Brian,
>I tried to send this to tt-tf(a)ripe.net but it was bounced.
Forwarded it.
> > Hardware:
> > Outdated hardware, hardware without a current contact person, etc, is
> > an ongoing concern for our operations people. We suggest the following
> > changes:
> > * We start writing off the hardware, say 3 or 4 years (perhaps
> > longer for the clocks). After this period, a box will be automatically
> > switched off unless a site buys new hardware.
>
>When you say 'new' hardware, are you talking about a new model of the
>same box (which may be difficult when it comes to Dell 350s?) or is
>the server base for the TTM box being upgraded? And if so, to what?
Replace by something that is available then and is suitable to run
the TTM software. At the moment, that would be a Dell 850, but I'm
certain that in a couple of years, it will be something different.
> > The host has to pay the entire sum at once, in advance. After that,
> > there will be no "hidden fees".
>
>Ah yes, most useful. Is there any fear that a higher price will put
>people off?
I thought of that and this is a possibility. OTOH, it now takes a lot
of time/money from our side to collect the service fees and some sites
get the service for "free" (by not paying). I don't think that is good
either. I personally prefer to have a smaller number of sites that all
pay, over a large number of sites where only a fraction pays.
> > * Sites will be asked to provide current contact information and to
> > regularly check this.
>
>Is the intention for the site to check this or for the RIPE NCC to
>prompt the site to check this?
I think prompting is the best way.
Henk
------------------------------------------------------------------------------
Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre http://www.amsterdamned.org/~henk
P.O.Box 10096 Singel 258 Phone: +31.20.5354414
1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445
The Netherlands The Netherlands Mobile: +31.6.55861746
------------------------------------------------------------------------------
1160438400 + 381600 = 1160820000.
>X-Original-To: henk(a)ripe.net
>Delivered-To: henk(a)ripe.net
>Date: Tue, 26 Sep 2006 10:07:18 +0100
>From: Brian Nisbet <brian.nisbet(a)heanet.ie>
>User-Agent: Mozilla Thunderbird 1.0 (Windows/20041206)
>X-Accept-Language: en-us, en
>To: Henk Uijterwaal <henk(a)ripe.net>
>Subject: Re: TTM Service 2007 - 2009
>X-Enigmail-Version: 0.92.0.0
>X-RIPE-Spam-Level:
>X-RIPE-Spam-Tests: BAYES_00
>X-RIPE-Spam-Status: U 0.285436 / -2.6
>X-RIPE-Signature: 20a856a8701b575655ec0799ed4b09ee
>
>Henk,
>
>I tried to send this to tt-tf(a)ripe.net but it was bounced.
>
> > Please see below. For discussion,
>
>I don't have a huge amount to add to this as, in general, I
>agree with it all, but there are a couple of points I'd like
>to note.
>
> > Draft proposal for the TTM service 2007-2009.
> >
> > Services on the boxes:
> > Starting with the current set of boxes and services:
> > * Delay/Loss/Jitter/Traceroute:
> > At RIPE52, there was clear interest in the data. We propose to
> > continue
> > these measurements. The software on the TBs exists as well as the code
> > to analyze it. We do want to make a couple of adjustments
> > * OWAMP will be added as a method to do measurements on demand,
> > either between TBs or outside the TTM network (subject to access
> > restrictions).
>
>That makes perfect sense and should widen the appeal.
>
> > * Data will be collected at a central point and initial processing
> > will still be done. However, we will reduce the number of plots and
> > lists produced by default on a daily basis to something like: daily
> > plots, daily overviews, weekly overviews. All other plots will only be
> > generated on demand. This will result in a bit of lag when one asks
> > for a plot. On the other hand, producing a large number of plots that
> > nobody ever looks at is not a good idea either.
>
>As one of those who pointed out we were still using the data that should
>satisfy our needs. We very rarely use the box as an up to the second
>resourse, so some delay for a more detailed plot is fine. I would
>guess, however I wouldn't want to presume, that other operators act
>in similiar ways.
>
> > * Network alarms will be revisited, is anybody ever looking at
> > them?
>
>I know I haven't looked in quite a while, but as Mike Norris used to
>use them a lot I will check with him, however I suspect you are right
>that their readership is extremely low. Is the overhead of generation
>all that high?
