We are all familiar with storing megabytes, gigabytes, and even terabytes of data but few have stored petabytes and even less have reached an exabyte of stored data. The European Centre for Medium-Range Weather Forecasts, ECMWF, has reached this data storage pinnacle by storing an exabyte of data on tape technology. Two of the key scientists from this organization were recently interviewed by the LTO program to get the inside scoop about this milestone achievement. This LTO BlogBytes article will review highlights from that interview and provide access to the complete conversation.
HOW BIG IS AN EXABYTE?
How big is an exabyte of data? It is a one followed by 18 zeroes which, to give that number context, equates to over 350 billion digital pictures. The ECMWF did indeed reach this immense data storage threshold, but first, what does the ECMWF do? As you will learn in the interview, this organization is a research institute and a round the clock service that produces global numerical weather predictions and other data for the European community.
They have one of the largest supercomputer facilities and meteorological data archives in the world. In a recent interview, Stephen Richards, team leader for the data handling system, and Ian Randall, senior system analyst, told how they were able to reach this threshold as well as move their complete data center to a new location in Europe. Let’s review some of the highlights from this candid interview conversation.
THOUSANDS OF TAPES
As mentioned, ECMWF has a very large high performance computing facility and Stephen Richards shared details about the growth of their storage system and specifically stated that “we’ve grown probably on average over … the last 10 or even 20 years around 30 or 40 percent per annum.” When asked to define their tape system environment Ian Randall stated that they have over 60,000 data tapes and 10 tape libraries with the backup copies of data stored on LTO tape technology which includes a mix of LTO-8 and LTO-9 tape.
So just how busy are these tape systems? Stephen said, ”It’s in the region of about 5,600 (which) would be a normal sort of mounting (cartridges per hour) or exchange rate. But we probably peak closer to a thousand mounts an hour out of the (tape) libraries…results in keeping 400 tape drives busy. We sort of use the tape archive as if it’s disk.”
VOLCANIC ERUPTIONS – BUSY DRIVES
As noted, the ECMWF stores tons of data. When asked how far back they keep the data, the answer was described by Stephen: “…for all observations and also … our model output, we will try and keep that for forever. We do an awful lot of reanalysis work, which tends to be the more valuable piece for … other research institutions and universities and commercial customers.” Ian went on to describe one example of access to the massive data archive where researchers from around the world inquired about the Icelandic volcanic eruptions and then other volcanic explosions for comparison, projection and analysis purposes. You can imagine that the tape drives were busy.
LET’S MOVE ACROSS EUROPE?
There was a detailed discussion of the system configuration and also how they moved the entire huge data center from Reading, England to Bologna, Italy. One of the key considerations for the move was how to keep the operation running and move all of the systems, active data and archived data to the new location. Moving hundreds of petabytes of data over a wide area connection was not practical and could take years. Stephen explained, “So really we looked at ways of just being able to run our operational service and have enough disk for about two weeks, although we knew we should only need it for about four, five, six days. So we switched off everything in the UK and we just basically ran the service from disk in Italy.” The tape systems were transported on trucks. Stephen added that “The (tape) media all went in flight cases and basically, as the libraries were being built (in Italy)… we would place the media back in the library.” More detail about this extraordinary move is found in the full interview.
As you can tell, ECMWF has a remarkable operation with tens of thousands of tapes, many tape libraries, and many tape drives, with a mountain of information to manage. There is a lot more nitty gritty info to be gleaned from this captivating discussion. Listen to the full conversation here. Enjoy!
.
