Thanks Adrian for the detailed review and suggestions. Please see replies
inline with ...
On 2019-04-17, 11:03 PM, "Adrian Farrel" wrote:
Hello authors,
Sorry about the clunk as this draft shifted between shepherd.
I have some fairly minor review comments (below). Could you please
address
these in a new revision.
Meanwhile, I will start on the shepherd write-up ready to move ahead as
quickly as possible.
Thanks,
Adrian
---
Odd leading space on 4th line of Abstract
Fixed.
---
Abstract para 2
s/Automatic bandwidth/Automatic bandwidth adjustment/ ??
Added – “The automatic bandwidth feature”
---
2.3
Down-Adjustment-Interval
s/lesser/less/
Fixed.
---
I found the definition of Maximum Average Bandwidth and the definitions
of Up/Down-Adjustment-Interval to be circular. Unless, perhaps, the
definition of Adjustment-Interval is missing.
That is, Maximum Average Bandwidth is defined as
max {Bandwidth-Sample(i)} for each sample interval i in the
Adjustment-Internal
But Up/Down-Adjustment-Interval both appear to be dependent on Maximum
Average Bandwidth.
Removed the Adjustment-Interval from the Max Average Bandwidth
definition. That should remove the circular dependency.
---
3.
s/the PCC, the LSP that/the PCC, which LSPs/
Fixed.
---
4.1
OLD
Auto-Bandwidth feature allows automatic and dynamic adjustment of the
reserved bandwidth of an LSP over time, i.e. without network operator
intervention to accommodate the varying traffic demand of the LSP.
NEW
The Auto-Bandwidth feature allows automatic and dynamic adjustment of
the reserved bandwidth of an LSP over time (i.e., without network
operator intervention) to accommodate the varying traffic demand of
the LSP.
END
Fixed.
---
4.1 has...
The bandwidth
adjustment uses the make-before-break (MBB) signaling method so that
there is no disruption to the traffic flow carried by the LSP.
I think this should be...
Bandwidth adjustment must not cause disruption to the traffic flow
carried by the LSP. One way to achieve this is to use the make-
before-break (MBB) signaling method.
This is the softest way I can think of saying what you don't want to
say which is that RSVP-TE signaling supports in-place bandwidth
adjustmnt simply by sending a new Path message. (Noting that failure to
make the adjustment can be seen either in a non-fatal PathErr or in a
Resv with unchanged bandwidth.)
Fixed.
Similarly in 4.3 maybe...
OLD
It should be noted that any bandwidth change requires re-signaling of
an LSP in a make-before-break fashion, which can further trigger
preemption of lower priority LSPs in the network.
NEW
It should be noted that any bandwidth change requires re-signaling of
an LSP, which can further trigger preemption of lower priority LSPs
in the network.
END
Fixed.
---
4.2 has...
When the Auto-Bandwidth feature is enabled, the measured traffic rate
is periodically sampled at each Sample-Interval (which can be
configured by an operator and the default value as 5 minutes) by the
PCC which is the head-end node of the LSP.
As you know, in the PCE architecture, the PCC is not necessarily the
head-end LSR. You need either to restrict this function to only
operating when the PCC *is* the head-end LSR, or you need to re-word
slightly. Either works for me.
Updated. “by the PCC, when the PCC is the head-end node of the LSP.”
---
I wonder whether you intend that Section 4 (in particular Section 4.2)
is normative in this document. It could be that you are just describing
the procedures in a general way - in which case a one line statement of
that would address my concern. Or it could be that you intend to define
how auto-bandwidth adjustment must be implemented.
That is, the document claims to describe changes to PCEP to support
auto-bandwidth, but this section appears to be describing how auto-
bandwidth must be implemented, and I don't think I agree that the
description is completely wise.
Added – “This section describes the Auto-Bandwidth feature in a
general way.”
For example...
The
PCC, in-charge of calculating the bandwidth to be adjusted, will
adjust the bandwidth of the LSP to the highest traffic rate sample
(MaxAvgBw) amongst the set of bandwidth samples collected over the
adjustment-interval period (in the Up or Down direction).
...means that a single spike in a potentially very small window over
a potentially v