NetApp filers?
Matt McLeod
people who are actually using the things would be cool.
Basically, we're looking at buying an E450 with a an
imperial arseload of disk (somewhere in the area of 100G
all-up, which is a lot for a unit of 70 people when they've
already got ~50G), or a D1000 with a similar amount of disk
to hang off our overloaded E4k.
As it happens, Filers are on the approved-list for hardware
at $VBC, so we're a bit curious. The main things we'd be
doing with it is user home-directories, application repository,
and ClearCase storage pools.
So -- are they as good as they sound? Are they reasonable
value for money?
--
"Don't put off 'till tomorrow, responsibilities.
They'll just come back to haunt you.
(Ignore them totally)"
-- TISM
Calle Dybedahl
> I've read the marketting guff, but some feedback from
> people who are actually using the things would be cool.
Well, I've worked with them before and at the moment I'm responsible
for Telefnordia's NetApps...
> As it happens, Filers are on the approved-list for hardware
> at $VBC, so we're a bit curious.
Ericsson currently has the world's biggest NetApp installation, here
in Stockholm. A clustered monster with a grand total of nine terabytes
of disk on it[1]. Original plan was to have all Swedish employees'
home directories on it. That has since been upgraded to "Heck, let's
put *everything* on it", after it has had time to prove both stable
and fast. It also didn't exactly hurt that the snapshot capabilities
made it possible to clean up a bad virus incident in about an hour
rather than the four days it took the people running on more ordinary
diskservers.
> So -- are they as good as they sound? Are they reasonable
> value for money?
Yes and yes. NetApps are the closest thing to fire-and-forget hardware
I know of. For typical usage (compiling, editing, that sort of thing)
they're often faster than local disk (given good enough network, you
may want one or more dedicated 100Mbit/s pipe(s) from the NetApp to
the CPU server), they're trivial to admin and extremely stable[2].
[1] Last time I heard, anyway. They upgrade it more or less
constantly. And they "only" distribute about 4TB to the users, the
rest vanishes as clustering overhead[3] (4TB) and temporary storage
for the backup server (1TB).
[2] As long as you only use NFS. If you use CIFS, it'll crash once a
month or so. I spoke to a NetApp technician about that, and got a
reply along the lines of "We know. We bought the CIFS code from
Microsoft to get maximum compatibility, and the terms of the
contract prevents us from fixing it."
[3] You don't normally get any overhead from clustering, but Ericsson
decided that this server is important enough that they have two
entire machines in different buildings mirroring each other. The
original plan was to have the machines on opposite sides of
Stockholm, but they changed their minds. So you can knock the
entire cluster out by nuking Kista.
--
Calle Dybedahl, Vasav. 82, S-177 52 Jaerfaella,SWEDEN | ca...@lysator.liu.se
"Our five main weapons are Invincible Ignorance, Not Invented Here,
FUD, derision, wild-eyed ranting, ad hominem attacks, straw men, and,
and...oh bugger." -- Joe Bednorz, A.S.R
Matt McLeod
that Calle Dybedahl did write:
>> So -- are they as good as they sound? Are they reasonable
>> value for money?
>
>Yes and yes. NetApps are the closest thing to fire-and-forget hardware
>I know of. For typical usage (compiling, editing, that sort of thing)
>they're often faster than local disk (given good enough network, you
>may want one or more dedicated 100Mbit/s pipe(s) from the NetApp to
>the CPU server), they're trivial to admin and extremely stable[2].
Cool. We're going to have one hell of a job selling the right
people on making any changes to the current setup (or at least
anything which will cost money), so might as well try to sell
them on the Right Thing.
From memory, there are three options on the "approved" list
for servers -- Sun kit, NT boxes, and NetApp stuff. We know
Sun kit, and hate NT, so naturally NetApp appeals.
--
More than a mere hindrance. It's a whole new barrier!
Stig Sandbeck Mathisen
>As a device they work quite well, but I wouldn't describe them
>as value for money given the amount of markup NetApp put onto
>their disk and disk shelves.
