Infrant ReadyNAS X6

Reviews: Networking

Infrant ReadyNAS X6

Author:
Editor:
Sponsor:
Published:

Richard Poelling
Brian
Infrant
Dec. 7, 2005

<< Previous

Next >>

Testing

X-RAID

To test the X6, I first began by inserting a single drive and then gradually expanding the array until I had all three hard drives functional. Even though I enabled the second and third drive simultaneously, it only recognized the second one and built the array as a mirror. Once that procedure had been completed, I rebooted the device and it automatically configured the third drive into the array thereby doubling the capacity. I did find myself looking through the menus after the second disk initialized since it did not initialize the third. Although the reboot wasn't an inconvenience, I would have liked to have been informed that I needed to perform a disk reboot and hadn't improperly connected something. This is just a reporting oversight, not an actual functional error. Since all three drives are up and running, this is how the X6 will be tested. This configuration is closest to a RAID 5 setting, but since the X-RAID system is proprietary and can not be changed, it is all just a matter of semantics anyway.

The total capacity of the final configuration is 692GB. I lost 37GB for snapshots and the rest is drive overhead. For the record I have a total drive capacity in the machine of 1.2 TB. This comes to a 42% drive overhead. The conventional wisdom on calculating the space a standard RAID 5 array should take up is (N-1)/N where N is the number of drives in the array. This means that if I used a standard RAID 5 configuration with no snapshots I should have gotten 800GB of storage, or 33% overhead. If I add another 400GB drive I should triple the original capacity to a little over 1 TB which would come to 37% overhead compared to standard RAID 5 which should come to 25%. In the end it is up to the user whether having the ability to expand the array at a later time supersedes the loss of 100GB of storage capacity.

Speed Tests

The test computer contains an ASUS A7N8X-Deluxe Motherboard with an AMD Athlon XP 3000+ CPU. This particular motherboard has dual onboard LAN ports, one of which is a Marvell 1 Gbit adapter. This will become important when we test throughput later on. To connect the X6 to the network I have it running through an SMC 8508T. This is an unmanaged switch, but it is one of the few which does support an MTU (Maximum Transmission Unit) of 9000, otherwise known as jumbo frames, and can be had for $100. This is also important since the X6 also contains support for jumbo frames.

To test the X6 I am using multiple programs which will put it through its paces. My test system is running Windows XP with Service Pack 2 and all security updates. The firmware revision of the X6 for testing is 2.00c1-p3. The tests are run three times and an average is taken. The tests are performed using the NVIDIA 100Mbit network connection or the Marvell 1Gbit connection. The maximum packet size is either set at 1514 (standard) or 9000 (Jumbo). For the testing I had the write caching enabled and full data journaling disabled. The data journaling can be enabled, but it is recommended that the X6 be protected by a UPS. This will ensure that if the device loses power data won't be lost since there is at least 24 MB of disk cache that could hold data on the 3 drives.

Jumbo Frames

Jumbo frames are not really out in the open just yet. Some companies support them (such as Infrant) while others do not. You will see jumbo frame support primarily on Gigabit Ethernet devices. The big question here is whether you should use jumbo frames. It has been my experience that jumbo frames work very well if everything supports their use. Obviously, having a network switch which supports frame sizes of 9000 is a start. To set up jumbo frames I usually had to go directly into the configuration properties of my network adaptor. I had to either set the maximum packet size=9000 or had to enable jumbo frame support. For systems which did not have jumbo frame support, there was a bit more lag in the initial connection, but after the systems settled on a mutual transfer speed, performance picked up. Also there was no noticeable difference in internet browsing for any Jumbo Frame enabled machine.

I will point out that my Windows XP test machine seemed to fair better than my Windows 2000 machine. The XP machine had a faster response when connecting to the X6. Because I was using separate machines with different hardware and NIC's from different manufacturers, I have no real way of knowing where the bottleneck truly resides. I could just have a screwball build on the Win2K machine!

<< Previous

Page 6 of 10

Next >>

Page 1: Introduction
Page 2: First Looks
Page 3: Installation
Page 4: Hard Drives
Page 5: Services
Page 6: Testing
Page 7: Testing: DiskBench
Page 8: Testing: Iometer
Page 9: Quality, Security & Error Reporting
Page 10: Conclusion

Subscribe to Storage [more info]

Add Comment

0 User Comments

Add Comment

To add a comment without being a member, you may omit the password field, but you must enter your name (or nickname) along with your comment. * Denotes required fields.