Re: iSCSI: Out of order commands

["Julian Satran" <Julian_Satran@il.ibm.com>]

Mallikarjun,

There are several other mechanisms that will have to be changed to allow 
for OOO.
Task abort relies on the fact that the task management request is 
delivered after the task to be aborted and on the same connection. 
Retry command cleaning relies on the fact that the "cleaning command" is 
shipped after the retried command.
And I suspect we may find others.

This thread is going nowhere IMHO for two reasons:

the proponents of OOO (Rod, Santosh and Bob Russell ) have faced us with 
an issue not a design - the validity of which is hotly contested
I and other authors do not seem to be willing to do all the work and make 
a design (or complete set of design changes) based on this sketch. Unlike 
other requests I could not see anything appealing enough to justify the 
time spent on doing it.

I suggest that Rod, Santosh and whoever else is willing to work with them 
put together a draft based on their ideas that should:

clearly state what is to be gained
indicate in sufficient what changes are needed in the current draft to 
accommodate the new design not only in command delivery but in all other 
areas (task management, recovery, logout etc.).

Once we have this draft I assure you that we will give it serious 
consideration.

Julo





"Mallikarjun C." <cbm@rose.hp.com>
Sent by: owner-ips@ece.cmu.edu
09-11-01 03:28
Please respond to "Mallikarjun C."

 
        To:     John Hufferd/San Jose/IBM@IBMUS
        cc:     <ips@ece.cmu.edu>
        Subject:        Re: iSCSI: Out of order commands

 

John,

Sorry, this note got a little longer than I would've liked, but....

I believe there are cases where OOO CmdSN handling is a
legitimate requirement on targets due to exception events -
    a) retransmitting a CmdSN on a command acknowledgement
      timeout (within-connection recovery class).  This manifests as
      an OOO CmdSN on the connection to a target if it didn't see
      the original copy due to a digest error.
    b) retransmitting the last few "lost" commands due to a connection
      failure on a new connection. If this new connection had already
      carried a CmdSN greater than these retransmitted commands
      (prior to connection failure), this again manifests as OOO CmdSN
      on the new connection to the target.

OTOH, I believe sending OOO CmdSNs on a connection as a
regular practice is counterproductive, since the target must continuously
re-order the initiator "optimization" leading to a zero-sum game.  I
would argue that the need to dispatch CmdSNs OOO due to immediate
data DMA (brought up by Rod) can be addressed by simple NIC
changes to prefetch data for the next command (or more simply use
unsolicited separate data PDUs, if negotiated).  [ You got to deal
with the case of all commands being writes anyway! ]

If we allow OOO CmdSNs on a connection (I'd advocate discouraging
it as a regular practice), I don't believe any of the stuff in error
recovery
breaks (nor does it affect the current reliance on ExpCmdSN).  Julian
perhaps can comment.
    - All the in-order assumptions are for DataSNs/R2TSNs/StatSNs, not
       for CmdSNs.
    - Any multi-connection session by definition must deal with OOO 
CmdSNs.
    - I belive that the current abort task scheme for immediate commands
      detailed in section 9.3 caters to OOO CmdSNs on a connection
      as well (we must be dealing with an immediate Abort arriving before
      the command today, since the command could have been hit with
      a digest error).

To summarize, here is what I suggested to Julian in a private email -

a)I suggest using a SHOULD for in-order dispatch of
  commands on a connection - for an initiator.

b)I suggest using a SHALL handle out-of-order commands
  on a connection - for the target (as Barry pointed out).

Hope that was useful.
--
Mallikarjun

Mallikarjun Chadalapaka
Networked Storage Architecture
Network Storage Solutions Organization
Hewlett-Packard MS 5668
Roseville CA 95747


----- Original Message -----
From: "John Hufferd" <hufferd@us.ibm.com>
To: <cbm@rose.hp.com>
Cc: <ips@ece.cmu.edu>
Sent: Thursday, November 08, 2001 1:40 PM
Subject: Re: iSCSI: Out of order commands