>
> > * Data will continue to be visible on the TB itself.
> > * We will set up a DQM mechanism, with a daily routine to look at
> > the data and see if the plots make sense (rather than what we do
> > today: are they being filled).
>
>Both great.
>
> > * DNSMON
> > * DNSMON is a valuable service and we will keep it. Features to
> > be added could include:
> > * IPv6
> > * Trends/Alarms
>
>Excellent.
>
> > See below.
> > * New: Protocol beacons
> > * The boxes will start to run protocol beacons, i.e. something
> > that sends out packets for a service on a regular basis, allowing
> > people to monitor their systems.
> > * An example is the multicast work. Another example are the
> > debogon prefixes on the RIS.
>
>Ah, this would be excellent and a welcome service, primarily because
>it would allow us to maintain one fewer server ourselves and we're
>always happy if we can manage that.
>
> > * New: Network experiments.
> > * We will set up shell accounts that will allow interested
> > parties to do one-off network experiments. (Proposal, measurement,
> > analysis, publication).
>
>Excellent, assuming the necessary strict controls etc.
>
> > * New: CBM reference point/server: the boxes will be set up such
> > that interested ISPs can use them for this purpose. The box will offer
> > the configuration site, plus a target for the measurements. The NCC
> > will look into aggregation of data, from reports for one user, to a
> > report for one ISP.
>
>While being broadly in favour of this I still feel we need to be
>extremely careful about how and to whom this data will be accessible.
>I think it could be an extremely useful resource for all concered, but
>also a potential way of making some ISPs (unfairly) angry at RIPE.
>
> > Ownership of the boxes:
> > * For both TTM and DNSMON, it is essential that there are a large
> > number of active TTM boxes at interesting locations.
> > * For this reason, the NCC proposes to install a number of boxes at
> > important locations for free. The NCC pays for the hardware and
> > service contract, the local site only for
> > power/connectivity/operations (making it essentially free). Rough
> > criteria for hosting a box:
>
>This is a fantastic idea and one that I am very glad to see.
>
>Roughly how many boxes would be expected to be deployed under this
>scheme?
>
> > Hardware:
> > Outdated hardware, hardware without a current contact person, etc, is
> > an ongoing concern for our operations people. We suggest the following
> > changes:
> > * We start writing off the hardware, say 3 or 4 years (perhaps
> > longer for the clocks). After this period, a box will be automatically
> > switched off unless a site buys new hardware.
>
>When you say 'new' hardware, are you talking about a new model of the
>same box (which may be difficult when it comes to Dell 350s?) or is
>the server base for the TTM box being upgraded? And if so, to what?
>
> > * We will start to support GPS-free boxes (TSC clock) once that
> > technology is mature. A nearby Stratum-1 NTP server may be required.
>
>Considering the problems I'm currently having getting the antenna for
>TT-35 to play nice, this is encouraging. I strongly suspect HEAnet
>will continue to want a Stratum-1 clock on our box, but anything that
>can be done to make installation easier for the majority is a good
>thing.
>
> > Administrative:
> > Our finance department faces a lot of problems collecting the service
> > fee. It is often not clear to people that the SF is an integral part
> > of the service. Also collecting money from hosts where the contact
> > people have disappeared. We suggest:
> > * The current service contracts will be revisited and replaced
> > (taking into account the text of the current contract) by a new
> > contract.
> > * In the new system, there will be one contract to cover:
> > * Hardware
> > * Service fee for a 3 year period
> > * Shipping, import/export duties,
> > The host has to pay the entire sum at once, in advance. After that,
> > there will be no hidden fees.
>
>Ah yes, most useful. Is there any fear that a higher price will put
>people off?
>
> > * Sites will be asked to provide current contact information and to
> > regularly check this.
>
>Is the intention for the site to check this or for the RIPE NCC to
>prompt the site to check this?
>
> > Lifetime of this proposal:
> > This is a 3-year plan, 2007-2009, to be evaluated afterwards.
>
>Overally I think this is an excellent plan and, on what is given
>here, I believe it will maintain those parts of the service that
>Gert and I were so forthright about in Istanbul while providing
>a model to stabilise and then expand the network.
