-D-announce 
to change your subscription settings.

Internet-Drafts are also available by anonymous FTP. Login with the 
username "anonymous" and a password of your e-mail address. After
(Continue reading)

Permalink | Reply |
headers
bill | 6 May 2004 20:07

RE: I-D ACTION:draft-ietf-ipoib-architecture-04.txt

Vivek has made a few clarifying chagnes around the phrases that IPv6 addresses are used as IB addresses that the
IESG asked for.

Please look this over and make any comments  by next Thursday (May 13th) so it can be resubmitted back to the
IESG for approval

Bill Strahm
co-chair
> -------- Original Message --------
> Subject: [Ipoverib] I-D ACTION:draft-ietf-ipoib-architecture-04.txt
> From: Internet-Drafts <at> ietf.org
> Date: Thu, May 06, 2004 8:26 am
> To: i-d-announce <at> ietf.org
> Cc: ipoverib <at> ietf.org
>
> A New Internet-Draft is available from the on-line Internet-Drafts
> directories.
> This draft is a work item of the IP over InfiniBand Working Group of
> the IETF.
>
> 	Title		: IP over InfiniBand(IPoIB) Architecture
> 	Author(s)	: V. Kashyap
> 	Filename	: draft-ietf-ipoib-architecture-04.txt
> 	Pages		: 24
> 	Date		: 2004-5-4
>
> 	InfiniBand is a high speed, channel based interconnect between
>         systems and devices.
>
>         This document presents an overview of the InfiniBand
(Continue reading)

Permalink | Reply |
headers
Internet-Drafts | 9 Mar 2004 21:57
Picon
Favicon

I-D ACTION:draft-ietf-ipoib-subnet-manager-mib-00.txt

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the IP over InfiniBand Working Group of the IETF.

	Title		: Infiniband Subnet Manager(SM) MIB
	Author(s)	: E. Tsang, C. Yang, C. Yang
	Filename	: draft-ietf-ipoib-subnet-manager-mib-00.txt
	Pages		: 105
	Date		: 2004-3-9
	
This memo defines a portion of the Management Information Base (MIB)  
   for use with network management protocols in the Internet community.   
   In particular, it defines objects for the InfiniBand Subnet Manager(SM) 
   (http://www.infinibandta.org).

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-ipoib-subnet-manager-mib-00.txt

To remove yourself from the IETF Announcement list, send a message to 
ietf-announce-request with the word unsubscribe in the body of the message.

Internet-Drafts are also available by anonymous FTP. Login with the username
"anonymous" and a password of your e-mail address. After logging in,
type "cd internet-drafts" and then
	"get draft-ietf-ipoib-subnet-manager-mib-00.txt".

A list of Internet-Drafts directories can be found in
http://www.ietf.org/shadow.html 
or ftp://ftp.ietf.org/ietf/1shadow-sites.txt

Internet-Drafts can also be obtained by e-mail.
(Continue reading)

Permalink | Reply |
headers
H.K. Jerry Chu | 9 Jul 2003 23:57
Picon

RE: I-D ACTION:draft-ietf-ipoib-architecture-02.txt

Bill,

I have reviewed this draft and thought it looks ok. One possible issue
is that the draft includes quite a bit of discussions on the
"rationale" aspect of the current IPoIB design, consequently
duplicating many materials from the other IPoIB drafts. 

Since the draft will be informational only, perhaps this redundancy
is not a problem?

Jerry

------------------------------------------------------------------------

I have received no issues with this draft.  Can I get people that have
read the architecture draft to ping me or the list and say so.

This is the end of the last call for the architecture, encapsulation and
link drafts.  As soon as possible (hopefully before Vienna) I will
contact Thomas and have him get these on the IESG agenda

Bill

-----Original Message-----
From: ipoverib-admin <at> ietf.org [mailto:ipoverib-admin <at> ietf.org] On Behalf
Of bill
Sent: Tuesday, June 17, 2003 11:44 PM
To: ipoverib <at> ietf.org
Subject: RE: [Ipoverib] I-D ACTION:draft-ietf-ipoib-architecture-02.txt
(Continue reading)

Permalink | Reply |
headers
Rosenstock, Hal | 24 Sep 2002 20:18

RE: I-D ACTION:draft-ietf-ipoib-subnet-mgmt-agent-mib- 03.txt

Hi Sean,

Looks good. Just some nits to pick below. The biggest issue looming over
this is that by the time this MIB comes out or perhaps even the next
version, IBA will be at 1.1 and there will need to be some modifications.
Nothing major.

