Clustering is not the only way to ensure server availability. Fault tolerant servers with multiple redundancy built in hardware, hitherto mostly used in the financial and telcos sectors, have crashed down in price and are making their way up the manufacturing ‘food chain’ from SCADA (supervisory control and data acquisition) systems towards ERP. Brian Tinham reports
Clustering is not the only way to ensure server availability. Fault tolerant servers with multiple redundancy built in hardware, hitherto mostly used in the financial and telcos sectors, have crashed down in price and are making their way up the manufacturing ‘food chain’ from SCADA (supervisory control and data acquisition) systems towards ERP.
Fault tolerant systems were expensive, largely proprietary beasts but in the last 18 months suppliers like Stratus Technologies have launched on mainstream Windows 2000 and Intel platforms – in Stratus’ case, bringing the cost down from £180,000 to £18,000.
And with 99.999% availability, that price/performance that compares very favourably with traditional mirrored cluster alternatives – not only on hardware costs but on the additional software, operating system and database licenses no longer required.
Jeremy Lovatt, European business development director for Stratus, claims the firm’s Wintel offering is now one third cheaper in terms of total cost of ownership than equivalent clustered high availability systems, and “considerably cheaper than that” where Unix is concerned.
Until recently, its systems have been used primarily on plant-critical HMI (human machine interface) and SCADA systems (running Iconix and Yokogawa process monitoring and control, for example) and for key factory shopfloor management systems. But, with the advent now of four-processor (2.4Ghz) FT Servers plus performance scaling with simple horizontal clusters, the bar is being raised to business systems at least in the mid range ERP space.
Systems like this will help beleaguered IT managers shift business thinking on critical IT infrastructure by enabling performance and availability SLAs appropriate to operations at what’s considered a realistic price.
Dave Chalmers, also with Stratus, says: “What we do is put the hard stuff down in the hardware and make availability simple. Our processors are now adequate for 80% of the market. In fact we’re now engaging with SAP on the Windows side for certification.”
Putting it in the hardware is inclusive: he means the CPU, all associated modules, boards, comms and I/O. There are also additions to the operating system to eliminate memory leakage – technology which Stratus is now sharing with Microsoft and which will appear in its .Net server later this year.
Incidentally, the system also captures application problems during runtime, enabling IT to trap and analyse application errors and provide a path to continuous system and robustness improvement.