Re: HTTP/2 flow control <draft-ietf-httpbis-http2-17>

Patrick McManus <> Tue, 10 March 2015 21:16 UTC

Return-Path: <>
Received: from localhost ( []) by (Postfix) with ESMTP id 802391A8A44 for <>; Tue, 10 Mar 2015 14:16:30 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -6.289
X-Spam-Status: No, score=-6.289 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, FM_FORGED_GMAIL=0.622, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_PASS=-0.001, SPF_PASS=-0.001, T_RP_MATCHES_RCVD=-0.01] autolearn=ham
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id kX0KP03WA60y for <>; Tue, 10 Mar 2015 14:16:21 -0700 (PDT)
Received: from ( []) (using TLSv1.2 with cipher DHE-RSA-AES128-SHA (128/128 bits)) (No client certificate requested) by (Postfix) with ESMTPS id EE8691A8A23 for <>; Tue, 10 Mar 2015 14:16:19 -0700 (PDT)
Received: from lists by with local (Exim 4.80) (envelope-from <>) id 1YVRSb-00044u-VC for; Tue, 10 Mar 2015 21:13:06 +0000
Resent-Date: Tue, 10 Mar 2015 21:13:05 +0000
Resent-Message-Id: <>
Received: from ([]) by with esmtp (Exim 4.80) (envelope-from <>) id 1YVRSS-00041f-GM for; Tue, 10 Mar 2015 21:12:56 +0000
Received: from ([] by with esmtp (Exim 4.72) (envelope-from <>) id 1YVRSP-0005PB-K6 for; Tue, 10 Mar 2015 21:12:56 +0000
Received: from ( []) by (Postfix) with ESMTPSA id 904EE3A015 for <>; Tue, 10 Mar 2015 17:12:31 -0400 (EDT)
Received: by qcvp6 with SMTP id p6so5469979qcv.1 for <>; Tue, 10 Mar 2015 14:12:31 -0700 (PDT)
MIME-Version: 1.0
X-Received: by with SMTP id h35mr43695129qgf.71.1426021951372; Tue, 10 Mar 2015 14:12:31 -0700 (PDT)
Received: by with HTTP; Tue, 10 Mar 2015 14:12:31 -0700 (PDT)
In-Reply-To: <>
References: <> <>
Date: Tue, 10 Mar 2015 17:12:31 -0400
Message-ID: <>
From: Patrick McManus <>
To: Bob Briscoe <>
Cc: HTTP Working Group <>
Content-Type: multipart/alternative; boundary="001a11396244da869b0510f59b5e"
Received-SPF: neutral client-ip=;;
X-W3C-Hub-Spam-Status: No, score=-3.6
X-W3C-Hub-Spam-Report: AWL=-2.454, BAYES_00=-1.9, HTML_MESSAGE=0.001, SPF_NEUTRAL=0.779, URIBL_BLOCKED=0.001
X-W3C-Scan-Sig: 1YVRSP-0005PB-K6 798d77bc45988c8abacff631b02b56fb
Subject: Re: HTTP/2 flow control <draft-ietf-httpbis-http2-17>
Archived-At: <>
X-Mailing-List: <> archive/latest/28923
Precedence: list
List-Id: <>
List-Help: <>
List-Post: <>
List-Unsubscribe: <>

Hi Bob - I think your comments are appreciated. Its just one of those
things where people have dispersed to other things and aren't necessarily
in a place to revisit all the ground work again at this stage for a new go
round. It was in large part the operational feedback and needs of the team, who has a lot of experience operating spdy at scale, that
created the flow control provisions. hopefully those folks will chime in
more authoritatively than my musings below:

I'm sure there is quite a bit to learn here - indeed poorly configured use
of the window_update mechanism has been (predictably) a source of
unintended bottlenecks during both spdy and h2 trials. The spec does try
and highlight that there can be dragons here and implementations that don't
need the features it can bring should provide essentially infinite credits
to steer clear of them.

