I meant exactly what was stated, you missed a bit of context.
I'll give the benefit of the doubt on your English since this forum is international... You mean host managed or host aware SMR, correct?
A great deal of experience.
I have a lot of that as well! That's why I was asking. I'm out of the business now, but I've built hundreds of RAIDs with thousands of drives for small businesses and have always been lead to believe that it was in my interest to avoid SMR drives for all of the reasons discussed in this thread by the manufacturer's PR literature, at least since the technology was introduced. For this reason, I've never used a SMR drive in a RAID. What exactly is your experience? Have the likes of WD released info on why SMR drives are now RAID appropriate?
Simple. Complexity. Enterprise RAID arrays, often have multiple tens(often hundreds) of drives. While redundancy is built-in, the complexity itself can sometimes be a liability all on it's own.
Statements like these lead me to believe you don't really know what you're talking about.
Generally we have groups or pools of drives in arrays. While there may be hundreds of HBAs or expander cards, a RAID is almost always controlled by a RAID controller or some software. Traditionally, per RAID array, even if there are hundreds of disks and dozens of LUNs/pools/mountable volumes, there is usually only one controller per RAID (I don't know much about virtualized and cloud instances, so forgive me here). I haven't noticed much of a difference across devices.
Consumer RAID arrays are rarely more than 4 or 5 drives and are, depending on configuration, often less prone to failure due to array desync.
More suspicious words! Anyone who cares about data knows RAID5 has become questionable since disks have exceeded 1TB, RAID6 sometime after, due to their rebuild times and the likelihood a disk would fail before being rebuilt, which again begs the question about SMR and random writes...
Even under heavy load. I have several of the drives that have a mixed mode of recording methods.
Again, I'll forgive the English, but it's suspicious since you're otherwise solid in translation!
Been running benchmarks and so far the slow down is minimal, less than 5%.
Please post! Years of anandtech reviews testing for this exact condition based on disclosure from manufacturers to said effect would like to have a discussion with your results
"Random" writes are so infrequent that they are not important to overall drive performance
On what OS? lol
and they are insignificant to RAID array performance.
What about the scenarios which I mentioned? What about during a rebuild or expansion? It feels awfully convenient that you dismissed all of my contentions by asserting your authority without producing any evidence to warrant said authority...
Mixed mode drives existing in a RAID array is not the problem. Whether or not those drives cause a problem is HIGHLY dependent on the array configuration.
I would really like to dismiss this as a non-english speaker translation issues, but it is pretty clear you are making things up based on cursory research... What is this "mixed mode" that you keep referring to? As I linked to above, and as is common language in all documentation and PR about SMR drives from manufacturers, the kind of thing you read if you buy lots of HDDs, especially if you work in the enterprise segment, there are three kinds of SMR drives: Device managed, host managed, and host aware.
Maybe when you say "mixed mode" you mean "host aware," but you must know, being in the biz and all, that those are not the kind of disks being discussed here...
I mean, maybe you mean "mixed mode" in the way that all SMR drives write in a CMR/PMR method until they begin to reach capacity, at which point they start to "shingle," and, in that way, are "mixed mode?"
The WD Red drives in question are device managed (according to WD). As I mentioned before, if they were host aware then this would not be an issue -- the drives themselves would simply disable SMR and no one would experience drops or slow-downs, because that his how device managed SMR drives are designed. Given that RAID controllers and almost every OS is expecting a PMR/CMR drive, you might understand why this is an issue.
According to WD's PR on the issue:
https://blog.westerndigital.com/wd-red-nas-drives/
WD said:
DMSMR is designed to manage intelligent data placement within the drive, rather than relying on the host, thus enabling a seamless integration for end users. The data intensity of typical small business/home NAS workloads is intermittent, leaving sufficient idle time for DMSMR drives to perform background data management tasks as needed and continue an optimal performance experience for users.
Yet, their own whitepaper:
https://documents.westerndigital.co...d-magnetic-recording-helioseal-technology.pdf
WD said:
Drive Managed SMR
...In short, Drive Managed SMR requires stronger caching algorithms and more random write space to temporarily hold non-sequential data.
Because the HDD constantly works to optimize the caching algorithm and indirection table handling, performance is unpredictable at certain workloads such as large block random write with high duty cycles. Due to the wide range of performance variability and unpredictability, Drive Managed SMR is considered impractical and unacceptable for enterprise-class deployments.
Drive Managed SMR is suitable for applications that have idle time for the drive to perform background tasks such as moving the data around. Examples of appropriate applications include client PC use and external backup HDDs in the client space. In the enterprise space, parallel operations of backup tasks become random write operations within the HDD, which typically result in unpredictable performance or significant performance degradation.
(You might also note that the whitepaper makes no mention of SMR's use in RAID)
