Test Interferometer Networking Plan - Review Report
===============================================================================
ALMA SOFTWARE DOCUMENTATION REVIEW PROCEDURE
REVIEW REPORT
===============================================================================
Date : 2002-January-08
Document : Test Interferometer Networking Plan, ALMA08006Nx0001
Author(s): Mick Brooks
Reviewers submitting comments:
B. Glendenning (bglendenning)
J. Mangum (jmangum)
R. Marson (rmarson)
B. Gustafsson (bgustafs)
===============================================================================
FINAL DECISION
The document has been accepted, pending the changes agreed to below. The
revised document will be posted before 2002-January-11. Comments received during
the review period are included below. Further comments were received after the
review period expired and these are included at the end of this document. The only
major clarifications added after the review period are regarding the number of
subnets, and the requirement to support simultaneous optical pointing. The issue of
splitting the cost of the network was raised in this document but will be finalized
at a later date. Purchase of networking equipment will proceed.
===============================================================================
COMMENTS RECEIVED DURING THE REVIEW PERIOD
--------------------------------------------------------------------------------
(jmangum)
General
I have gotten only comments from you and Brian on the cabling
plan. I suspect that I should just finalize it and distribute it,
but I am unsure as to how it should be distributed. One idea is
that it be an attachment to your network plan.
REPLY:
I don't think that the ALMA software team is the right review group to look at
the cabling plan since the plan includes LO distribution fibers for example.
I will bring this up with the Systems Group.
--------------------------------------------------------------------------------
(rmarson)
General
Where's the data go?
I'm left wondering how the astronomical signals sampled at each
antenna get back to the correlator. Is this on a separate path? If
so that should be stated in the introduction. Similarly the data
path for users looking at the correlated output on workstations in
the SLOB should be mentioned (maybe thats what connection to the CCC is?)
REPLY:
The data path between the antennas and the correlator is outside the scope of
this document since that data is carried on a dedicated fiber system being
developed by Dick Sramek's group. I will mention this in the introduction.
Users may look at any data, including monitor data and correlator output, from
any workstation on the network. I will mention this in Section 4.
--------------------------------------------------------------------------------
(rmarson)
General
I printed this on a monochrome laser printer. The distinction
between red an blue is entirely lost on me. Perhaps you can make
some of the lines dashed.
REPLY:
The high priority stations now have dotted lines to clarify this.
--------------------------------------------------------------------------------
(rmarson)
General
I am not an expert on acronyms. Perhaps you could spell them out
either when first used or in an appendix or make a reference to
the ALMA glossary. Many of them I can figure out but I'm still
pondering some (like ARTM).
REPLY:
I will include a reference to the ALMA Software Glossary.
--------------------------------------------------------------------------------
(jmangum)
p.4 The official name for the Test Interferometer has recently been
changed to ALMA Test Facility (ATF). You might want to mention
this in the introduction. I don't think that it is necessary to
change all reference to the old TI name, though.
REPLY:
I will mention this in the introduction.
--------------------------------------------------------------------------------
(jmangum)
p. 5 In figure 1, you show two trailers. In fact, there will
be one trailer and two contractor containers. I think that we can
hub the ethernet off of the trailer to feed ethernet to the
contractor containers.
REPLY:
I have made this change to the diagrams.
--------------------------------------------------------------------------------
(jmangum)
p.5 It would be better to call the subnet "atf.alma.nrao.edu" to
reflect the recent name change.
REPLY:
OK.
--------------------------------------------------------------------------------
(bglenden)
p.5 Are there only 2 meteorology towers now? At one time I thought there
were more.
REPLY:
I believe that there are only two met towers. Since there was no correction from
Jeff, I'll leave it at two.
--------------------------------------------------------------------------------
(bglenden)
p.5 Currently there is no Gb path from a central computer to an antenna,
even for testing purposes. Can we hang one off the 3508? If so I would show
a test computer just to show that it's possible. If it's not possible I
think we should describe how it would be possible to do this. (I suppose we
don't have any Gb computers at the antenna end either. I see that you list
this as an issue on p.8. I think we should ultimately be able to go
end-to-end to at least one Linux computer and one VxWorks computer if the
costs aren't prohibitive. Not a priority for initial installation.).
REPLY:
I have added a note about this to the issues on p.8. For maximum benefit, the
real time machines should be upgraded to use a GB link as well. In both cases
hardware selection and device driver testing would need to be performed. Unless,
a genuine bottleneck is found, this may not be particularly necessary.
--------------------------------------------------------------------------------
(bglenden)
p.5 GB fiber or Gb fiber?
REPLY:
IEEE standard is "b" for bits. I have changed the document and the figures to
reflect this.