During the various trials I've seen h2 per stream flow control deployed
successfully for a couple of use cases - both of them essentially deal with
unsolicited data.

The primary one is essentially a more flexible version of h1's
100-continue. When a client presents a large message body (e.g. a file
upload) a multiplexing server needs a way of saying "these are how many
buffers I've got available while I figure out where I'm going to sink this
incoming data (perhaps to another server I need to connect to)". Presenting
this on a per-stream basis allows the server to limit one stream while
another (with a distinct sink) can proceed independently.  IMO this value
should represent resources available and should be independent of BDP. This
is why in practice you see clients with extremely large stream windows -
most circumstances just want the data to flow at line rate (as you
describe) and aren't trying to invoke flow control. The firefox default
window is 256MB per stream - that's not going to slow down the sender nor
require frequent window_update generation.

The other use case is when the server pushes resources at a client without
them being requested, which is a new feature of h2. This is conceptually
similar to the server receiving a POST - suddenly there is a large amount
of inbound data that the implementation might not have the resources to
store completely. We can't just let TCP flow control take care of the
situation because the TCP session is being multiplexed between multiple
streams that need to be serviced. In this case the client accepts "some" of
the stream based on policy and resource availability and can leave the
stream in limbo until an event comes along that tells it to resume the
transfer by issuing credits or reject it via cancel.

hth a least a bit.

On Tue, Mar 10, 2015 at 4:31 PM, Bob Briscoe <> wrote:

