by Jane McConnell
OSDU Management Committee Member / O&G Data Architect, Teradata
I’ve been given the honour of writing the first blog for the new OSDU Marketplace website. It seems fitting to start with some history about why the whole OSDU endeavour exists, so I thought I would re-post an old article of mine called “10 Things I Hate About Subsurface Data Management”. You might have seen it before. I wrote it back in 2015, not long after I’d joined Teradata.
In my role within Teradata’s Oil and Gas Practice I have the fun job that involves not just explaining Big Data and modern data management to Oil Companies, but also to try to explain the weird world of subsurface data management to IT professionals who have never worked in Oil and Gas. The more I tried to justify what we had been doing in subsurface data management for the previous two or three decades, the more ridiculous it started to sound, even to myself. But on the other hand, the complexity of the data management problems opportunities that we find in subsurface was more than the traditional horizontal data management solutions – or those “expert” data management consultants – could cope with. Frustrated by the situation – I wrote what can most honestly be described as a rant.
Fast forward 5 years, and here I am, serving as an OSDU Management Committee member, part of this amazing coming-together of oil companies, oil service companies, cloud providers, technology companies, consulting companies – more than 157 member companies as of this writing – who are committed to building an open source data platform fit for the needs of the Upstream Oil and Gas industry.
I strongly believe that what we are building will give us a chance to tackle some of these fundamental problems that have hindered us for so long in subsurface data management. And in the follow-up posts I’ll talk more specifically about how the OSDU data platform allows us to address some of those “hates”. But for now, in all its glory, here’s the original rant blog post.
10 Things I Hate About Subsurface Data Management
Published on June 19, 2015
Talk about big data analytics in the subsurface domain and what’s the most common response? “Can’t change.” “Won’t work.” Sure, there are some perfectly valid historical reasons why we in the subsurface data management community look after our data the way we do, but isn’t it time to break from tradition? Why? Because new and evolving E&P business processes are driving whole new workflows and data architectures across much wider value chains, with shorter time frames and stronger cost control. If we don’t change our practices, if we do not build big data analytics into our data management strategies, we won’t be able to deliver what the business needs.
Here then are the top ten things that I (and lots of others) hate about subsurface data management today, things I see as barriers to change.
Seismic SEG-Y files go in one database. Well logs in DLIS, LIS, LAS in another. One for raw log files, and another one for corporate spliced and composited logs. Core data, yet another database. Same with core photos. Geochemistry – separate. Biostratigraphy – yup, you guessed it – different again. Why? Did we miss the lesson on modelling multiple types of data into a single integrated data model? Did we miss the lesson on high performance databases, where it’s OK to store millions of rows of data in a single database table?
Keeping data in separate systems with separate indexes, separate master data management issues and often separate physical hardware, only means extra work, master data management problems, and unnecessary hassle when we try to bring the data together so we can analyse it as a whole.
What could be worse than data silos? It’s got to be: storing data inside proprietary application databases or data structures. What’s the result? Pools of data that can only be accessed through an application API. Too many reasons why this is a bad idea, here’s just a few:
- The data storage strategy is based on requirements for application data access, not on data management principles. Governance and lineage are not a priority
- Most applications still store data sets as files or blobs for performance reasons, limiting your ability to use this data for analytics
- You are locked in to which applications you can run against the data store – killing your ability to choose “best of breed”
- Requiring access through an API prevents you from using mass-market SQL-based tools (visualisation, data quality, MDM) to manage and access the data
- You are beholden to the application vendor and the changes they choose to make to the data model across versions – and you need to implement their upgrades, which are often costly service engagements
Library style data management
OK – I understand the history. Our main role used to be to catalogue tapes. But now that the data is often kept on-line on spinning disk, why are we still cataloguing files as closed entities, “black boxes”, rather than cracking the files open and loading the contents into a data structure where we can work directly with the data? Is it because we always did it this way, or because we really don’t believe the alternative is possible? [See 1 – did we miss the lesson on high performance databases, where it’s OK to store millions of rows of data in a single database table?]
Project, corporate, or master?