1. p.9 ibSMANodeType indicates other (4) but the description indicates
reserved for 4.

2. p.11 ibSmaNodeLid can only go up to 48K, perhaps be permissive (0xFFFF).

3. p.15 ibSmaMulticastFdbTableNum goes up to 16382 rather than 16383 as the
permissive LID is at 65535.

4. p.16-17 ibSmaSwDefaultPort, ibSmaSwDefMcastPriPort, and
ibSmaSwDefMcastNotPriPort should all go up to 255 rather than 254 as this is
a valid invalid port.

5. p.48 ibSmaRandomLMC is limited to 5 rather than 7 despite being 3 bits.

6. p.57 ibSmaPortLinkStateChange DESCRIPTION should read
"128:Link State of at least one port of switch at <LIDADDR> has changed." 
rather than
"128:Link State of at least one port has changed." 

I also get the following error putting this through mosy:
capability ibSmaFullRouterCACompliance:
    syntax error
(Continue reading)

Permalink | Reply |
headers
Larry Butler | 19 Aug 2002 18:33

RE: I-D ACTION:draft-ietf-ipoib-ibif-mib-03.txt

All,

I'd like to share one more observation about the IPoIB Interface MIB from an
implementer's perspective.

Counters (e.g. PortXmitPkts, PortXmitData, etc.) are implemented in the IBA
Silicon because of performance and because RDMA operations are transparent
to kernel mode device drivers.

An ifIndex does not always map one-to-one with a Channel Adapter port
because there may be multiple IB Partitions running over a Channel Adapter
port. Most IBA Silicon does not track statistical counters based upon IB
Partition, rather, they are based upon a physical port.

My point is, the Interface MIB counters are not going to be accurate when
multiple partitions are running over a Channel Adapter port for most IBA
chips as the counters will be counting statistics from a port level view and
not from a partition view.

Larry Butler

-----Original Message-----
From: Internet-Drafts <at> ietf.org [mailto:Internet-Drafts <at> ietf.org]
Sent: Monday, July 01, 2002 3:36 AM
Cc: ipoverib <at> ietf.org
Subject: [Ipoverib] I-D ACTION:draft-ietf-ipoib-ibif-mib-03.txt

A New Internet-Draft is available from the on-line Internet-Drafts
directories.
This draft is a work item of the IP over InfiniBand Working Group of the
(Continue reading)

Permalink | Reply |
headers
Larry Butler | 14 Aug 2002 19:35

RE: I-D ACTION:draft-ietf-ipoib-ibif-mib-03.txt

All,

I am having a problem implementing the ibIfVLTrafficTable and the MIB-II
ifTable counters.

The reason I am having difficulties is that the IBA chip that we are using
defines port based performance counters that must be set up in a sampling
mode just like a PMA would set up sampling of these counters.

Here is an experpt from the chip data sheet:
"The management application first determines the node's sampling
capabilities via the PerformanceGet
performance-management method in the PortSampleControl attribute. The
performance-management agent
responds with the number and width of available counters, the quantities
that can be sampled, and the basic
time intervals. The application can use this information to compute the
maximum test interval that does not
cause the counters to overflow. The management application initiates the
PerformanceSet performance-management
method in the PortSampleControl attribute. If another application is using
the sampling mecha-nism,
the management application is notified about it via the
PerformanceGetResponse performance-management
method. Once a sample is started, the application waits until the sampling
is completed and
initializes the PerformanceGet method in the PortSampleResult attribute to
get the sampling result."

SNMP expects these counters to be continously counting, not setup prior to
(Continue reading)

Permalink | Reply |
headers
Rosenstock, Hal | 27 Jul 2002 00:10

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt

Hi Bill,

Not even all managers register themselves. When I saw your last response on
verbs (and standard interfaces), I could see the rational here. It was just
a thought on things we might want to know about a CA, but unfortunately the
information is from many different places, some of which may not be
standardized
in IB. Does that mean it shouldn't be in an IETF MIB ?

Also, your point about an SA MIB reminds me, whatever did happen to the SM
MIB ?
Is it out of scope with the current charter ?