>  HTTP/2 folks,
> I know extensibility had already been discussed and put to bed, so the WG
> is entitled to rule out opening healed wounds.
> But have points like those I've made about flow control been raised
> before? Please argue. I may be wrong. Discussion can go on in parallel to
> the RFC publication process, even tho the process doesn't /require/ you to
> talk to me.
> If I'm right, then implementers are being mandated to write complex flow
> control code, when it might have little bearing on the performance benefits
> measured for http/2.
> Even if I'm right, and the WG goes ahead anyway, /I/ will understand. My
> review came in after your deadline.
> However, bear in mind that the Webosphere might not be so forgiving. If h2
> goes ahead when potential problems have been identified, it could get a bad
> reputation simply due to the uncertainty, just when you want more people to
> take it up and try it out. Given you've put in a few person-years of
> effort, I would have thought you would not want to risk a reputation flop.
> I'm trying to help - I just can't go any faster.
> Bob
> At 14:43 06/03/2015, Bob Briscoe wrote:
> HTTP/2 folks,
> As I said, consider this as a late review from a clueful but fresh pair of
> eyes.
> My main concerns with the draft are:
> * extensibility (previous posting)
> * flow control (this posting - apologies for the length - I've tried to
> explain properly)
> * numerous open issues left dangling (see subsequent postings)
> The term 'window' as used throughout is incorrect and highly confusing, in:
> * 'flow control window' (44 occurrences),
> * 'initial window size' (5),
> * or just 'window size' (8)
> The HTTP/2 WINDOW_UPDATE mechanism constrains HTTP/2 to use only
> credit-based flow control, not window-based. At one point, it actually says
> it is credit-based (in flow control principle #2 <
> >
> ), but otherwise it incorrectly uses the term window.
> This is not just an issue of terminology. The more I re-read the flow
> control sections the more I became convinced that this terminology is not
> just /confusing/, rather it's evidence of /confusion/. It raises the
> questions
> * "Is HTTP/2 capable of the flow control it says it's capable of?"
> * "What type of flow-control protocol ought HTTP/2 to be capable of?"
> * "Can the WINDOW_UPDATE frame support the flow-control that HTTP/2 needs?"
> To address these questions, it may help if I separate the two different
> cases HTTP/2 flow control attempts to cover (my own separation, not from
> the draft):
> a) Intermediate buffer control
> Here, a stream's flow enters /and/ leaves a buffer (e.g. at the app-layer
> of an intermediate node).
> b) Flow control by the ultimate client app.
> Here flow never releases memory (at least not during the life of the
> connection). The flow is solely consuming more and more memory (e.g. data
> being rendered into a client app's memory).
> ==a) Intermediate buffer control==
> For this, sliding window-based flow control would be appropriate, because
> the goal is to keep the e2e pipeline full without wasting buffer.
> Let me prove HTTP/2 cannot do window flow control. For window flow
> control, the sender needs to be able to advance both the leading and
> trailing edges of the window. In the draft:
> * WINDOW_UPDATE frames can only advance the leading edge of a 'window'
> (and they are constrained to positive values).
> * To advance the trailing edge, window flow control would need a
> continuous stream of acknowledgements back to the sender (like TCP). The
> draft does not provide ACKs at the app-layer, and the app-layer cannot
> monitor ACKs at the transport layer, so the sending app-layer cannot
> advance the trailing edge of a 'window'.
> So the protocol can only support credit-based flow control. It is
> incapable of supporting window flow control.
> Next, I don't understand how a receiver can set the credit in
> 'WINDOW_UPDATE' to a useful value. If the sender needed the receiver to
> answer the question "How much more can I send than I have seen ACK'd?" that
> would be easy. But because the protocol is restricted to credit, the sender
> needs the receiver to answer the much harder open-ended question, "How much
> more can I send?" So the sender needs the receiver to know how many ACKs
> the sender has seen, but neither of them know that.
> The receiver can try, by taking a guess at the bandwidth-delay product,
> and adjusting the guess up or down, depending on whether its buffer is
> growing or shrinking. But this only works if the unknown bandwidth-delay
> product stays constant.
> However, BDP will usually be highly variable, as other streams come and
> go. So, in the time it takes to get a good estimate of the per-stream BDP,
> it will probably have changed radically, or the stream will most likely
> have finished anyway. This is why TCP bases flow control on a window, not
> credit. By complementing window updates with ACK stream info, a TCP sender
> has sufficient info to control the flow.
> The draft is indeed correct when it says:
> "   this can lead to suboptimal use of available
>    network resources if flow control is enabled without knowledge of the
>    bandwidth-delay product (see [RFC7323]).
> "
> Was this meant to be a veiled criticism of the protocol's own design? A
> credit-based flow control protocol like that in the draft does not provide
> sufficient information for either end to estimate the bandwidth-delay
> product, given it will be varying rapidly.