Well, you don't *have* to use disk and shelves from NetApp. You
can use "unoriginal" ones as well. Far cheaper, but without the
support, so you have to think about that as well.
--
SSM - Stig Sandbeck Mathisen
Trust the Computer, the Computer is your Friend
Rune Bakken
> They used to use Storageworks shelves, for their SCSI stuff. The fiberchannel
> shelves are something I haven't seen before.
I am not sure who makes the FCAL shelves, but they are apparently "Off
the Shelf". My Boss[1] spent a few hours one afternoon - and digged
out how where and how much he could save on non-original[2] shelves.
Somewhere in the range of 2/3 savings from NetApp prices.
>>>>>>>>>>>>>>>>>>.rune
1/ A magician when it comes to finding alternatice sources
2/ ie not been trough NetApp QA and without their warranty[3]
3/ The manufacturers warranty is better that NetApps ...
--
Rune Bakken ru...@bakken.no
Maur Hansens v 34, N-2013 Skjetten, Norway
Ph: (+47) 63 84 16 95
void
>
>[1] NFS locking *really* feels like Sun designed this stateless
> file sharing protocol and at the last minute someone said
> "What happens if we want to lock files ?" - the result was
> rpc.lockd.
Er, I think that actually is the case. Anyone know for sure?
--
Ben
"Cleverness is serviceable for everything, sufficient for nothing."
-- A fortune cookie, lunch, 25 December 1999
Scott Hazen Mueller
>http://teaparty.mathworks.com:1999/toasters/
>toas...@mathworks.com (standard majordomo list)
Heh. 'Toaster' was the nickname we applied to NetApps when we (Tandem) beta-
tested them some years back. Glad to see it stuck somewhere.
\scott
Scott Hazen Mueller
>Somewhere in the range of 2/3 savings from NetApp prices.
What's funny is that the NetApp folks originally intended to be a software
company. They were going to build competitors to Auspexen that ran on stock
PC hardware, and something happened along the way to that goal...they found
that they needed control over the hardware environment in order to guarantee
reliability, so they had to build their own.
Which just reinforces the point that what hardware integrators do is add
reliability.
\scott
Thorfinn
Scott Hazen Mueller <sc...@zorch.SF-Bay.ORG> wrote:
[ a followup without References: ]
Oi. Fix your references header.
ook,
Thorf
--
<a href = "http://netizen.com.au/~thorfinn/">thor...@netizen.com.au</a>
When someone says "I want a programming language in which I need only
say what I wish done," give him a lollipop.
-- BSD fortune file
Roger Burton West
Matt McLeod <matt+...@netizen.com.au> wrote:
>As it happens, Filers are on the approved-list for hardware
>at $VBC, so we're a bit curious. The main things we'd be
>doing with it is user home-directories, application repository,
>and ClearCase storage pools.
>
>So -- are they as good as they sound? Are they reasonable
>value for money?
IME: mostly pretty good, fairly reliable, but with a disturbing
tendency suddenly to go completely tits-up for no discernable
reason and refuse to talk to anything ever again, after about a
year of use when you're finally starting to think about trusting
the things.
Mind you, the ones I'm thinking of were being used for news spools.
Roger the ex-bob
Nick Manka
Calle Dybedahl <ca...@lysator.liu.se> writes:
> [2] As long as you only use NFS. If you use CIFS, it'll crash once a
> month or so. I spoke to a NetApp technician about that, and got a
> reply along the lines of "We know. We bought the CIFS code from
> Microsoft to get maximum compatibility, and the terms of the
> contract prevents us from fixing it."
That's interesting, our only CIFS-related crash has to do with an
overzealous sanity checker between WAFL and CIFS. It suddenly makes
more sense.
Tony Finch
>
>Heh. 'Toaster' was the nickname we applied to NetApps when we (Tandem) beta-
>tested them some years back. Glad to see it stuck somewhere.
