HEPDB

Forget ADAMO tables: Make Entries in a Notebook

2006-04-12T00:56:00.000-07:00

Data Structures for Particle Physics Experiments: Evolution or Revolution? - the 14th INFN Eloisatron project.

We all bogged off to Erice in Sicily for a fun workshop in November 1990, fighting out the battle between memory managers (Zebra, (Y)Bos, Cheetah, Jazelle, ...) and languages (Fortran, Fortran 90, C++, ...)

Quite a lot of the focus was on data modelling at a time when entity-relationship modelling was all the rage in HEP.

ERICE = "Even Rene Is Convinced Eventually"

My dinner-time contribution was "Can A Relational Model mean Entirely Nothing?"

Can I Have My Data Back?

2006-03-03T03:32:00.000-08:00

A marvellous anecdote in "The Difference Between God and Larry Ellison" worth (IMHO) buying the book for...

Back in the days of the VAX - where the early Oracle service for the construction of the LEP machine was run - we were forever restoring data from backup tapes.

Unfortunately, the tape handling of the VMS BACKUP utility was far from perfect and often led to the wrong tape being mounted. Worse still, current practice at that time was to reinitialise (label) tapes before each use, meaning that you could just lose last night's backup...

On one such occasion, temperatures were running high. (Important data had been lost / corrupted and we needed to get it back - trouble is, the backup tape had been relabelled and was now 'empty'.)

Luckily, I had just been working on some new tape handling software (bits of which were later used by some of the LEP experiments online - also at FNAL and SLAC - but that's another story...)

Putting a bold face on it, I casually walked into the machine room and mounted the offending tape. Casually skipped past the offending tape mark and copied the data over.

BACKUP successfully restored the file - complaining mildly about wrong block count in the trailer label.

My reputation as a VAX guru and wizard had been made.

Pity we no longer use 'em.

Odd Mugs

2006-02-21T04:46:00.000-08:00

The standards group for Object Databases was called the Object Data(base) Management Group. I was the representative for CERN, along with a bunch of database vendors (O2, Poet, Versant, Objectivity etc.)

Chaired by Rick Cattell, the group used to flip between East and West coast - and at least thrice to Europe (Annecy - the "CERN" meeting, Versailles - O2 - and Nice). East coast included the Florida keys and once Providenciales, in the Turks and Caicos Islands. The TCI used to have the slogan "Where on Earth are the TCI?". Noone really seemed to know - the Islands flag used to feature a salt heap, as that was one of the main industries there (they turned out a substantial fraction of the world's salt, until it magically became uneconomic, just about the time of the Cuban missile crisis when it was important to have a military base "nearby". Great Inagua - nearby - still has a working salt industry). Someone this got misinterpreted as an igloo - somewhat unlikely for a Caribbean island.

Although I didn't know it at the time, at least one person previously at the SSCL had bought a property directly on Grace Bay. I soon found out though, when on an otherwise deserted beach, someone walked up to me and started asking questions about HEPDB...

So much for going incognito...

Seven Wonders

2006-02-15T22:33:00.000-08:00

How many people can name the Seven Wonders of the Ancient World? Not that it matters much - there's only one left (plus fragments that found their way into the British Museum in London).

VLDB 2000 was held at the foot of the Great Pyramid of Giza. Apart from predicting yotabyte databases by the year 2020, this was somehow a turning point (in time) for databases in HEP.

More interesting, perhaps, is the riddle of the Rosetta stone, where the physicist known to all school kids - Thomas Young - played a key role. (The real Rosetta stone is also in the British Museum, which is how I know about the fragments of the Seven Wonders also there...)

Going inside a pyramid has very much a "Raiders of the Lost Ark" feel about it, but has to feature somewhere on the 1001 list... (Luckily, I've already done it...)

(There's all sorts of interesting stuff about the evolution of the pyramids themselves, the Sphinx, good ol' King Tut, but you really need to go and see for yourself).

Lost in Translation

2006-02-14T22:32:00.000-08:00

I gave a talk on "Building the world's largest scientific Grid" at an Oracle Tech day in Moscow, 2004. My slides were translated into Russian and there was simultaneous translation.