>
>Thanks,
>
>Brian.
>
>--
>Brian Nisbet,
>Senior Network Engineer, HEAnet
>Email: brian.nisbet(a)heanet.ie
>Tel: +35316609040/Fax:+35316603666
------------------------------------------------------------------------------
Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre http://www.amsterdamned.org/~henk
P.O.Box 10096 Singel 258 Phone: +31.20.5354414
1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445
The Netherlands The Netherlands Mobile: +31.6.55861746
------------------------------------------------------------------------------
1160438400 + 381600 = 1160820000.
Hi all,
Please see below. For discussion,
Henk
- - - -
Draft proposal for the TTM service 2007-2009.
Services on the boxes:
Starting with the current set of boxes and services:
* Delay/Loss/Jitter/Traceroute:
At RIPE52, there was clear interest in the data.
We propose to continue these measurements. The
software on the TBs exists as well as the code to
analyze it. We do want to make a couple of adjustments
* OWAMP will be added as a method to do
measurements on demand, either between TBs or
outside the TTM network (subject to access restrictions).
* Data will be collected at a central point
and initial processing will still be done.
However, we will reduce the number of plots and
lists produced by default on a daily basis to
something like: daily plots, daily overviews,
weekly overviews. All other plots will only be
generated on demand. This will result in a bit of
lag when one asks for a plot. On the other hand,
producing a large number of plots that nobody
ever looks at is not a good idea either.
* Network alarms will be revisited, is anybody ever looking at them?
* Data will continue to be visible on the TB itself.
* We will set up a DQM mechanism, with a
daily routine to look at the data and see if the
plots make sense (rather than what we do today: are they being filled).
* DNSMON
* DNSMON is a valuable service and we
will keep it. Features to be added could include:
* IPv6
* Trends/Alarms
See below.
* New: Protocol beacons
* The boxes will start to run protocol
beacons, i.e. something that sends out packets
for a service on a regular basis, allowing people to monitor their systems.
* An example is the multicast work.
Another example are the debogon prefixes on the RIS.
* New: Network experiments.
* We will set up shell accounts that
will allow interested parties to do one-off
network experiments. (Proposal, measurement, analysis, publication).
* The TB will just offer the framework
for the experiment and a little operator support,
whoever proposes the experiment will be responsible for doing the work.
* Similar models have been proposed
(VINI, PlanetLab, ), we will see what we can
copy and where we can collaborate.
* This is intended as a community
services for R&D that requires access to real hosts in the real Internet.
* Example from the past: Kroot anycast measurements.
* New: CBM reference point/server: the boxes
will be set up such that interested ISPs can use
them for this purpose. The box will offer the
configuration site, plus a target for the
measurements. The NCC will look into aggregation
of data, from reports for one user, to a report for one ISP.
Ownership of the boxes:
* For both TTM and DNSMON, it is essential
that there are a large number of active TTM boxes at interesting locations.
* For this reason, the NCC proposes to
install a number of boxes at important locations
for free. The NCC pays for the hardware and
service contract, the local site only for
power/connectivity/operations (making it
essentially free). Rough criteria for hosting a box:
* Major AS
* Only one box per AS, if a site wants more, they have to pay
* Commitment from the site to operate the box for a 3 year period.
Selection will be a beauty contest.
* Other sites can still order boxes as
before, the same goes for AS who want to install more than one box.
Hardware:
Outdated hardware, hardware without a current
contact person, etc, is an ongoing concern for
our operations people. We suggest the following changes:
* We start writing off the hardware, say 3 or
4 years (perhaps longer for the clocks). After
this period, a box will be automatically switched
off unless a site buys new hardware.
* Current hardware older than 3 years will be phased-out and/or replaced.
* We will allow people to build their own
box, subject to the restriction that our software
will run on it without modifications. This means
that FreeBSD should run on the box without modifications.
* We will start to support GPS-free boxes
(TSC clock) once that technology is mature. A
nearby Stratum-1 NTP server may be required.
Administrative:
Our finance department faces a lot of problems
collecting the service fee. It is often not clear
to people that the SF is an integral part of the
service. Also collecting money from hosts where
the contact people have disappeared. We suggest:
* The current service contracts will be
revisited and replaced (taking into account the
text of the current contract) by a new contract.
