Most organizations aspire to high availability. However, high availability typically comes at the cost of implementing highly complex and difficult to manage solutions, such as Microsoft’s failover clustering. Stratus Technologies’s Stratus ftServer 4500 is a high-availability server that can provide five 9’s of availability with very little added complexity. Much like the NEC 5800 unit that I reviewed earlier this year (see “NEC Express5800/R320”), the Stratus ftServer 4500 is a 4U rack-mounted server that provides dual copies of all system components. In other words, there are two motherboards, two sets of CPUs, and two sets of RAM and storage. Each of these sets is contained in its own 2U unit. Stratus calls each unit a CPU-I/O Enclosure. Each of these CPU-I/O Enclosures slides into a shared rack-mounted chassis. The two CPU units run in lockstep, with the current memory and CPU instructions shared between each of the two CPU-I/O Enclosures. Figure 1 shows the Stratus ftServer 4500.
Figure 1: Stratus ftServer 4500
From when I first opened the shipping container, it was apparent that the Stratus ftServer is in a league apart from ordinary servers. Instead of arriving in a plain corrugated box, the Stratus 4500 arrived on a palette. The Stratus ftServer shipped in three main pieces: two CPU-I/O Enclosures and one industrial-strength steel chassis. These components were quite heavy, so the ftServer installation took a little doing. To install the unit, I first installed the chassis into my server rack and then slid each of the CPU-I/O Enclosures into the chassis. I used thumbscrews at the front the unit to secure the CPU-I/O Enclosures. The chassis provides internal connectors that are used to plug in each CPU-I/O Enclosure. These connectors are how the CPU-I/O Enclosures communicate and stay in sync.
The Stratus ftServer 4500 that I tested came equipped with two logical Intel Xeon E5504 quad-core CPUs running at 2GHz. The system also made use of the Intel 5500 chipset. The unit that I tested came configured with 16GB of RAM and 136GB of Serial Attached SCSI (SAS) disk storage spinning at 15K rpm. In the case of this system, the key word is logical because the ftServer actually has two physically matching sets of CPU, motherboard, RAM, and disk storage—one set per CPU-I/O Enclosure. This duplication of system components is what enables the fault tolerance. Each CPU-I/O Enclosure can support up to 96GB of RAM running at 800MHz and up to 4.8TB of SAS disk storage.
Internally, each CPI-I/O Enclosure had two PCI Express 2.0 expansion slots and four more optional PCI Express 1.0 or PCI-X expansion slots. On the back of each CPU-I/O Enclosure, there were three 1GB network ports. Two of the network adapters were intended for client networking activity, whereas the other network adapter was reserved for remote management. In addition, each CPU-I/O Enclosure also had an additional two-port 1GB network adapter. Between both of the CPU-I/O Enclosures, there were eight client network ports, which were configured as a team using Intel’s Advanced Network Services (ANS) technology. This teaming technology provides networking fault tolerance.
Each CPI-I/O Enclosure in my test unit also had a Fibre Channel adapter. The connections for the video display, keyboard, serial ports, and USB ports were on the chassis—not on each CPU-I/O Enclosure. The video used a standard nine-pin VGA port. An integrated video controller provided 8MB of RAM and supported a maximum of 1024 ´ 768 display resolution. Notably, the Stratus ftServer 4500 had no PS/2-style mouse and keyboard ports. The mouse and keyboard connections are USB only; you can use the port on the front of the unit or the three USB ports on the back of the unit. Because two of these ports are required by the mouse and keyboard, I wished the unit had more USB ports available—especially on the front of the system. The front of the chassis also provided a vertically mounted DVD-RW drive.
The Stratus ftServer 4500 that I tested came with Windows Server 2008 R2 x64 Enterprise Edition preinstalled. You can also order it with VMware vSphere 4 or Red Hat Enterprise Linux (RHEL) 5.
Despite its fault-tolerant configuration, managing the system was essentially the same as managing a standard Windows Server system. All the management tools that you typically use, such as Control Panel, Services, Event Viewer, and Device Manager, were essentially just like you’d expect. In addition, there was an ftServer Management Tools icon on the desktop that lets you work with the fault-tolerant configuration. The Stratus ftServer 4500 provides a remote-management facility called the Virtual Technician Module (VTM), which Figure 2 shows, that lets you manage the system remotely. Notably, the VTM lets you power the server on and off. The VTM works even when the system is in a powered-down state, because as long as the system is connected to power it never completely powers off. At idle state, the system consumed about 53W.
Figure 2: Virtual Technician Module
To test the system, I configured four Microsoft Hyper-V virtual machines (VMs). Each VM was running Server 2008 Enterprise and a single instance of SQL Server 2005 Enterprise Edition. The VMs were configured to use 512MB of RAM, and the VM files were stored on the local drives. This test suite consisted of a mixed workload of database queries. The database tests ran a set of 27 queries against each virtual SQL Server instance. The Stratus ftServer 4500 proved to be an excellent performer, with test scores comparable to the other high-end servers we’ve tested in the Windows IT Pro labs.
When the system is running in fault tolerant mode, there’s a Ready to Pull light that’s lit on the front of the unit. To test the ftServer’s fault-tolerant capabilities, I pulled the plug out of the back of each of the units while the system was running the test workload. In addition, I tried removing the network cables from one of the units, as well as pulling out the hot-swap drive. In all cases, the unit lived up to its five 9’s reputation and continued to function with no end-user interruption. The ftServer 4500 continued running the workload with no noticeable slowdown and absolutely no interruption of services.
After I reconnected the power to one of the CPU-I/O Enclosures, the unit took a few minutes for the two CPU modules to resynchronize. Again, the workload continued to run with no interruption and no noticeable slowdown. The resynchronization process was completely automatic, and there was no manual intervention necessary.
The time required for synchronization depends in part on the workload the unit is handling. Under heavy workload, the resynchronization took about 10 minutes. When the system was idle, the resynchronization completed in about a minute. During the resynchronization period, the unit isn’t fault tolerant, and I needed to wait until the Ready to Pull light was relit to perform another test. When the Ready to Pull light came back on, the ftServer 4500 was once again fully fault tolerant. While running under the workload generated by our virtualization test suite, the unit consumed about 574W.
A unique availability feature in Stratus’s series of computers that goes way beyond the availability offerings from most vendors is Stratus’s ActiveService technology. ActiveService lets the server automatically contact Stratus support and even automatically order replacement parts if a hardware failure is detected. Stratus informed me that replacement parts are shipped next day. I didn’t experience any real hardware failures during the 3-month testing period, but I did run across a couple of problems during testing, related to misconfiguring the unit. Stratus’s support provided expert help and was able to resolve my issues quickly. ActiveService customers have 24 ´ 7 support, and support personnel can connect to the server remotely for problem remediation.
Overall, I found the Stratus ftServer 4500 an excellent choice for a high-availability server. The ftServer 4500 brings five 9’s of availability, at a price that’s within the reach of most businesses. In addition, managing the unit is almost the same as a standard server. If you’re in the market for a new server for a mission-critical workload, or you’re looking into other high-availability technologies, I highly recommend the Stratus ftServer 4500.
Stratus ftServer 4500