If your business is enhanced storage, your margin in storage infra is directly relevant to your COGS. I got to learn this the hard way, having once worked in the field. (SpiderOak)
We like to match drive types in chassis and sometimes cabinets - that way we don't have 8 different kids of spares to juggle. Right now we just happen to have a lot of 4TB arrays in use.
But yes, we will deploy arrays with 6TB very soon.
Once you go past the 45 drive JBOD (which costs $2000 give or take) the price per slot starts to go up quickly ... the last quote we had for the supermicro 90 drive JBOD was $9000.
So if you are optimizing for rackspace, it might make sense, but if you're keeping costs down, the 45 drive chassis is the best bet.
Yes, from the very beginning, but they are based on SATA expanders and have relatively very poor performance. They are really built specifically for that workload and not for a lot of other storage workloads.
That was depressing circa the first BB pods where there really wasn't another option (other than using a Sun Thumper/x4500, which we also considered) ... but now the chassis options from supermicro are so good - ranging from 45-90 drives - that it's not a problem at all.
The Dell DSS7000 chassis without the computer nodes on the back (which are a requirement (45 drives per compute node)) is cheaper than the 90 disk JBOD from SuperMicro. I can't disclose pricing, but it may be worth reaching out to Dell for what they can offer there.
You might be able to get better density for similar pricing. Just a suggestion :-)
That's a refreshing attitude. I wish AWS had something like that warning... "If you only want a VPS to idle in IRC with... stop reading, this is not for you."
However, it would be nice to know a little about the company. I perused your website, but didn't see any pages about the company history and current employees. Did I miss that? I did finally see the "CEO page", but it doesn't seem to be linked from the main page. It also appears to be something a technical person shows their CEO, rather than something that is about the CEO of rsync.net. I also didn't even see a physical address displayed anywhere.
Maybe it's just me, but I wouldn't feel comfortable dealing with such a black box business situation.
Perhaps you need to add a second disclaimer?
If you want to know who you're doing business with,
our product is Not for You.
We just redesigned our website and I guess our old "about" page didn't make it in the redeploy.
FWIW, I'm John Kozubik, owner and founder, and we started offering this service in 2001 as an add-on to JohnCompanies colocation (the first VPS provider)[1][2]. We spun rsync.net out as a standalone entity in 2006.
Address is 524 San Anselmo Ave., Suite 107, San Anselmo, CA 94960. Although data is housed in either San Diego, Fremont, Denver, Zurich or Hong Kong datacenters.
[1] Yes, really.
[2] We called them "server instances" - someone else coined "VPS" and it stuck.
Does anybody take the encryption support in attic/borg seriously? They seem to be implementing their own AE construction. Maybe they did it right, but it makes be a bit uneasy. While they're at it wrapping OpenSSL, I don't know why they wouldn't use one of OpenSSL's AEAD EVPs...
Which of Attic or Borg would you recommend for new usage? Deduplicated always-incremental backups have appealed to me ever since I started following the development of obnam, and I'd love to switch away from duplicity and stop requiring periodic full backups.
Yes, that's the reason everyone is eschewing duplicity (which we've always loved and supported) for attic and borg.
borg appears to be much healthier as a project, has a lot of steam behind it, and has some additional flexibility in block sizes, etc., so we think that's the train to hitch your wagon to ...
It will probably stop working. Aerodynamics are crucial to the way a hard disk works. One cannot not simply increase the density of the gas inside the drive by 10x and expect it to continue working.
The helium reduces resistance which results in lower vibration of the read/write heads, which in turn allows for reduced space between the platters and the heads, which allows for a reduced space between the platters and more platters per drive.
If the same drive was filled with air rather than helium, the increased vibration of the heads would cause them to crash into the platters - which would scrub the disk clean (which would be bad).
In a vacuum, the head would crash into the platter. The gas provides a cushion. Flying height these days is something like 3nm (yes, you read that right), so the tiniest vibration could cause a head crash in a vacuum.