> ==b) Control by the ultimate client app==
> For this case, I believe neither window nor credit-based flow control is
> appropriate:
> * There is no memory management issue at the client end - even if there's
> a separate HTTP/2 layer of memory between TCP and the app, it would be
> pointless to limit the memory used by HTTP/2, because the data is still
> going to sit in the same user-space memory (or at least about the same
> amount of memory) when HTTP/2 passes it over for rendering.
> * Nonetheless, the receiving client does need to send messages to the
> sender to supplement stream priorities, by notifying when the state of the
> receiving application has changed (e.g. if the user's focus switches from
> one browser tab to another).
> * However, credit-based flow control would be very sluggish for such
> control, because credit cannot be taken back once it has been given (except
> HTTP/2 allows SETTINGS_INITIAL_WINDOW_SIZE to be reduced, but that's a
> drastic measure that hits all streams together).
> ==Flow control problem summary==
> With only a credit signal in the protocol, a receiver is going to have to
> allow generous credit in the WINDOW_UPDATEs so as not to hurt performance.
> But then, the receiver will not be able to quickly close down one stream
> (e.g. when the user's focus changes), because it cannot claw back the
> generous credit it gave, it can only stop giving out more.
> IOW: Between a rock and a hard place,... but don't tell them where the
> rock is.
> ==Towards a solution?==
> I think 'type-a' flow control (for intermediate buffer control) does not
> need to be at stream-granularity. Indeed, I suspect a proxy could control
> its app-layer buffering by controlling the receive window of the incoming
> TCP connection. Has anyone assessed whether this would be sufficient?
> I can understand the need for 'type-b' per-stream flow control (by the
> ultimate client endpoint). Perhaps it would be useful for the receiver to
> emit a new 'PAUSE_HINT' frame on a stream? Or perhaps updating per-stream
> PRIORITY would be sufficient? Either would minimise the response time to a
> half round trip. Whereas credit flow-control will be much more sluggish
> (see 'Flow control problem summary').
> Either approach would correctly propagate e2e. An intermediate node would
> naturally tend to prioritise incoming streams that fed into prioritised
> outgoing streams, so priority updates would tend to propagate from the
> ultimate receiver, through intermediate nodes, up to the ultimate sender.
> ==Flow control coverage==
> The draft exempts all TCP payload bytes from flow control except HTTP/2
> data frames. No rationale is given for this decision. The draft says it's
> important to manage per-stream memory, then it exempts all the frame types
> except data, even tho each byte of a non-data frame consumes no less memory
> than a byte of a data frame.
> What message does this put out? "Flow control is not important for one
> type of bytes with unlimited total size, but flow control is so important
> that it has to be mandatory for the other type of bytes."
> It is certainly critical that WINDOW_UPDATE messages are not covered by
> flow control, otherwise there would be a real risk of deadlock. It might be
> that there are dependencies on other frame types that would lead to a
> dependency loop and deadlock. It would be good to know what the rationale
> behind these rules was.
> ==Theory?==
> I am concerned that HTTP/2 flow control may have entered new theoretical
> territory, without suitable proof of safety. The only reassurance we have
> is one implementation of a flow control algorithm (SPDY), and the anecdotal
> non-evidence that no-one using SPDY has noticed a deadlock yet (however, is
> anyone monitoring for deadlocks?).
> Whereas SPDY has been an existence proof that an approach like http/2
> 'works', so far all the flow control algos have been pretty much identical
> (I think that's true?). I am concerned that the draft takes the InterWeb
> into uncharted waters, because it allows unconstrained diversity in flow
> control algos, which is an untested degree of freedom.
> The only constraints the draft sets are:
> * per-stream flow control is mandatory
> * the only protocol message for flow control algos to use is the
> WINDOW_UPDATE credit message, which cannot be negative
> * no constraints on flow control algorithms.
> * and all this must work within the outer flow control constraints of TCP.
> Some algos might use priority messages to make flow control assumptions.
> While other algos might associate PRI and WINDOW_UPDATE with different
> meanings. What confidence do we have that everyone's optimisation
> algorithms will interoperate? Do we know there will not be certain types of
> application where deadlock is likely?
> "   When using flow
>    control, the receiver MUST read from the TCP receive buffer in a
>    timely fashion.  Failure to do so could lead to a deadlock when
>    critical frames, such as WINDOW_UPDATE, are not read and acted upon.
> "
> I've been convinced (offlist) that deadlock will not occur as long as the
> app consumes data 'greedily' from TCP. That has since been articulated in
> the above normative text. But how sure can we be that every implementer's
> different interpretations of 'timely' will still prevent deadlock?
> Until a good autotuning algorithm for TCP receive window management was
> developed, good window management code was nearly non-existent. Managing
