LSI MegaRAID 9271-8i PCIe Raid Controller Review

Day to Day Usage Testing

I received this card from LSI in early Febuary of 2013. Originally the plan was to do our normal short testing cycle on hardware. Something that for me tends to turn into a more involved process. However as life is, it turned into a extended period of time of testing. A series of months instead of weeks.

This series of time has spoken well for the the controller. When we spoke with LSI about obtaining a card for review part of goal was for stability. As most overclockers and general enthusiasts know. Hardware failure is common, loss of data across SSD’s seems to be even more common in some situations. Due to firmware changes and bugs. Its not uncommon to loose functionality of an SSD. Even more common is in the pursuit of performance to loose all the data on a SSD, due firmware updates that suppose to be none destructive. Under Most situations loss of an drive can be a pain. Loss of a SSD is generally worse, as any hope of data recovery is pretty much none existent.

The easy way around data loss this is to backup on a regular basis. However even routine backups are not a great solution. Unless you work in an environment that requires constant backups, most are done quite hit and miss. Same with the frequency that most people do windows updates. Even if you do adhere to a daily backup, there is still a chance of loosing enough work that its a pain to retrace your steps and recover from a data disaster.

Here is where a RAID controllers typically come into their own. Despite that performance increase gained in some drive setups. The ability to actively mirror an setup parity to other drive to enable the loss of a drive or multiple drives is definitely an advantage. Some RAID types also present a strong performance boost in conjunction with this level of redundancy  Such setups can typically loose a drive while still allowing access to the data on the drive in question. At the same time as rebuilding the lost drive to a spare.

The Setup

For everyday usage testing the setup is very similar to our initial Sandy Bridge test setup. With a few changes.  This is my everyday setup, so I have some large bulk storage drives thrown into the mix. There is also an additional SSD just for games. What games it has on it tends to vary depending on what I am playing at the time. It tends to mostly be reserved for MMO’s. My machine stays on 24/7. I don’t use speed step, and it is generally always overclocked. Sometimes further than others.

As you can tell this is a fairly high strung setup. Running stable generally with a reboot every 30 days or so. Its not uncommon under an overclock of this level to experience periodic Drive corruption when making performance tweaks for testing. Which is another obstacle a good raid controller can help cope with.

You can also notice that at this point I have switched from a RAID 0 setup on the main array, to a RAID 6 array. Allowing me the loss of two drives at once in the array. This also provides slightly better performance than RAID 5. Not to mention less overall read/write overhead for the controller compared to a RAID 0 setup, which starts experiencing diminishing returns with the more drives added, and zero fault tolerance.

Every day usage

As more enthusiast do, I throw a large mix of tasks as my machine. A large amount of web browsing, watching TV shows and movies, music, and gaming. This makes up a fairly consistent level of inter-mingled applications that tend to be in use at just about any time. As most games at this point in time don not really use four cores its fairly safe to say that my machine spends most of its time at about 35-50% CPU usage.

The things that really are important and generally intermixed into the usage are Video Editing, Multi-Track Sound editing, and Photoshop. These three applications by themselves can all bring a machine to max load. Not to mention a mix of time and computation needed to complete tasks in some cases can take days. This makes a active backup and data integrity that much more important. It also makes disk performance an important factor.

At this point This where generally humming along nicely for a few months. The drive performance was nothing short of exceptional. The state of mind knowing that if I lost a couple drives my data would still be safe. Coupled with a daily backup of the essential data across my local network at night. In fact I was having a hard time coming up with the requirement for this much storage speed at times. Video editing certainly was not limited by drive performance at this point. Swap performance for Multi-track sound editing was beyond exceptional. As loading 24 tracks and then some of studio recordings became a breeze. The thought of overkill was definitely on my mind.

The Failure

Early April I started having stability problems. At first I wanted to finger a faulty display driver. Being one who generally ends up running beta display drivers for performance increases in games. I started more closely monitoring everything in my system. As a number of changes had happened at once along with rising local temperatures.

I resorted to strapping thermal probes to number of devices along with short duration logging and visual logging of temperatures during loads. It wasn’t quite consistent enough at this point to finger anything. As anything short of 30 days of stability is considered a problem in my book. Most would not bat an eye lash at a failure after 5 days of excessive CPU load.

Then it happened a hard crash and an alarm that put a house hold smoke alarm to shame. Painful would be the correct term for it really. As I was about three feet away at the time of it going off. Having messed around with drive failure during the testing process I already had a good idea what was going on.

After the hard crash one of my five drives in the array had become corrupted. (I had also at that point in time had just tracked down the source of the stability problem. My trusty CPU of around the last few years was getting rather unhappy with being pushed quite so hard and decided it wasn’t happy at 5ghz anymore.) After  the alarms where silenced much to the appreciation of my house mates, (and my neighbors). The rebuild was started, since these drives are so small and fast, it took a relatively short amount of time. Once the rebuild was complete. I changed some bios settings, having intentionally forced the bios into a default safe level of settings before turning the machine back on in the first place.

Since then everything has been running like clock work. If this had been a single drive, a stripe or a mirrored setup even. The potential for lost time along with the requirements to restore from a previous days backup would most likely of been a requirement. Followed by a large amount of cursing and trying to remember exactly everything I had done project wise that day.

Day to Day Testing Conclusion

The fact that I lost maybe 10-20 minutes of time in this situation instead of 8-12 hours of time, proves a point. The controller is stable and its meant to be. A high end controller should be the last concern when it comes to system stability. It should also be a priority in a system build that requires such stability. The quick and easy recovery process is also another great advantage. Not to mention the software tools to monitor your controllers health and array health, can help alert you to problems that might be controller or array associated. Allowing problems to be caught before a disaster  You can even automate the recovery process by having a hot spare in the system, in case of failure.