> Flying height these days is something like 3nm (yes, you read that right), so the tiniest vibration could cause a head crash in a vacuum.
Wikipedia goes on to say the head could crash into a fingerprint or _smoke particle! So... how did my spinning rust laptop drive ever survive on my lap?
How well have the first generations of helium drives lasted?
I'm wary of trusting my data to something that relies on incredibly good gas seals. But if there are no observed reliability issues, it seems that problem has been solved.
You mean their almost certainly doomed by flash drive 10K and 15K drives? (A class of drive I've been using since 1995 and just stopped buying new ones; in fact, a 15K Savio mirrored system disk recently failed and I replaced both with an Intel enterprise SSD rather than send it in for warranty repair. So far I'm not looking back.)
Officially, Constellation drives are "Enterprise", 5 year warranty (which still works well, did that on a 2TB drive last year I think), available in SAS as well as SATA interfaces. Although they're now saying that you shouldn't transfer more than 550TB of data per year through the best of this class of drives.
In a 8 disk nas i replaced 6 oft them in one year because of bad smart data. Some lastet only Werks before marked as faulty.
This series is really bad.
Place where I used to work had large NetApp array's using HGST's Helium filled drives.
During the time that I was there (1 year, these had been in service longer than that), out of the 250+ drives in service we didn't have a single failure.
None ever failed me (several hundreds in production). In general, HGST drives are much more reliable than Seagate and WD anyway, but the He are supposedly even better.
The helium is not under pressure, it should not leak much. A metal to metal seal should do a pretty good job of sealing it, especially with no pressure.
"Guaranteed" is too strong a word, but I'm sure that engineering a seal that would hold reliably for years and could be mass-produced was a nontrivial problem. This is why everyone isn't already using helium.
The great thing is that it doesn't use SMR. I have an 8TB SMR from Seagate and its write performance is bad and variable, much like the pre-TRIM SSDs. It's ok if you write a lot of stuff and then just read it (you can get >100MB/s out of it), but I often see small random writes (updates) at 3-4 IOPS, because it has to re-shingle the whole track.
The USG had/has a store the National Helium Reserve [1]. It was being sold off so I am unsure how much is still in it, but to answer your question you store it underground.
Edit. It looks that the National Helium reserve will be sold off by 2018 [2].
I wonder if these will be recoverable by traditional data recovery means, or if opening them and exposing the platters to normal air instead of helium would mean they can't spin/run correctly.
If helium-filled HDD technology becomes more widespread I am curious about the potential impact to the data recovery market.
Which would add a significant up-front cost that would have to be shouldered by the data recovery company. Potentially an issue for small shops (which most data recovery companies are.)
Get one of those vacuum sealing luggage bags, cut holes in side and glue in some rubber gloves into these holes. Put drive and tools in. Vacuum seal it, then Fill with helium you get from the party store.
Probably not but neither are SSD's, judging by how many data recovery places have went out of business I don't think it's a major consideration any more.
If it dies in an enterprise environment it would be part of a redundant array of some sort (I can't think of a single use case for such a large and expensive drive in a single drive setup), gone are the days when the 30GB HDD in the CEO's tower died and you had to rush it to some 48 hours recovery service to get that "super important" presentation they had on it (AKA his vacation pictures).
I'm pretty sure that like 90% of the current data recovery cases are from people that still carry mechanical portable drives or haven't switched to a NAS with RAID support which will never be the market segment for those drives anyhow.
I think these are gonna be too little too late for the platter drive market, especially with SSD capacities skyrocketing recently (thank you 3D NAND). Toshiba and Samsung both have announced plans for 128TB SSDs within the next 3 years - I can't really see platter drives competing with that.
66 comments
[ 3.9 ms ] story [ 166 ms ] threadExciting because we love the idea of deploying a 450TB zpool in a 4U chassis ...
Frustrating because we will have to keep reworking our price per GB.
With that in mind, our new attic/borg[1] support has dropped down to 3 cents per GB, per month to kind-of-sort-of match S3 pricing[2].
