UK-based delivery company, TNT Express, has completely refreshed its aging Teradata systems, as the amount of data it is holding has increased tenfold in five years and the level of analytical queries carried out by employees has soared.
Computerworld UK spoke to Mike Robinson, database and business intelligence manager at TNT, who explained that the company has a longstanding relationship with Teradata and was faced with the decision to either add on some new hardware to the company’s existing kit, or to conduct a ‘floor sweep’ and carry out a complete refresh.
TNT Express was using three generations of existing Teradata hardware, which had been bolted together and added on over the years, to carry out both traditional business intelligence reporting and real-time analytics across operations.
“We use Teradata as an enterprise data warehouse, which is the data source for our IBM Cognos reporting system. We do traditional reporting, where a batch job will run and create a data output that will feed into a report for someone in Head Office,” said Robinson.
“However, we don’t only run business intelligence reports, we also have a number of other real-time systems that we use the it for. One example is our customer service system, which allows a customer support person to get a call from a customer and pull up 13 months’ worth of historical transactions from the data warehouse in fractions of a second.”
Robinson explained that bolting on new hardware to the existing kit wasn’t the best option for TNT, as the performance benefits of the new technology would be slowed down by the aging systems.
He said that when putting the business case together, TNT assessed that it can receive a return on its refresh investment in less than three years by upgrading to the new 4 node 6690 with hybrid storage SSD/HDD kit.
“We had utilised all of the resource that was available across those three generations of boxes. The oldest generation we had was almost six years old and it was coming to the end of its life. We had long discussions with Teradata about the benefits of adding a fourth generation onto that, but one of the issues is that the new system would only run as fast as the slowest box,” he said.
“So we would be spending quite a lot of money to bolt on a new box but we wouldn’t be getting the full benefits. However, if we did a full floor sweep we could get the benefits of everything on the floor and all the new technology that comes with that.”
TNT got the new hardware delivered in November of last year and has been working with Teradata to transfer the data across.
According to Robinson there is always a concern that in transferring the data from the old system to the new one there will be a problem, but he has run this process a number of times in recent years and never experienced a glitch.
“Teradata has a process that lifts the data off of the old boxes, where a server basically sits in-between the old and new systems and transfers it over. As part of that process it rebalances the data evenly across all the spinning disks and different nodes,” said Robinson.
At the time of speaking Robinson and his team had just completed TNT’s first monthly batch job using the upgraded Teradata system. This batch job is the company’s biggest processing suite and usually takes approximately six hours to complete – with the new technology it takes just three hours. Robinson said: “We were all a bit stunned as to how quickly it ran.”
Robinson explained that Teradata now has enough capacity to see TNT through the next three years and the recently implemented system is only operating at 67.5 percent. This increased capacity was important given the data surge the company has experienced in recent years.
“We currently have just under 20TB of data on the system – in 2008 we were at 2TB. In that period we have gone from running just standard reports with users running the occasional ad-hoc query, to becoming a lot more knowledgeable about what the data warehouse can bring,” he said.
“As a result there have been a lot more applications that have come along, for different parts of the business – some in sales, marketing, operations etc.”
He added: “Three years ago we were running a couple of thousand tactical queries a day, we now run on average 45,000 an hour.”