Code of good conduct - suggestions to the users of the OTL service

Göteborg, January 5, 2022

Considering that
See below for making a rough prediction of processing time (wall clock).

Since December 5, 2021, the following has been arranged:
  1. The managing script includes a reshuffling of jobs an the basis of the time needed for computation and the time of waiting in the queue. Thus, recently submitted heavy jobs will be disadvantaged for some period.__
  2. When many requests are building up in the normal ("fast track") queue, the system resorts to side-tracking. It opens a directory folder where jobs from a few clients are placed in waiting, to be switched back into processing when the fast track is empty.
Suggestion:
Side-tracking can be avoided if clients with heavy requests (50 to100 stations and high-resolution ocean models) submit their requests at two to three hours intervals.
Redundant computations:
Since Dec. 26, 2021 the system rejects jobs with identical request parameters that are submitted within a short time interval (details here).
The system also rejects jobs with duplicated station names. This is a coarse criterion to avoid calculations when stations are located close together. The measure attempts to stifle carelessness in composing the station lists. Users can help to reduce the load on the computer by considering
  1. Computation of displacement at a station that only differ in the height coordinate with respect to another station is redundant (in gravity and coastal settings, it is not of course).
  2. In general, the loading parameters for displacement will not differ significantly if stations are closely collocated and sit far inland. As a rule of thumb: take the distance of the nearest coast to the centre of a station array, D, and the distance between array stations, dij. If  20 × dijD, the loading parameters will be identical at the sub 1-mm level. (Concerning gravity, the low number of such stations in the world makes us tolerate the extra computational load not withstanding its insignificance.)   If in doubt, you can bracket your cases and submit the a set of three sites with the shortest and farthest mutual distances, and submit consecutive requests as long as the parameters are different measured by your own criteria.
Before feeling tempted to announce a shame prize for the user (part of the email address) with the most redundant requests of the year, you can make a voluntary effort to the benefit of the user community.
Suggestion:
Take the two items above as a guideline and, in addition,
visit the job queue web page and utilise time segments when the system is idle or the queue is only occupied with requests for a low number of stations, one to 10.    
Rejected jobs, no results returning:
If you notice that your jobs have never entered the queue and are not in ongoing processing, don't resubmit them. Other hurdles can cause delays. e.g. email routing; or spam filtering in our central mail system at chalmers.se; or your institution's mail server has blacklisted the chalmers.se domain.

Requests are rejected due to missing keywords; results won't arrive if you misspell the email address  (direct mailers, please revisit the guidelines).

Tampering with the web page is to no avail - that transport system of requests, occurring locally on the computer named holt, has its own spam detector; it's very picky at the verification of a form filled in with the required information. For example, missing to select an ocean model in the drop-down manu reverts to a decoy DFLT model, used an indicator that a spammer has used a tampered page to send us ads (the OTL programs won't ever think of buying Rayban sunglasses or click on any hyperlink with a mouse it doesn't have). Sorry if you, decent user, have forgotten to select the model. For good reason you will not receive a notice.

The system does reply with an automatic message concerning a request denied
if
If you think your request should have been accepted but is not shown in the queue (granting a reasonable time margin),
  1. Using the web page http://holt.oso.chalmers.se/loading/ : send us a copy / screen clip of the "Thank you" page
  2. Sending direct mail to loading_kanelbulle_holt.oso.chalmers.se  : send us a copy of the mail
to hgs_kanelbulle_chalmers.se

If you must repeat a request because the results have been lost, just swap two lines of the station list or add a nuisance station, and the jobs will be admitted as they look formally different. The correction of a typo in e.g. the email address will also make the job different with respect to the lost one.

Ocean tide models with a high resolution of 1/8 × 1/8 degrees:
DTU10  EOT08a  EOT11a  EOT20  FES2004  HAMTIDE

Jobs with ocean tide models of higher resolution, 1/16 × 1/16 degrees
FES2012 FES2014b TPXO.8 TPXO.9 TPXO_atlas
are forwarded to the Cloud-based service, all others are computed at Chalmers/OSO. 
Models with lower resolution than 1/8 × 1/8 degrees do not pose the severe challenges this page elaborates on, mostly because they are not of much interest any more in precise applications in global geodesy. 

Job metrics:
High resolution model,
The exact surplus depends on the fraction of sea coverage in a three-by-three degrees area and the complexity of the coastline, which is hard to enumerate; 30 seconds is a typical median. 

The Cloud service needs about two minutes or less per station, similar to the one at Chalmers despite the higher model resolution.  

Presently, only the second job in a row (next in the queue) is rejected. We are going to widen the scope to the ten most recent submissions. Some of the identical submissions may occur by mistake, others owing to a user's impatience. The decision to reject a job is automatically taken right before the job goes into the CPU. 

Hans-Georg Scherneck, Machiel Bos