Linux.com

Feature: Software

Review: Stratus ftServer 4300

By Joe 'Zonker' Brockmeier on December 18, 2006 (8:00:00 AM)

Share    Print    Comments   

Lots of companies sell Linux servers, but how many promise 99.999% uptime? Not very many, but Stratus Technologies sells systems that offer the kind of fault tolerance that will handle mission-critical applications and leave admins with peace of mind. I had a chance to test out one of the company's ftServer 4300 systems, and it's an impressive system.

Since Stratus didn't let me keep the system around for a year (99.999% uptime means less than six minutes of downtime per year), I can't confirm the reliability figure that it promises, but if my limited testing is any indication, the ftServer is one good way to achieve that kind of uptime.

What you get

The server is actually two complete servers configured to work as one machine. The system consists of two 2U servers that are inserted into a 4U rack enclosure. Each system includes its own CPU, RAM, Ethernet interfaces, power supply, disk drives, and so forth. The enclosure provides the USB, serial ports, a modem, and VGA output for the system.

I received the system from Stratus with Red Hat Enterprise Linux Advanced Server (AS) 4 pre-installed, along with the fault-tolerant utilities. Stratus offers several configurations; I received the 4300, which included two 3.20GHz Intel Xeon CPUs, four 74GB 10K RPM SATA hard drives, two 160GB 7,200 RPM SATA drives, and two sets of two sticks of 1GB DDR2 RAM. Since everything is redundant, that breaks down to one virtual CPU, 2GB of RAM, two 74GB drives, and a 160GB drive.

Setting up the ftServer, physically, is easy -- you can have it unboxed and up and running within 10 minutes. After the units have been loaded and locked within the enclosure, powering up the master unit brings up both systems.

Once I powered the unit on, it displayed the Stratus startup screen and then the familiar Red Hat boot sequence. For the most part, you won't notice any difference between one of Stratus' ftServer systems and any other system running AS. The standard utilities are all the same, and there's no difference in the way you manage services such as Apache or Postfix on the ftServer compared to how you manage them on any other hardware.

Redundancy in action, Redundancy in action

The system performs well enough, but it seems a shame to have all that additional hardware "going to waste." Actually, it's not being wasted -- each system processes commands simultaneously, so that if a component goes out, the other one can step in and take over. With this setup, there's really no single point of failure with the system itself. It's extremely unlikely that the power supplies or CPUs or RAM would fail in both systems simultaneously.

Each 2U unit has two Ethernet interfaces. The interfaces are "bonded," meaning that the first interface on the first and second unit act as a bonded pair, and the second interface on the first and second unit also acts as a single Ethernet interface as far as the operating system is concerned. While I was running the system I tried pinging to and from the system and unplugging one interface at a time -- the system never missed a packet.

I also tried unplugging one of the units while the system was running. Again, aside from some error messages from the monitoring system, there was no evidence that anything was amiss at all. Users connecting to the system wouldn't notice a thing.

The modem I mentioned earlier is included so that the system can "phone home" if a component does go offline. Even though the system can continue to operate if a part fails in one system, that leaves the ftServer with a single point of failure, so it's imperative that a replacement part be installed as soon as possible. The folks at Stratus tell me that if the system does go offline, they'll be alerted and will then go through their contact list to connect with someone at the organization where the system is installed to notify them of the outage.

This redundancy does come at a cost. You are limited to Red Hat's enterprise offerings (or, ugh, Windows) on the ftServer. Stratus doesn't offer SUSE or other Linux distros as supported options, so if you're running a certified or homegrown Linux app, you'll need to be sure it's good on Red Hat. However, it shouldn't require any customization or tweaks to run on the ftServer.

The Stratus documentation I received with the system is an OK reference, but not a very good manual. I think the system would benefit from a shorter "getting started" guide that covers the utilities you use to monitor and control the ftStratus.

Not for everyone

I was impressed with the ftServer. It's a nifty piece of engineering, and should deliver the kind of uptime necessary for some applications. However, the system is overkill for many setups. In many cases, it would be more effective to have several separate systems in a high-availability configuration than a single ftServer.

The system I reviewed, as configured, clocks in at $30,755. According to the Stratus folks I spoke with, the entry-level systems start around $10K and larger dual-core systems can run more than $60K. They recommend Stratus systems for "applications where the consequences of downtime are costly, in terms of money or protection of life or property." If one of those situations apply to you, put Stratus on the shortlist of vendors to talk to.

Share    Print    Comments   

Comments

on Review: Stratus ftServer 4300

Note: Comments are owned by the poster. We are not responsible for their content.

Notably Missing...

Posted by: Anonymous Coward on December 18, 2006 09:38 PM
Notably missing from the article are minor details like it's NOT only available with Red Hat AS as the article suggested. It is also available with Windows and their own custom Linux which they call VOS.

Also notably missing is the source code. This system seems to be, from the article, a rather good fault tolerant clustering system built from GPLed Linux. Linux needs a good fault tolerant clustering system.

Where's the source???

#

Re:Notably Missing...

Posted by: Anonymous Coward on December 18, 2006 11:25 PM
If the fault-tolerant code doesn't run in the kernel, then there is no (legal) reason for it to be GPL'ed. So the question is, how did they write their system - as a kernel module, or in userspace? (I have no idea, never seen this product)

#

implementation

Posted by: Anonymous Coward on December 19, 2006 06:38 AM
I recall hearing that it's purely hardware failover from local hardware folks. We had a chance to look closer but it didn't work out in that project.<nobr> <wbr></nobr>:-)

--
Michael Shigorin

#

Re:Notably Missing...

Posted by: Anonymous Coward on December 19, 2006 02:13 PM
VOS is anything but Linux, it's a proprietary, pain in the ass stone-age mainframe-alike operating system.

#

Re:Notably Missing...

Posted by: Joe Barr on December 18, 2006 10:38 PM
You may have missed this in your first reading:


"This redundancy does come at a cost. You are limited to Red Hat's enterprise offerings (or, ugh, Windows) on the ftServer."

#

Re:Notably Missing...

Posted by: Anonymous Coward on December 18, 2006 10:49 PM
I did miss it. Thanks for keeping me honest!

#

Myth of the nines?

Posted by: Anonymous Coward on December 19, 2006 02:03 AM
<a href="http://en.wikipedia.org/wiki/Myth_of_the_nines" title="wikipedia.org">http://en.wikipedia.org/wiki/Myth_of_the_nines</a wikipedia.org>

#

Re:Myth of the nines?

Posted by: Anonymous Coward on December 19, 2006 10:51 PM
I wonder if Adam and Jamie from MythBusters will prove the myth?<nobr> <wbr></nobr>:)

#

This story has been archived. Comments can no longer be posted.


 
Tableless layout Validate XHTML 1.0 Strict Validate CSS Powered by Xaraya