I was really impressed with the interpreter - every time I made a joke the room erupted in laughter.

I found out later what the translation was - "He's making a joke - laugh".

CHEPDB

2006-02-13T22:29:00.000-08:00

Check out the 'Databases in HEP' presentations given at CHEP 2006 (http://www.tifr.res.in/~chep06/) in Mumbai and at IWLSC 2006 (www.veccal.ernet.in/~vecpage/tkmp/vecc/symposium.htm) in Kolkata.

http://indico.cern.ch/contributionDisplay.py?contribId=12&sessionId=4&confId=048

From the Databases in HEP panel at CHEP 1992 in Annecy, France until more or less the present day...

The Years of FAT

2005-10-07T08:27:00.000-07:00

My first visit to Oracle HQ was during 'the years of FAT' (i.e. when working on File and Tape Management: Experimental Needs - FATMEN).

(On the way back from an IEEE Mass Storage Symposium in Monterey).

The idea was to discuss with "Smokey" on some possible future directions...

I suggested a distributed lock manager.

Na, doesn't scale.

I was somewhat amused when a while later they came out with... a DLM...

Ah well... Can't win 'em all...

Whither HEPDB?

2005-10-06T05:21:00.000-07:00

So during my 2nd DB encounter, I was involved in something that some viewed as a File Catalog, but in reality offered 'location and platform independent access to data'... (etc.) Sounds familiar?

The main target was the LEP experiments at CERN (adopted eventually by DELPHI, L3 and OPAL, but not ALEPH).

Whilst I was touting it around, an oft repeated comment was "Very interesting but what we really need is..." The 'this' was a detector conditions database. Well, quite a few common ideas featured in the file catalog and the conditions DB solution of one of the experiments and an informal suvery revealed significant overlap with the functionality of that used by another (and indeed many other HEP experiments at CERN and elsewhere.)

So a 'common project' was born to produce an experiment-independent solution to a common problem. The name adopted was "HEPDB", with the implementation re-using the DBL3 package and also many of the ideas from OPCAL (from OPAL).

HEPDB runs still on node hepdb.cern.ch, accessed through /hepdb.

So Groundhog Day again...

Persistence...

2005-09-28T09:11:00.000-07:00

Somehow I don't seem to get away from databases. Either I follow them, or they me.

My first job for CERN was to support the VAXes on which Oracle (and other applications) ran. After 5 years, I moved to 'File and Tape Management', which meant both Oracle and SQL/DS (of which more later.) (Another ~5 years...)

As we prepared to make the jump to OO, Object Databases and the Object Data Management Group (ODMG) came up. (The next 1/2 decade...)

And then it was back to Oracle. (For guess how long...)

Even now, when focussing on getting the LCG Services in full operation ready for LHC operation, most of the services involve a database - Oracle and/or MySQL.

Well, not so many hemi-decades to go, unless I take it up as a hobby later...

Why this blog?

2005-09-28T07:33:00.000-07:00

I have been asked to give a talk on 'the evolution of databases for HEP' (see below). As there are many stories related to databases & HEP - some not necessarily suitable for a conference - but still part of our collective history, I thought that I'd start a blog where I recorded some of these.

Hopefully not too painfully ego-centric... We'll see...

Dear Jamie,
The next CHEP has quite an important role, as it is the last before the "scheduled" data taking. And even if data taking is delayed this will be the last time we can meet and think...

...

We will ask people to present contributions of a more "philosophical" nature, a sort of critical review of their major field of action. HEP-LHC computing has gone through a period of high- turbulence in the years 1993-2002, and then dust has settled (I hope :). In dawn of the "proof of reality" I think it will be good to look back and try an objective review of what happened, why it happened, why it worked and what did not.

...

I would be very glad if you could come and give a contribution on the evolution of databases for HEP, and your view on it. We would like to assemble the contributions and publish proceedings of this meeting, a sort of short history of LHC computing. Suggestions are of course welcome! Let me know,

Federico Carminati