* In the new system, there will be one contract to cover:
* Hardware
* Service fee for a 3 year period
* Shipping, import/export duties,
The host has to pay the entire sum at once, in
advance. After that, there will be no hidden fees.
* HW support for the boxes will cover the
lifetime of the boxes. In other words, one knows
in advance that the service will be there for 3 years.
* The service for sites who have not paid for
several years and who we dont consider interesting will be stopped.
* Sites will be asked to provide current
contact information and to regularly check this.
Lifetime of this proposal:
This is a 3-year plan, 2007-2009, to be evaluated afterwards.
Operational (for the NCC):
* The number of operational problems will be
reduced, as the H/W will be newer as well as more uniform.
* Contacts for the boxes should be well known again.
------------------------------------------------------------------------------
Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre http://www.amsterdamned.org/~henk
P.O.Box 10096 Singel 258 Phone: +31.20.5354414
1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445
The Netherlands The Netherlands Mobile: +31.6.55861746
------------------------------------------------------------------------------
1160438400 + 381600 = 1160820000.
------------------------------------------------------------------------------
Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre http://www.amsterdamned.org/~henk
P.O.Box 10096 Singel 258 Phone: +31.20.5354414
1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445
The Netherlands The Netherlands Mobile: +31.6.55861746
------------------------------------------------------------------------------
1160438400 + 381600 = 1160820000.
Dear All,
The third thing I promised for this week...
Some years ago, Darryl Veitch at U.Melbourne suggested to use the TSC
register found in modern Pentium chips as a reference clock reference.
The TSC register is related to the internal clock of the chip and very
stable even when the chip is mounted inside a regular PC. Darryl and
colleagues published a paper on the principle at IMC.
If this works as expected, it will allow for a TTM setup without the need
for a local GPS antenna. This will obviously make deployment a lot
easier. The setup will require a stratum 1 NTP server a few hops away
and there will be some loss of accuracy. All this will have to be
quantified and calibrated.
Over the last years, we have been trying to us this idea in TTM. Mark
Santcroos worked on this while still at the NCC. When he left last year,
the system worked but it had to be calibrated. This was to be done by
the U.Melbourne. For various reasons, they only started on this in June
this year.
We now have a setup where the user can switch between standard GPS-based
timestamping and TSC-based timestamping. We still have to add a switch
and some code to the TTM software to do this for TTM. The idea is to do
measurements with this setup in various circumstances, switching back and
forth between timekeeping models, and compare the results.
First results look good, but a lot of work still has to be done. The
goal is to have this ready by RIPE53.
Henk
------------------------------------------------------------------------------
Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre http://www.amsterdamned.org/~henk
P.O.Box 10096 Singel 258 Phone: +31.20.5354414
1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445
The Netherlands The Netherlands Mobile: +31.6.55861746
------------------------------------------------------------------------------
Dear All,
At RIPE 52, we voluntered to be members of a task force to study the future
of TTM. I planned to start the discussion on this before the summer, but
for a variety of reasons this didn't happen. However, we still have time
before the RIPE meeting and we should be able to come up with something.
In the next week, I'm planning to do the following:
1. Wake y'all up (by sending this mail).
2. Post a short report on our experiences with Luca Deri's prototype
implementation of the CBM proposal.
3. Post a write up on the TSC work (as this might be very relevant for
a GPS-free TB).
4. Post some notes on our work on the (Gert Doering) Multicast
proposal.
5. Post a draft proposal for TTM in the future, as a starting point
for our discussion.
I'd like to invite you all to comment on these docs, come up with
good ideas, throw tomato's in my direction if you disagree, and
participate in the discussion as you see fit.
We do have a little over a month, this should be sufficient time, but
we cannot afford to wait forever to get started,
Henk
------------------------------------------------------------------------------
Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre http://www.amsterdamned.org/~henk
P.O.Box 10096 Singel 258 Phone: +31.20.5354414
1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445
The Netherlands The Netherlands Mobile: +31.6.55861746
------------------------------------------------------------------------------
1160438400 + 381600 = 1160820000.
