NEC briefed Serverwatch last week about the value proposition of its latest fault-tolerant server and software offerings. The basic argument two servers are cheaper than one. Sound logical? It actually does make sense when you look at it from a fault-tolerant perspective over a several year span. But first, let's cover what NEC offers in this area.
The latest FTserver from NEC is the Express5800/R320a. It is a two-socket server equipped with the Intel Nehalem chips (both the Xeon 5504 at 2 GHz and the Xeon 5550 at 2.93 GHz are available). This server will ship later this month.
The Express5800 is packaged as two servers in one box for fault-tolerance. It operates as a single server: If one fails, it fails over to the other one right away. As well as the Nehalem processors, most of its specs have been enhanced. That means 96 GB of RAM, 8 SAS drives and up to 5 PCI slots. The previous version, the Express5800/320Fd, will continue to be sold until the end of the year.
The other element is ExpressCluster WAN business continuity (BC) software, which fails over to another site. The software does both asynchronous and synchronous data mirroring. Synchronous operates for sites up to 200 miles apart.
"If you have your DR site greater than 200 miles away, you can switch to asynchronous for as far away as you need to go," said Ken Hertzler, director of product marketing for servers and software for North America at NEC Corp. of America, based in Irving, Texas.
Essentially, this is a software clustering product the servers in the secondary site are maintained in a hot standby mode. This reduces the startup time down to a minute or two for many applications. The product has been out since the beginning of this year and supports Linux, Windows and several virtualization platforms.
While NEC is pushing the combo of the Express5800 with the ExpressCluster WAN as a complete solution, Herztler said that the software cluster isn't married to NEC hardware.
"Availability can now be achieved with the implementation of both NEC solutions for a comprehensive business continuity plan," said Hertzler. "FTservers combined with ExpressCluster software over a wide area network will protect organizations against business disruption due to planned or unplanned system outages, while delivering maximum uptime."
When I was in high school, our math teacher delighted in proving mathematically how 1 = 2. NEC's trick is to show how two servers are actually cheaper than one. But it works out well as covered below.
The comparison involves a single server for a price of $5,000 with no BC solution hooked up. It would take $4,000 to restore it and an estimate of $50,000 for downtime. That's a conservative estimate at least when it comes to large enterprises. A Meta Group study puts the cost of downtime at more than $1 million per hour for industries such as energy, telecom, manufacturing, financial services and retail.So with no downtime over a five-year period, a price tag of $5,000 for the server without any fault-tolerant capabilities looks cheap. But just one major failure and that cost goes up to around $60,000 $5,000 for the initial hardware plus $4,000 for the restore plus $50,000 in estimated downtime. Two failures take you up above $100,000.
Now compare that to a solution like the NEC Express5800. $28,000 for initial hardware, no restores or downtime despite any number of failures and a five-year warranty is available for purchase. Hence, two servers are cheaper than one.
What about the contention that a software cluster is cheaper than fault-tolerant hardware? NEC numbers price the software cluster plus server at $14,000 to $18,000. Factor in warranties and consultant time for restoration of systems in the event of a failure, and you end up with a price of $24,000 for no failures over a five year span, 28,000 for one failure, $32,000 for two failures, $36,000 for three failures and so on.
"Software cluster costs include hardware, software, restoration, travel, consultant fees and so on," said Hertzler. "It's a myth that fault-tolerant servers are more expensive than software clusters."
He also points out the additional benefits to a hardware approach to fault tolerance: a single OS license, no SAN required, support for multiple OSes, no performance overhead, no configuration management and servers resynching in lockstep with no IT intervention required.
"Our fault-tolerant servers suffer no overhead for lockstep," said Hertzler. "Our Gemini engine chipset handles lockstep and not the Xeons."
Drew Robb is a freelance writer specializing in technology and engineering. Currently living in California, he was originally from Scotland where he received a degree in Geology/Geography from the University of Strathcyle. He is the author of Server Disk Management in a Windows Environment (CRC Press).