Alex Gerulaitis
Forum Replies Created
-
[David Roth Weiss] “Sorry Alex, but there’s absolutely nothing real-world about those numbers, and so I’m gonna have to ask for your source on this one.”
Didn’t say anything about the real world, only “published”. If you have anything real-world, would love to hear about it.
The TL;DR version:
As far as “real-world”: there is a number of studies on the subject including this one where published AFR, MTBF, URE rate numbers deviate from real-world ones – however:
1. Nobody I know compared reliability of desktop vs. enterprise drives in any scientific manner.
2. Doing a statistically sound comparison like that is truly a gargantuan task not easily accomplished.
This is why we have to go by published numbers – which support reliability figures I mentioned above.
As Ron and Bob said several times, there’s also the matter of what’s supported and warranted for a specific purpose. Few desktop drives if any are warranted for high density parity RAIDs with a 24/7 duty and if you’re using them for that purpose, the customer should sign a waiver.
-
[David Roth Weiss] “…this discussion should really be about, i.e. DATA SECURITY.”
The discussion was originally about the differences between desktop and enterprise drives, and various brands, wasn’t it?
Data security isn’t the only component of what matters to an editor: reliability and performance (i.e. uptime) are up there too. If your array is rebuilding three days in a year bringing performance down to unacceptable levels in addition to higher risks of data loss while the array is degraded – that costs money, too.
-
[Paul King] “Using desktop drives doesn’t make a difference to overall data integrity, just TLER susceptibility.”
Just differences in MTBF and URE numbers alone make enterprise drives two to 100 times more reliable than desktop ones. It’s another matter whether (a) you can believe published numbers, (b) it justifies the cost difference if you do.
-
Paul,
I don’t think anybody disputed that “enterprise” drives are desktops drives that passed the enterprise exam: better tolerances, less vibration, longer warranty, different firmware. They may all come from the same batch; those that pass certain tests, get the “enterprise” label.
Some enterprise drives though just don’t exist in desktop versions – SAS SFF ones: full duplex, longer I/O queues, and really expensive. (Love those, wish there were desktop versions.)
The real question is, does the enterprise label offer tangible benefits for the price? I think it does for some, and does not for others. I am not sure it’s all about greed though.
-
[David Gagne] “But we NEED TBolt drive enclosures to come down in price.”
Agreed; I am pessimistic about it though: enclosure, cooling, power, circuitry, etc. – you’re basically re-creating a small computer with that enclosure. The market is not big enough for them to come down in price, they’ll keep costing a dear penny. Storage boxes – more so, because each has to house a full-featured RAID controller, and you can’t expand it outside the box. What if I wanted a 48TB array with a 2GB/s throughput? Stripe two (or four) Promise boxes together over multiple TB ports? It’ll get expensive faster than it’ll get fast.
What Apple did (again, like with Mac Mini, iMac, MB Air)) was create an all-in-one box that you can’t mess with much, at the expense of expandability and upgradeability. You can’t add a SAS RAID controller or a Fibre Channel HBA to it. 10GbE? An extra box, lots of extra $$. More storage? More boxes, more $$. Each with their own fans and power supply – very efficient (not).
Sure, it’s the right thing to do in the face of the shrinking PC market, in general. Is it the right approach for a high-end editing system though?
-
How secure do you need this (erasure) to be? Expert hacker level or random uninitiated user level?
For the 1st, you’d want to fully rewrite the drives (with zeros or random data) – individually or in RAID0, 2nd – what you did is enough.
-
[Chris Burns] ” We’ve had some problems with the RAID ejecting itself while editing, so we’ve ordered a new one that should be arriving tomorrow.”
Chris, what are the specs on the array you ordered? Doesn’t make sense to recommend RAID level without knowing the details.
Also, the place you ordered that array from – have you tried asking them what they’d recommend?
-
[Bob Zelin] “Jumbo frame support”
Indeed, all that remains is to figure out how to enable / disable it… Unless it’s on full auto.
-
[Vadim Carter] “ZFS storage pools can span multiple vdevs (virtual devices)”
Vadim, I was talking about how the width of a stripe affects performance. Can’t improve performance without widening a stripe one way or the other, correct? I.e. even in RAIDZ, you’d have to re-stripe the array to improve performance. (At least that’s my understanding.)
-
[Bob Zelin] “I don’t remember if the 708 defaults to jumbo frames on, or if there is a toggle for in the the management.”
That’s the thing, no mention of jumbo frames, MTU sizes or related technologies (802.11) in User Guide.
I don’t have access to the switch, so maybe the setting is there, it’s just the User Guide doesn’t mention it.
BTW, seen a few write-ups out there saying jumbo frames improve performance only marginally if at all on 10GbE. What’s your take on it?