> hundreds of interdependent stream buffers is a much harder problem. But
> implementers are being allowed to just 'Go forth and innovate'. This might
> work if everyone copies available open source algo(s). But they might not,
> and they don't have to.
> This all seems like 'flying by the seat of the pants'.
> ==Mandatory Flow Control? ==
> "      3. [...] A sender
>        MUST respect flow control limits imposed by a receiver."
> This ought to be a 'SHOULD' because it is contradicted later - if settings
> change.
> "   6.  Flow control cannot be disabled."
> Also effectively contradicted half a page later:
> "   Deployments that do not require this capability can advertise a flow
>    control window of the maximum size (2^31-1), and by maintaining this
>    window by sending a WINDOW_UPDATE frame when any data is received.
>    This effectively disables flow control for that receiver."
> And contradicted in the definition of half closed (remote):
> "  half closed (remote):
>       [...] an endpoint is no longer
>       obligated to maintain a receiver flow control window.
> "
> And contradicted in 8.3. The CONNECT Method
> <>,
> which says:
> "  Frame types other than DATA
>    or stream management frames (RST_STREAM, WINDOW_UPDATE, and PRIORITY)
>    MUST NOT be sent on a connected stream, and MUST be treated as a
>    stream error (Section 5.4.2) if received.
> "
> Why is flow control so important that it's mandatory, but so unimportant
> that you MUST NOT do it when using TLS e2e?
> Going back to the earlier quote about using the max window size, it seems
> perverse for the spec to require endpoints to go through the motions of
> flow control, even if they arrange for it to affect nothing, but to still
> require implementation complexity and bandwidth waste with a load of
> redundant WINDOW_UPDATE frames.
> HTTP is used on a wide range of devices, down to the very small and
> challenged. HTTP/2 might be desirable in such cases, because of the
> improved efficiency (e.g. header compression), but in many cases the stream
> model may not be complex enough to need stream flow control.
> So why not make flow control optional on the receiving side, but mandatory
> to implement on the sending side? Then an implementation could have no
> machinery for tuning window sizes, but it would respond correctly to those
> set by the other end, which requires much simpler code.
> If a receiving implemention chose not to do stream flow control, it could
> still control flow at the connection (stream 0) level, or at least at the
> TCP level.
> ==Inefficiency?==
>  5.2. Flow Control
> <>
> "Flow control is used for both individual
>    streams and for the connection as a whole."
> Does this means that every WINDOW_UPDATE on a stream has to be accompanied
> by another WINDOW_UPDATE frame on stream zero? If so, this seems like 100%
> message redundancy. Surely I must  have misunderstood.
> ==Flow Control Requirements===
> I'm not convinced that clear understanding of flow control requirements
> has driven flow control design decisions.
> The draft states various needs for flow-control without giving me a feel
> of confidence that it has separated out the different cases, and chosen a
> protocol suitable for each. I tried to go back to the early draft on flow
> control requirements <
> >, and I was not impressed.
> I have quoted below the various sentences in the draft that state what
> flow control is believed to be for. Below that, I have attempted to
> crystalize out the different concepts, each of which I have tagged within
> the quotes.
> * 2. HTTP/2 Protocol Overview
> <> says
>   "Flow control and prioritization ensure that it is possible to
> efficiently use multiplexed streams. [Y]
>    Flow control (Section 5.2) helps to ensure that only data that can be
> used by a receiver is transmitted. [X]"
> * 5.2. Flow Control
> <>
> says:
>   "Using streams for multiplexing introduces contention over use of the
> TCP connection [X], resulting in blocked streams [Z]. A flow control scheme
> ensures that streams on the same connection do not destructively interfere
> with each other [Z]."
> * 5.2.2. Appropriate Use of Flow Control
> <>
> "  Flow control is defined to protect endpoints that are operating under
>    resource constraints.  For example, a proxy needs to share memory
>    between many connections, and also might have a slow upstream
>    connection and a fast downstream one [Y].  Flow control addresses cases
>    where the receiver is unable to process data on one stream, yet wants
>    to continue to process other streams in the same connection [X]."
> "  Deployments with constrained resources (for example, memory) can
>    employ flow control to limit the amount of memory a peer can consume.
> [Y]
> Each requirement has been tagged as follows:
> [X] Notification of the receiver's changing utility for each stream
> [Y] Prioritisation of streams due to contention over the streaming
> capacity available to the whole connection.
> [Z] Ensuring one stream is not blocked by another.
> [Z] might be a variant of [Y], but [Z] sounds more binary, whereas [Y]
> sounds more like optimisation across a continuous spectrum.
> Regards
> Bob
> ________________________________________________________________
> Bob Briscoe,                                                  BT
>  ________________________________________________________________
> Bob Briscoe,                                                  BT