A former colleague of mine once sold one Megabyte of memory for a million dollars.
This was in the late 1960s, when 1 MB was actually a lot of memory – but all the same, $1,000,000 was a lot of money, especially so back then. Fast forward to the mid-1990s and technology had moved on to the point where even the original Sony PlayStation games console came with a 1 MB memory card.
Nowadays, memory (even on portable devices) is measured in Gigabytes and that million dollars of memory can now be had for a fraction of a cent. Relational databases were invented in the 1970s, when memory was rare and expensive. As a result, databases were designed to use disk to perform substantially all of the data processing – and disk is several orders of magnitude slower than memory. You would have thought that the coming of plentiful, affordable memory would have profoundly influenced the design of these database …
… Except that it didn’t. Most databases from the 1970s and 1980s that are still on the market today, still have a design that dates back to the days when memory was a million dollars per megabyte. Exasol took a different approach. Based on research in German Universities in the 1990s, we have built the world’s fastest database – not by doing anything complicated with disk access, or by building on an existing out-dated disk-based databases – but simply by using memory as the medium for performing substantially all data processing activity.
The upshot is a database that is both extremely fast, while being simple to use, maintain and implement. It is also designed to optionally run on a cluster of machines – either physical, virtualised or in the cloud – and so it is capable of scaling to cope with hundreds of terabytes of data and beyond.
The Exasol database doesn’t make the disk-based databases redundant – there is still a need for on-line transactional processing for example. But if you are using such a database for analytical queries, widely used in Business Intelligence applications, then you definitely have the wrong tool for the job. The main problem you will find is that query performance will drop off a cliff as data volumes increase.
You should expect analytical queries to complete in a matter of seconds and not a matter of hours. If you find yourself unable to achieve this with your existing database, then I can highly recommend that you try out the Exasol database. Within a few hours, you could be running your queries on your data and seeing for yourself the difference that in-memory processing can bring.
Just sign-up here – either for a virtual machine you can install on your own hardware or for an account on our EXACloud hosted service.
It won’t cost you a cent – which you could instead use to buy 1.74 MB of memory at today’s prices! (1)
(1) 2x 8GB DIMM DDR3-1600 = $91.99 + free shipping