As if we don’t have enough silos with our project databases, we thought we should add some more. Today’s E&P application vendors hawk a suite of solutions – the problem? Each only deals with a part of the data management problem. Sometimes the split between products is by design, but often it’s an accident – the result of acquisitions or solutions developed for a single company. E&P application vendors are not experts in data management. Don’t buy the marketing story about the reasons why you need a separate database for this one data type – they are just excuses.
Never fixing the data
When we find incomplete or incorrect headers, when we track down that the wrong CRS conversion has been used, or we finally identify the source CRS – why not fix the data once and for all? Why are we happy to preserve the mistake for posterity in our archived dataset? It’s one thing to have provenance and lineage, a system of record – but it’s another to refuse to fix the metadata or master data because “it’s the original”. Horizontal data management solutions have many options for maintaining a system of record.
Big data vs “lots of data”
“In the Oil Industry we have always had big data.”
No. In the Oil Industry we have lots of data. Normally in a cupboard, sometimes still on tape [See 3]. Some of it loaded (multiple times) into proprietary application silos that control what we can and can’t do with the data [See 2].
Big data is still our Achilles heel. Awkward and unwieldy data formats trip us up when we try to run analytics on that data, preventing us from realising the true value of all that costly-to-acquire data in its full business context. New tools and techniques could allow us to do things differently.
Decisions by PowerPoint
Billion dollar decisions are made on information presented on PowerPoint that contains no lineage information back to the original data on which the interpretations or models were made. Our use of siloed applications and manual data management makes it very difficult to forensically dissect previous decisions, and learn from our success or failure.
It’s what everyone else does
Well, everyone used to think the world was flat, and that the sun orbited the earth. Why are we so quick to dismiss alternatives for data management that have thrived for decades in other industries? If the O&G company that you benchmark against hasn’t adopted it, that doesn’t mean you shouldn’t. History doesn’t remember everyone, but it does remember Christopher Columbus and Galileo Galilei.
Unit conversion issues
With the amount of scientific data we have, there are literally thousands of unit conversions required. And we need it to be accurate. Which means knowing what units our data was recorded in. Metres, feet? Or for geospatial data – we have eastings and northings and we know the data is projected in UTM zone 32 but was the datum ED50 or WGS84? Get this one wrong and your position could be wrong by 200m – and that is not OK for a drilling target.
We know how important this is – and yet we are content to rely on the conversions built-in to applications, many of which are out of date or incomplete. When you couple this one with 5 [Never fixing the data] and 2 [Application silos], you get this ridiculous world where some people know that you shouldn’t use the “convert on unload” to export data from PetroBank MDS if it is ED50 north of 62 degrees and loaded before 2005. And the others? Well, they just have to prepare to fail.
The low bar of “not losing stuff”
We talk all the time about professionalising E&P Data Management, but at the same time we consider our role to be somewhere between geodata loading monkeys and librarians, with the low bar of not losing the data. Of course, there is the mundane, commoditised data custodianship that still demands fantastic domain expertise. But why are we selling ourselves short? We can provide a whole new set of capabilities, routinely used by other industries to add value to the business.
Does this list strike a chord? Does any of this sound familiar? I’d like to know what you think.
Oh, and please stay tuned for my next installment when I’ll start to explain how the OSDU Data Platform lets us look at subsurface data management differently.
Jane McConnell Bio
Coming from an IT background, Jane fell into Upstream Oil & Gas with its specific data management needs back in 2000, and has been consulting, developing product, and implementing solutions in this area since. She has done time at both Landmark and Schlumberger – in R&D, product management, consulting, and pre-sales – and in one role or another, she has influenced subsurface data management projects for most major oil companies across Europe. She chaired the Education Committee for the European oil industry data management group ECIM, and regularly presents at Oil Industry data management events. Along the way she’s learned enough about geophysics, geology, and the world of hydrocarbon exploration and production to be dangerous.
For the last 8 years has been working in Teradata, a big data and analytics company, in their Oil and Gas Practice. She is also serving on the OSDU Management Committee.
Jane is Scottish and has a stereotypical love of whisky (especially Islay malts). She knits and reads almost to obsession.
You can contact her at firstname.lastname@example.org, on the OSDU Slack channels, or connect on LinkedIn.