[1] http://www.stavros.io/posts/holy-grail-backups/
[2] http://www.rsync.net/products/attic.html
</sarcasm>
Storage is cheap when you own it vs renting it: https://i.imgur.com/WZriUP5.jpg
But yes, we will deploy arrays with 6TB very soon.
Dell also has their DSS7000, which is 90 drives in a 4U system, so 900 TB in a single 4U package...
9 Petabyte of storage in a single 42U rack...
So if you are optimizing for rackspace, it might make sense, but if you're keeping costs down, the 45 drive chassis is the best bet.
That was depressing circa the first BB pods where there really wasn't another option (other than using a Sun Thumper/x4500, which we also considered) ... but now the chassis options from supermicro are so good - ranging from 45-90 drives - that it's not a problem at all.
You might be able to get better density for similar pricing. Just a suggestion :-)
Hey, thanks! I stopped reading there. :)
However, it would be nice to know a little about the company. I perused your website, but didn't see any pages about the company history and current employees. Did I miss that? I did finally see the "CEO page", but it doesn't seem to be linked from the main page. It also appears to be something a technical person shows their CEO, rather than something that is about the CEO of rsync.net. I also didn't even see a physical address displayed anywhere.
Maybe it's just me, but I wouldn't feel comfortable dealing with such a black box business situation.
Perhaps you need to add a second disclaimer?
FWIW, I'm John Kozubik, owner and founder, and we started offering this service in 2001 as an add-on to JohnCompanies colocation (the first VPS provider)[1][2]. We spun rsync.net out as a standalone entity in 2006.
Address is 524 San Anselmo Ave., Suite 107, San Anselmo, CA 94960. Although data is housed in either San Diego, Fremont, Denver, Zurich or Hong Kong datacenters.
[1] Yes, really.
[2] We called them "server instances" - someone else coined "VPS" and it stuck.
You don't have to pass every savings on to the consumer, you know.
borg appears to be much healthier as a project, has a lot of steam behind it, and has some additional flexibility in block sizes, etc., so we think that's the train to hitch your wagon to ...
"... but HGST's use of helium, which offers significantly less air resistance to the spinning platters, is probably the main one."
But then you have to have a case strong enough not to bend under the pressure difference.
At 1 atm, of course, helium has much lower density (and therefore lower viscosity). Nitrogen weighs 28 g/mol; helium just 4 g/mol.
Helium is a nice, nonreactive noble gas.
Wikipedia goes on to say the head could crash into a fingerprint or _smoke particle! So... how did my spinning rust laptop drive ever survive on my lap?
I'm wary of trusting my data to something that relies on incredibly good gas seals. But if there are no observed reliability issues, it seems that problem has been solved.
Officially, Constellation drives are "Enterprise", 5 year warranty (which still works well, did that on a 2TB drive last year I think), available in SAS as well as SATA interfaces. Although they're now saying that you shouldn't transfer more than 550TB of data per year through the best of this class of drives.
During the time that I was there (1 year, these had been in service longer than that), out of the 250+ drives in service we didn't have a single failure.
Edit. It looks that the National Helium reserve will be sold off by 2018 [2].
1. https://en.wikipedia.org/wiki/National_Helium_Reserve
2. https://en.wikipedia.org/wiki/Helium_production_in_the_Unite...
If helium-filled HDD technology becomes more widespread I am curious about the potential impact to the data recovery market.
Wear gloves, pickup tools and dismantle drive.
And pressurize it with helium or a noble gas.
If it dies in an enterprise environment it would be part of a redundant array of some sort (I can't think of a single use case for such a large and expensive drive in a single drive setup), gone are the days when the 30GB HDD in the CEO's tower died and you had to rush it to some 48 hours recovery service to get that "super important" presentation they had on it (AKA his vacation pictures).
I'm pretty sure that like 90% of the current data recovery cases are from people that still carry mechanical portable drives or haven't switched to a NAS with RAID support which will never be the market segment for those drives anyhow.