-- Hal

-----Original Message-----
From: billiam33 <at> attbi.com [mailto:billiam33 <at> attbi.com]
Sent: Friday, July 26, 2002 5:32 PM
To: Rosenstock, Hal
Cc: ipoverib <at> ietf.org
Subject: RE: [Ipoverib] I-D
ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt

I think the reason for this is that the CA itself does 
have a mechanism to advertise the SM via it's on the IB 
subnet  query mechanisms, but other managers are assumed 
to advertise as ServiceRecords with the SA. Now if we 
had a SA MIB, I suppose we could look there for a few
items of interest..... ;^ )
(Continue reading)

Permalink | Reply |
headers
billiam33 | 27 Jul 2002 00:38
Picon

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt


There are instances that the location of a manager is not
desired to be advertised, as in the case of a B1 
implementation of, say, a security service. Yes, that is 
esoteric, but some companies have expressed the desire 
to do something like this. To each his own.

The managers in IB are somewhat relative -- there could 
be a number of device managers, for example, which
are defined relatively to each port via settings in
each Manager classes ClassPortInfo attribute. So the
advertisement is on the lines of "I am your 
device manager". If one was paranoid, one could
choose to just tell the device agent but not the
SA as to a device manager's location.

Personally I would find an SA MIB useful. There was
an SM MIB proposed by TopSpin, this was dropped, 
apparently, not clear on the particulars. It was
not in Sean's charter presentation. An SA MIB
would provide a fairly complete mapping of the IB  
fabric universe, and also provide hours of entertainment 
for those writing all of the table mechanisms:-)

There was also a Device Management MIB, I have 
converted this to SMIv2, but have not republished 
as I had not gotten a lot of feedback on it. There
was also no charter for this, either. 

--
(Continue reading)

Permalink | Reply |
headers
billiam33 | 26 Jul 2002 23:31
Picon

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt

I think the reason for this is that the CA itself does 
have a mechanism to advertise the SM via it's on the IB 
subnet  query mechanisms, but other managers are assumed 
to advertise as ServiceRecords with the SA. Now if we 
had a SA MIB, I suppose we could look there for a few
items of interest..... ;^ )

--
Bill Swortwood 
Sr. Software Architect                                
OmegaBand Inc.
(512) 401-8208 X144 (Office)
(512) 401-8268      (Fax)
(512) 431-2876      (Cell)
9020-I Capital of Texas Highway North,
Suite 270 Austin, TX. 78759
Email : billiam <at> omegaband.com
Web:  http://www.omegaband.com
Obligatory Famous Quote : "It's all just code..."
> Hi again,
> 
> I have another comment on the CA MIB.
> 
> I am not sure why the SM is singled out (ibCaHasSubnetManager). Indication
> of whether any of the other optional managers (PerfMgr, BM, DM, ...) are
> supported should also be indicated.
> 
> -- Hal
> 
>
(Continue reading)

Permalink | Reply |
headers
Rosenstock, Hal | 26 Jul 2002 18:05

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt

Hi again,

I have another comment on the CA MIB.

I am not sure why the SM is singled out (ibCaHasSubnetManager). Indication
of whether any of the other optional managers (PerfMgr, BM, DM, ...) are
supported should also be indicated.

-- Hal
Permalink | Reply |
headers
Rosenstock, Hal | 26 Jul 2002 16:00

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt

Hi,

I have a few comments on the Channel Adapter MIB.

Specific comments on what is in the MIB:

It appears there is one port GUID per CA (ibCaPortGuid) but isn't this per
CA port ? Should this be moved 
into ibCaPortAttrInfo ?

In ibCaPortAttrInfo, rather than 3 separate truth values for type of
physical interface, why not one enumerated value with the three types ? 

As multipathing is required, by ibCaSupportsMultipathing, I presume you are
referring to LMC not 0 as it appears that is included in the IBA definition
of multipathing.

Isn't ibCaValidatesInPktDlid a requirement so why is this needed ?

What about the optional inbound PKey checking ? Shouldn't there be
ibCaInboundPkeyChecking ?

Also, a couple of comments on what might be added:

There are other CA parameters which could be added. Things like number of
QPs, number of WQEs, number of CQs supported, number of EECs, max message
size, class of fault behavior, ... If this is a direction which is of
interest, I can provide a better list of these.

