64-bit processing and DBMS

For over a year now AMD has been delivering a 64-bit processor which Intel will finally match in the next few months with a processor capable of 64bits in x86 mode. On the OS side , since the beginning of 2004 there have been ports to AMD64 Opteron-based machines from Suse and Mandrake. Now there are at least 8 implementations of AMD64 including Debian, L-OS, RockLinux, Gentoo, Fedora/RedHat, Knoppix versions of Linux plus FreeBSD and Solaris Server OS. Blackdown Java 1.4.12 supports 64bits.

On the Microsoft front, Redmond has announced support and has issued beta code for AMD64 but will officially release Windows 2003 64bit in early 2005 for use with Itanium, Xeon , and AMD processors (but in what order has not been specified). See the following sites for more details: x86-64 and TuxMobil .

The bottomline is that the traditional power Dell plus Microsoft have been late to the 64bit low-cost parade. Linux, OpenSource, HP, IBM (but not aggressively) and Sun have been leaders on the 64bit bandwagon. This is important because low-priced, 64bit processors are going to have an enormous impact on database and software development.

Database Point of Inflection

Duke University is going to give to each and everyone of its 2004 Freshmen a 20GB Apple iPod. Why an iPod ? Because having 20GB of highly portable, easily loadable storage space is atractive to the univesrity and its many departments. But for our purposes it also important to help underline the fact that storage has become so cheap. If you check Froogle.com you can easily find a 100GB hard drive for around a $100. And all experts are calling for that price to drop substantially or the equivalently priced drive to at least double in size by early 2005.

Storage price-performance has been “outperforming” computings vaunted Moores Law for the past 4-6 years. Disk storage has been doubling in size every 8-12 months for the same price compared to 18-24 months for computing power. But all that storage has been waiting for an operating system+hardware cheap and fast enough to take advantage of the gigabyte going onto terabyte surplus. But 32 bit processors are confined to a 4GB maximum memory space. And database applications just eat up memory and computing power both. Yet database implementation opportunities abound; they are awaiting the right price-performance and ease of development tools.

Enter the Opteron with 32 or 64GB of memory capacity and roughly double the speed of any 32bit processor at the same MHz rating. This is music to DBA and database application developers. For 1/10th the cost of current top tier Unix, mainframe and Windows systems, they can load up their complete databases into a dual or quad CPU servers memory and deliver up to 10 times the speed.

Let me repeat for those not attending – 1/10 the cost yet up to 10 times the speed for big databases now – soon to become huge. In calculus class this is called a point of inflection – when the rate of change itself starts to accelerate. And Harvard Business School says there is no great shakes IT innovations left to happen. Hmmm … we would beg to differ.

And what operating system has been delivered with an update that greatly enhances the memory size it can support plus number of concurrent CPUs ? No – Windows 2003 64bit is not due out until early 2005 … no make that mid year 2005. Rather the Linux 2.6 kernel has been out since spring and its loaded for bearing the huge database and memory requirements engendered by 64bit CPUs. And as we noted above it has the longest track record for supporting the cheapest and fastest x86 AMD Opteron chips. Solaris and FreeBSD are not far behind.

Forget Longhorn, its 3D GUI, radical rewrites, blatant attempt to destroy the browser and web interface by absorbing those functions into the OS, plus bloated conversion and hardware requirements. The real OS action for the next 2-5 years is going to be cheap 64bit processing and ironclad security, reliability and availability. And like Lance Armstrong reaching the Pyrenees and then the Alps, it looks like Linux is the front runner with Solaris (Sun has at least 3-4 years of 64bit experience to draw upon) and FreeBSD being serious challengers. Microsoft 2003 64bit is more like sprinters in the cycling flats – great show but still having lots to prove in the Enterprise mountains.

And of course the key to 64bit computing race is what OS software can deliver the best database, application server, and development software package. So it is no surprise that Visual Studio 2005 and SQL Server 2005 have been delayed to be fine tuned (rumors abound about getting performance and security out of .NET 2.0) and to await Windows 2003 64bit. But BEA, CA, IBM, Oracle, Sybase, Sun and dozens of others have been preparing for the new 64bit reality. What will be interesting to see is who among them can jump to the 64bit hyper-database-space with the most elan ?