--------------------------------------------------------------------------------
(bglenden)
p.5 Do we now not use the "control" room at all?
REPLY:
I assume that machines such as the ACC and the DNS/DHCP server will be in the
control room since they are not rack-mounted. However, people actually
controlling the antenna will most likely be in the trailer rather than the
control room.
--------------------------------------------------------------------------------
(bglenden)
p.5 I have a (single) network printer budgeted. Should it show up in the
diagram?
REPLY:
As the printer can be located anywhere, and the diagrams are already quite
crowded, I will omit this from the diagrams.
--------------------------------------------------------------------------------
(bglenden)
p.5 Are the 4 workstations and OTC doubled (one per trailer?). I only had
this total for the TI. (And assumed Europe might pay for half of them.
REPLY:
There is only one trailer, so they are not doubled. Even when there are two
antennas and two optical telescopes it is unlikely we will want to do
simultaneous optical pointing, so your total seems correct.
--------------------------------------------------------------------------------
(bglenden)
p.5 Perhaps it now has to be atf.alma.nrao.edu.
REPLY:
Yes, fixed.
--------------------------------------------------------------------------------
(bustafs)
p.5 Figure 1: What are these Trailers? No explanation anywhere.
Where in the diagram are the long haul fibers? I guess between the 3508
switch and the 3512 switches. Are you planning to test also with a very long
fiber (20-30 km)?
REPLY:
The trailer is the main control location as planned by the Antenna Evaluation
Group (AEG). I will add an explanation.
The long haul fibers are indeed between the 3508 and the 3512 switches at the
antennas. We will initially test with 5 km GBICs and later test with long-haul
100 km GBICs. I have added explanation of this to the text.
--------------------------------------------------------------------------------
(bustafs)
p.6 section 4 last sentence. Do you really want to give highest priority
to VOIP?
I understand that it might not work otherwise, but it should not effect
the M&C traffic!
REPLY:
I agree with you, but our network specialist, Gene Runion does not agree with
us. I have added clarification of this sentence to suggest that thorough testing
will be required to determine the best priority for VoIP traffic relative to
M&C traffic.
--------------------------------------------------------------------------------
(jmangum)
p.7 figure 2: I believe that the red-coloured computers
located at each antenna are the ABMs. Might want to label them as
such.
REPLY:
Yes, I have clarified this and the blue laptop at the antenna as well.
--------------------------------------------------------------------------------
(jmangum)
p.8 Regarding the issue of an additional T1 for the VLA link, I think
that we definitely need it. My experience with the webcam and the
existing T1 indicates that the current link could not possibly
handle the increased traffic that the ATF will require.
REPLY:
OK, this decision is definite.
--------------------------------------------------------------------------------
(jmangum)
p.8 In section 5 you raise several issues. Let me just comment on
them:
-- Initial support for three antennas: Probably best to buy the
equipment in phases. Outfit a single antenna in 2002.
-- Your proposal for payment of the equipment by the partners
sounds good to me.
-- As you suggest, computer networking and telephony costs should
be borne by Computing. The Systems group has been working
under the assumption that meteorology, holography, trailers,
etc. would be purchased from their own budgets.
-- Delay the purchase of the CallManager until we can figure out
how to interface to the VLA PBX, which is, as you point out,
from a different vendor (NEC).
REPLY:
Accepted. Document reflects these decisions.
--------------------------------------------------------------------------------
(bustafs)
p.9 What is the purpose of the media conveters?
REPLY:
Where 100 Mb Ethernet ports need to run over longer distances and outside the
RFI shielded buildings (such as the links to the trailer and the SLOB), then
fiber should be used for RFI reasons. The media converters take 100 Mb twisted
pair and convert it onto fiber.
--------------------------------------------------------------------------------
(jmangum)
p.10 Appendix A: You have 3 telephones allocated to the control room.
Now that the control room will be located in the trailer, perhaps
it would be better to shift 2 of these control room phones to the
trailer?
REPLY:
OK.
--------------------------------------------------------------------------------
(bustafs)
p.12 From the table I understand that you alternatively use the WS-G5486
units or the WS-G5487.
How can it be that for the first one you need three for the antennas
(one per each) and for the 5487 you need only one?
Did I misunderstand something here?
REPLY:
This is directly from a quote from Cisco, where I asked them for prices for
outfitting one antenna with the 5487 and one with the 5486. Given this pricing
we will outfit the Vertex antenna with a 5486. In short, the tables in Appendix
should be viewed for pricing information only and not for final equipment parts
numbers. The parts numbers are in Appendix A.
