Server partners Stratus Technologies and NEC have revamped their fault tolerant server lineups to take advantage of Intel's quad-core Nehalem EP Xeon 5500 servers.
The two companies typically like to get their fault tolerant machines out the door within a quarter of a new chip launch from Intel.
This time, though, the big shift from the old frontside bus on Xeon chips to the new QuickPath Interconnect used with the Xeon 5500s, plus other changes in the system BIOS, forced them to do more engineering work and testing to hit the market. In a few weeks they will begin shipping the fault-tolerant machines they created together, with their respective labels.
Fault tolerant servers are distinct from the more popular clusters in that they are two completely mirrored systems running two copies of an operating system and their applications are kept in absolute lockstep by a chipset and electronics in Intel's Xeon chips.
High-availability clusters, by contract, link one or more server nodes and replicate data between machines so they can take over each others' work in the event one node in the cluster goes down.
Historically, clustering has been cheaper to do even if it is more complex for IT shops to manage. The advent of x86 and x64 servers from Stratus and NEC, though, have brought the price of an fault-tolerant setup down even as the performance has gone up with every successive Xeon chip generation.
With support for the Xeon 5500s, even a two-socket fault tolerant box is going to have enough oomph for a lot of workloads, and for many kinds of applications - such as police, fire, and other emergency responders, who like fault tolerant boxes. These are environments, though, where the IT skills are a little thin and paying extra for mirrored machines that manage themselves is easier than trying to build and support an high-availability fail-over cluster. And where the bugets for doubling up on software licenses are kinda thin.
To make the current generation of fault-tolerant machines - this is the fifth generation of Xeon-based FT boxes from Stratus, but the sixth generation from NEC, which did its own before partnering with Stratus - the two companies collaborated on the design of the GeminiEngine chipset. This chipset accesses the lockstepping functions inside the Xeon chip and allows for the two server modules in the fault-tolerant box to be kept in absolute synch in terms of CPU, memory, disk, and network processing.
Stratus and NEC also tweaked the server's BIOS to support their own system monitoring and management software. NEC actually manufactured the fault-tolerant machines, much as it manufacturers the high-end of the Unisys SMP server line using Intel's six-core Xeon 7400 processors. These use the old frontside bus and scale up to 96 cores under a similar partnership that has Unisys and NEC peddling the boxes independently.
The basic hardware in the fault-tolerant machines are the same, with Stratus and NEC only supporting the 2GHz E5504 and 2.93 GHz X5570 versions of the Nehalem EP chips. The two-socket server module supports six memory slots per socket, and can use 1GB, 2GB, 4GB, or 8GB DDR3 memory, for a maximum of 96GB of main memory.
This is the logical capacity of the fault-tolerant machines, of course. Physically, there are two complete servers in an fault-tolerant setup. The system includes the management controller in the base price, which is new this time around. Customers used to have to pay extra for that, and were no doubt annoyed since it is really a requirement.
Stratus and NEC package the boxes a little differently. NEC's Express5800/R320a-E4 and Stratus' ftServer 4500 are essentially the same machine, using the slower E5504 processors - with either one or two logical processors, or four or eight cores, in the box - and either 48GB or 96GB of logical memory. The machine has two logical Gigabit-Ethernet ports and two integrated PCI-Express 1.0 peripheral slots. The machines have an optional additional two PCI-Express 2.0 or two PCI-X slots.
NEC's Express5800/R320a-M4 is essentially the same as the Stratus ftServer 6300. These machines come with two logical X5570 processors and up to 96GB of logical memory. The machine comes with the two base PCI-Express 1.0 and both vendors support an additional two PCI-Express 2.0 slots. All four of these machines come in a 4U rack unit - that's two 2U servers stacked.
The server chassis now has a shared DVD for both nodes in the fault-tolerant cluster, and the move to 2.5-inch disks means that customers can now put eight drives into a server node. That's an increase from three in the prior generation, and users can stripe their data for performance if they want to.
To chase smaller customers with lower budgets, Stratus is selling a slimmed-down machine, the ftServer 2600, which only supports one logical processor socket, up to 16GB of main memory, and only has the base two PCI-Express 1.0 slots on the system board of each node.
According to Denny Lane, director of product marketing at Stratus, this entry ftServer 2600 with one processor and 4GB of memory can be had for $13,000 on the street.
A reasonable midrange configuration of the ftServer line runs in the range of $25,000, according to Lane, and a high-end configuration can run to around $40,000. Fully loading an ftServer with the latest Intel processors, plus lots of memory and disks, can push the price up towards $100,000.
NEC's director of product marketing Ken Hertzler says the suggested list price for a base Express5800/R320a-E4 with no memory or disk runs to $17,000, with a decently configured box costing between $20,000 and $25,000. A heaver configuration with two logical processors, something in the range of 20GB of memory, and four disks costs over $30,000.
Both Stratus and NEC say they will start shipping their respective boxes using the Xeon 5500s within a few weeks. Both will support Microsoft's Windows Server 2003 and Windows Server 2008 on the machines at ship time, and plan to get support for Red Hat's Enterprise Linux 5 out the door in the fourth quarter.
Stratus says it will have support for VMware's ESX Server 4.0 hypervisor and its vSphere 4.0 management stack on the new ftServers by the first quarter of 2010. NEC merely says it has support for other operating systems and hypervisors coming down the pike.
Customers who need to support ESX Server 4.0 or RHEL 5 today can run them on the earlier ftServers from Stratus that employed the Xeon 5400 processors. NEC supports RHEL 5 and ESX Server 3.0 on its earlier generation of fault-tolerant machines. ®