Will ibCaSystemImageGuid be added in for IBA 1.1 ? (There may also be some
(Continue reading)

Permalink | Reply |
headers
billiam33 | 26 Jul 2002 19:52
Picon

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt


Hello Sean and Larry - 

Just 2 brief comments marked [whs].
 --
Bill Swortwood     Sr. Software 
Architect                                
OmegaBand Inc.
(512) 401-8208 X144 (Office)
(512) 401-8268      (Fax)
(512) 431-2876      (Cell)
9020-I Capital of Texas Highway North,
Suite 270 Austin, TX. 78759
Email : billiam <at> omegaband.com
Web:  http://www.omegaband.com
Obligatory Famous Quote : "It's all just code..."
> Dear IPoIB WG,
> 
> I have a couple of very minor comments on the Channel Adapter MIB in
> addition to those I made earlier.
> 
> I'd like to suggest that ibCaHasSubnetManager (OBJECT-TYPE ::= {
> ibCaAttrInfo 14 }) be changed to ibCaSupportsSubnetManager. The DESCRIPTION
> section says "Flag that indicates whether this CA supports a Subnet Manager
> (SM) instance." From an implementation perspective, host based systems may
> not be able to easily tell if they "have" a Subnet Manager running as this
> is a process that may not be visible at the CA level. But they will know if
> they have the capabilities to run a Subnet Manager on the CA.

[whs] The CA via the verbs layer is set by the SM to advertise
(Continue reading)

Permalink | Reply |
headers
Larry Butler | 26 Jul 2002 17:58

RE: I-D ACTION:draft-ietf-ipoib-channel-adapter-mib-02 .txt

Dear IPoIB WG,

I have a couple of very minor comments on the Channel Adapter MIB in
addition to those I made earlier.

I'd like to suggest that ibCaHasSubnetManager (OBJECT-TYPE ::= {
ibCaAttrInfo 14 }) be changed to ibCaSupportsSubnetManager. The DESCRIPTION
section says "Flag that indicates whether this CA supports a Subnet Manager
(SM) instance." From an implementation perspective, host based systems may
not be able to easily tell if they "have" a Subnet Manager running as this
is a process that may not be visible at the CA level. But they will know if
they have the capabilities to run a Subnet Manager on the CA.

There are typographical errors in the REFERENCE section of
ibCaSupportsMultipathing (OBJECT-TYPE ::= { ibCaPortAttrInfo 6 }. The
REFERENCE section should read, "InfiniBand Architecture Release 1.0.a Vol 1.
Section 17.2.1.3 and Table 255. Also, Section 7.11.1 Multipathing
Requirements on End Node."

There are typographical errors in the REFERENCE section of
ibCaValidatesInPktDlid (OBJECT-TYPE ::= { ibCaPortAttrInfo 7 }. The
REFERENCE section should read, "InfiniBand Architecture Release 1.0.a Vol 1.
Section 17.2.1.3 and Table 255."

There are typographical errors in the REFERENCE section of
ibCaMaxGidsPerPort (OBJECT-TYPE ::= { ibCaPortAttrInfo 8 }. The REFERENCE
section should read, "InfiniBand Architecture Release 1.0.a Vol 1. Section
17.2.1.3 and Table 255." I'd also like to suggest that adding an additional
statement to the DESCRIPTION (for clarification), "Each port will support at
least one (1) GID." Which brings me to another suggestion, that perhaps the
(Continue reading)

Permalink | Reply |
headers
Rosenstock, Hal | 25 Jul 2002 17:19

RE: I-D ACTION:draft-ietf-ipoib-ibif-mib-03.txt

Hi,

I have a few light weight (low intensity) comments on the latest IB
interface MIB.

On p.8, the example for ifName might be better as (e.g. ib0) rather than
(e.g. lan0)

p.15 What is the purpose of:
ibIfNotifications OBJECT IDENTIFIER ::= { ibIfMIB 2 } ? I do not see this
anywhere else in the MIB. Are there any notifications ?

p.21 VL in the names should be consistent.

IbIfVLTrafficEntry ::= SEQUENCE {
    ibIfVlTrafficIfIndex              InterfaceIndex,
    ibIfVlIndex                       IbVirtualLane,
    ibIfVLOutOctets                   Counter64,       
    ibIfVLOutPkts                     Counter64,       
    ibIfVLInOctets                    Counter64,       
    ibIfVLInPkts                      Counter64
}

so ibIfVlTrafficIfIndex should be ibIfVLTrafficIfIndex, and ibIfVlIndex
should be ibIfVLIndex.

p.22 Should there be a ibIpoibLinkLayerAddrNumber so you can know how many
ibIpoibLinkLayerAddrEntry in table rather than walking/iterating over the
table until you get no such entry ?
(Continue reading)

Permalink | Reply |

Gmane