===============================================================================
COMMENTS RECEIVED AFTER THE REVIEW PERIOD
--------------------------------------------------------------------------------
(jmangum)
Mick> (bglenden)
Mick> p.5 Are the 4 workstations and OTC doubled (one per trailer?). I only had
Mick> this total for the TI. (And assumed Europe might pay for half of them.
Mick> REPLY:
Mick> There is only one trailer, so they are not doubled. Even when there are two
Mick> antennas and two optical telescopes it is unlikely we will want to do
Mick> simultaneous optical pointing, so your total seems correct.
Actually, we will likely want to do simultaneous pointing, especially
optical pointing, with the two (three?) prototype antennas.
REPLY:
I will amend the document to mention this point.
--------------------------------------------------------------------------------
(graffi)
p.5 Fig.1 SLOB and trailer
I am concerned that when antennas get tested in parallel, all the
2 (or 3) teams can fit into the trailer. What is the role of SLOB then?
I assume you need 4 workstations for one team (or antenna sub-team).
This situation is confirmed by Jeff in his comment as well:
Actually, we will likely want to do simultaneous pointing, especially
optical pointing, with the two (three?) prototype antennas.
But even if there was only other technical work to do, we need
to be able to do technical tests, check optical telescopes etc in parallel.
Given the space problem in a standard trailer,
would it be possible to assume that only technical relatively low level work
gets done in the trailer and higher level tests in SLOB.
Looking at your text, comments from Brian and Jeff, I still do not have a
clear picture of how we will work. I suspect you give for granted something
I do not know (remember) here.
REPLY: (from jmangum)
Let me try to explain the planned office/work space layout. There
will be a 60 foot x 14 foot (18.3m x 4.3m) trailer located about 100
ft (30.5m) to the north of the east-west prototype antenna baseline.
This trailer will serve as the main control room. The other office
which will serve as a secondary control room is located on the second
floor of the VLA control building, located about 100m to the
south-west of the prototype antennas. This room is about 6m x 6m or
so in size, but does not have a clear line-of-sight to the prototype
antennas. The SLOB is a detached building located to the west of the
VLA control building which used to serve as the offices for the
VLA scientific staff before the AOC was built. A number of offices
have been set aside for general ATF staff use in this building. These
will likely just be offices where people can go to work, away from the
activity of the ATF.
The following link also include useful information about the ATF site facilities:
http://www.tuc.nrao.edu/~jmangum/alma/atf/
--------------------------------------------------------------------------------
(graffi)
p.5 3.1 Subnets ...48.0 to 51.0 This subnet should be named atf.
(from Erik) You have potentially 4*256 hosts and it looks a lot for putting them
on 1 subnet, if this is what you mean. You could specify masking in such a way
to have different subnets for different kind of traffic. Different subnets
provide an easy way to group different kinds of traffic.This could be used for
example to isolate VOIP rather than relaying on Quality of Service and
Priorities.
Or one might decide to have a red subnetwork (m&c traffic) separate from
the generic subnetwork.
REPLY:
I will amend the document to include the possibility of more than one subnet being
used with the available IP address range.
--------------------------------------------------------------------------------
(graffi)
p.8 5 Issues
Outfit a single antenna in 2002.
Yes and ESO will pay equipment for the other antenna early 2003 (right?).
In case we had to pay our 50% of shared equipment this year, I would like
to have the approximate amount by Brian sometimes.
REPLY:
OK.
--------------------------------------------------------------------------------
(bglenden)
> p.8 5 Issues
> Outfit a single antenna in 2002.
> Yes and ESO will pay equipment for the other antenna early 2003 (right?).
> In case we had to pay our 50% of shared equipment this year, I would like
> to have the approximate amount by Brian sometimes.
I think it's clear that each partner will pay for the equipment in their
antenna. What's less clear to me is who pays for the common equipment. Does
NRAO pay for all of it and Europe can pay for more of something else later
(my preference - keeps things simpler), or do we split the costs up front?
If we split the costs, what counts as common equipment? An ALMA specific
printer? Increased network costs?
I suggest that we separate these "who pays for what" issues from the
technical discussion in the paper. I've prodded my management to think about
the issue. I'll separately send Gianni my current FY02 budget. The common
"stuff" could range up to $100k depending on what one wants to include.
REPLY:
OK.
--------------------------------------------------------------------------------
(kmorita)
p.8
"Each ALMA partner should pay 1/3 of total networking costs."
I basically agree with this idea. ALMA-J is willing to pay reasonable
cost for the test interferometer experiment. On the other hand, ALG
(ALMA Liaison Group) are discussing about Japanese contribution to
total costs of TI under current 2-way ALMA structure.
I think, our contribution to total networking costs should be
consistent with this discussion.
REPLY:
OK.
_____oOo_____