>
> Mallikarjun,
> Could you comment on the concept of OOO on the ErrorRecoveryLevel>0.  I
had
> thought that "in order delivery" was part of the detection of missing 
PDUs
> and needed for timely Recovery.  I was wondering if this changes the way
we
> would use the ExpCmdSN, etc.
>
> I think your opinions on this part of the OOO discussion would be
valuable.
> For example, how would you contrast the differences in detecting a 
problem
> and recovering from that problem etc., today vrs the OOO approach (if
any).
>
>
> .
> .
> .
> John L. Hufferd
> Senior Technical Staff Member (STSM)
> IBM/SSG San Jose Ca
> Main Office (408) 256-0403, Tie: 276-0403,  eFax: (408) 904-4688
> Home Office (408) 997-6136, Cell: (408) 499-9702
> Internet address: hufferd@us.ibm.com
>
>
> "Mallikarjun C." <cbm@rose.hp.com>@ece.cmu.edu on 11/07/2001 09:41:05 AM
>
> Please respond to cbm@rose.hp.com
>
> Sent by:  owner-ips@ece.cmu.edu
>
>
> To:   Santosh Rao <santoshr@cup.hp.com>, ips@ece.cmu.edu
> cc:
> Subject:  Re: iSCSI: Out of order commands
>
>
>
> Santosh,
>
> I have only one comment on your responses.
>
> > Even a single connection target *MUST* implement a scoreboard. The
> > reason being that it can see out-of-order arrival of commands due to
> > commands being dropped on digest errors. In such a case, it must block
> > further command processing until holes are filled.
>
> I made two convenient assumptions if you noticed, :-), one of which
> is that target forces session recovery on *any* error that it sees
> (ErrorRecoveryLevel=0) - including a dropped command due to a digest
> error.  With that assumption, a target can afford not to implement
> a scoreboard.
>
> As I said in a private note, I guess what primarily bothers me about
> OOO commands on a connection is that it requires the receiver to
> undo this "optimization" on its end - most notably on a single
> connection.  TCP experts may comment on how/if they dealt with a
> similar issue.
>
> OTOH, you had some valid comments on exceptions to ordering during
> connection recovery.  Perhaps we can move on by making Julian's
> proposed stipulation a SHOULD....
> --
> Mallikarjun
>
>
> Mallikarjun Chadalapaka
> Networked Storage Architecture
> Network Storage Solutions Organization
> MS 5668   Hewlett-Packard, Roseville.
> cbm@rose.hp.com
>
>
> Santosh Rao wrote:
> >
> > Mallikarjun,
> >
> > Some comments below.
> >
> > Regards,
> > Santosh
> >
> > "Mallikarjun C." wrote:
> > >
> > > Rod and Julian,
> > >
> > > This has been an interesting thread of discussion.  Some
> > > comments -
> > >
> > > 1.My first reaction was - allowing out-of-order command
> > >   transmission on the same connection deprives targets of
> > >   an implementation choice.  Targets which support only
> > >   single-connection sessions and only support session
> > >   recovery (reasonable assumptions in my mind) can no
> > >   longer afford *not to* implement a command scoreboard.
> >
> > Even a single connection target *MUST* implement a scoreboard. The
> > reason being that it can see out-of-order arrival of commands due to
> > commands being dropped on digest errors. In such a case, it must block
> > further command processing until holes are filled.
> >
> > Thus, there is no getting away from implementing a sequencer at the
> > target. Given this, I think it is unreasonable to restrict initiator
> > implementation flexibility by imposing a strict ordering requirement
> > within the connection.
> >
> > > 2.Any end-node efficiency that is sought to be achieved
> > >   by transmitting CmdSNs out-of-order from the initiator
> > >   would be lost on the other end-node, since the target
> > >   now must wait for re-ordering the commands.
> >
> > It has to handle this situation anyway to deal with holes caused by
> > digest errors. This scenario occurs even with initiators that issue
> > commands in order.
> >
> > >
> > > 3.The flipside is that out-of-order transmission saves
> > >   link badwidth (albeit at the expense of end-node efficiency),
> > >   compared to idling the link waiting for outbound DMA.
> > >   We have to determine if this is a reasonable trade-off.
> > >
> > > 4.I can see Rod's point that prefetching all immediate
> > >   data can be a burden on the NIC resources.  But, two
> > >   questions -
> > >         - could the NIC not use unsolicited separate data
> > >           PDUs in these cases? [ I realize that InitialR2T
> > >           has to be "no" to let it happen... ]
> > >         - could the NIC have a memory architecture that
> > >           allows data prefetching for the next command (so
> > >           this is a non-issue from the protocol perspective)?
> > >           This scheme incurs one DMA delay for every new
> > >           burst of commands.
> > >
> > > 5.Another (perhaps radical at this point) option is to do
> > >   away with immediate unsolicited data, to stick only with
> > >   separate unsolicited data.  I would personally be okay
> > >   with the choice, particularly if this feature (that
> > >   helps software implementations) starts making hardware
> > >   design complicated/expensive.
> > >
> > > So, to summarize -
> > >
> > > option                         immediate         allow
> > >                                data in spec?     out-of-order?
> > >
> > > (A) (5) above                  no                no
> > > (B) No real reason to do this. no                yes
> > > (C) (4) above                  yes               no
> > > (D) pros & cons (1), (2) & (3) yes               yes
> > >
> > > >From the arguments I heard so far, I am leaning towards
> > > option A, and option C in that order.
> > >
> > > Comments?
> > > --
> > > Mallikarjun
> > >
> > > Mallikarjun Chadalapaka
> > > Networked Storage Architecture
> > > Network Storage Solutions Organization
> > > MS 5668 Hewlett-Packard, Roseville.
> > > cbm@rose.hp.com
> > >
> > > Rod Harrison wrote:
> > > >
> > > > Julian,
> > > >
> > > >         I don't understand what you are proposing here, what do 
you
> mean by
> > > > "multiplexed" DMA?
> > > >
> > > >         The problem is that the DMAs take some time, the more 
there
> are
> > > > queued the longer the last DMAs queued take to complete. Some
> commands
> > > > require DMAs to complete before they can be sent, i.e. Writes with
> > > > immediate data, some commands do not, i.e. Reads and writes with 
no
> > > > immediate data. The iSCSI HBA wants to be able to send commands as
> > > > soon a possible, which for a read after a write can be before the
> > > > write's DMA has completed. Maintaining an ordered queue for 
commands
> > > > to be sent on the HBA is expensive and redundant since the target
> > > > already knows how to queue commands before committing them to its
> SCSI
> > > > layer.
> > > >
> > > >         The iSCSI HBA and its host driver are not at liberty to
> change the
> > > > order of commands from the OS, but the DMAs those commands need 
are
> > > > unlikely to complete in the same order, and as I mentioned some
> > > > commands need no DMA. If the HBA can't send commands out of CmdSN
> > > > order it has to maintain an ordered queue of commands waiting to 
be
> > > > sent, and potentially buffer a lot of data. For an HBA this makes
> > > > immediate data almost impossible to support.
> > > >
> > > >         I don't see the problem with allowing out of order 
commands
> given
> > > > that the target already has to deal with very similar problems. I
> > > > think we are getting in to the area of implementation choices 
here,
> > > > which is inappropriate for a specification.
> > > >
> > > >         - Rod
> > > >