NetApp call NetApps toasters.
Tony.
--
I'm the dot in dot at
Matt McLeod
>In article <86g4ll$t6n$1...@triad.bpfh.net>, Simon Burr <si...@bpfh.net> wrote:
>>Another problem I have with NetApps is that unless you put a
>>tape drive onto them, its kinda hard to back the damn things
>>up - you are forced into doing the backup over NFS which kinda
>>sucks.
>
>You can use their "dump" program via "rsh".
Hm. Would it be safe to assume that they don't really
play well with Amanda?
--
"Perhaps he thought I was just
servicing the master?"
-- Damien, Prince of Darkness, Eater of Souls
Russ Allbery
> Kendall Libby <fu...@pir.net> writes:
>> What's the work on EDM ? We might be getting and EMC and that's
>> obviously their recommended choice.. I haven't yet heard anything good
>> about it.
> Well, I've had no direct experience with the thing. However the murmers
> I've been getting from Ops is that its a rather expensive white elephant
> which has very sucky support.
I'm not parsing the original completely, but if you're asking about
experience with EMC, ours almost couldn't have been better. Apart from
one controller board failure we ran into, they just work. Solid as a
rock, performance is amazing, people show up on your doorstep to change
disks, their support folks have managed to fix some fairly hairy problems
that we ran into when we were first getting started with them, and tons of
storage management hassles that we'd been dealing with before just
disappeared.
Of course, you pay through the nose (and several other orifices) for that,
but if you've got more money than people....
--
Russ Allbery (r...@stanford.edu) <URL:http://www.eyrie.org/~eagle/>
David Iain Greig
>In article <p3u2k4l...@kavringen.nextel.no>,>The manufacturer provides a support contract that guarantees 4 hour response
>to failures?
Onsite?
--D.
David Iain Greig
>>Another problem I have with NetApps is that unless you put a
>>tape drive onto them, its kinda hard to back the damn things
>>up - you are forced into doing the backup over NFS which kinda
>>sucks.
>
>You can use their "dump" program via "rsh".
Dumping the better part of a terabyte over LAN is an exercise
I'd only do over ATM or gigabit ethernet. Might as well buy a tape.
--D.
Rune Bakken
> Now, if it support ssh..... :)
They do, although they wont export, it contains cryptography ... [1]
>>>>>>>>>>>>>>.rune
1/ We obtained license to IPSec software from Cisco, but NetApp cant
find a way to legally export their software [2]
2/ Why would USofA care about software that can only be used to talk
to an appliance?
Rune Bakken
> In article <p3u2k4l...@kavringen.nextel.no>,
> Rune Bakken <ru...@bakken.no> wrote:
> >3/ The manufacturers warranty is better that NetApps ...
>
> The manufacturer provides a support contract that guarantees 4 hour response
> to failures?
No, I keep a a few spare disks on top of the NetApp, and several more
in my office. Then when one fails, I ship the disk for a warranty
replacement.
Although I do like the response from my NetApp supplier here in Oslo,
aproximately 2 minutes after the disk failed, they called and asked if
this was an original NetApp disk, and if so, should they send it to
the office.
But alas, I've heard that response varies with the location. One of
the netapp I administers is in Italy, and when the motherboard failed,
I was forced to call NetApp Europe to get a response from the Italian
supplier.
>>>>>>>>>>>>>>>:rune
Simon Cozens
>Eh? Palm VT100 exists, works, and is fun. :)
>
>I think there might be a couple of other terminal emulator programs
>apart from that, too.
There is ssh. I kid you not.
Eek. A sigmonster.
--
Never go to bed mad. Stay up and fight.
-- Phyllis Diller, "Phyllis Diller's Housekeeping Hints"
Russ Allbery
> There is ssh. I kid you not.
Person I used to work with (who may well be reading this) hooked a
Ricochet modem up to his Palm (larger than the Palm itself) and has I
believe done AFS volume releases remotely that way.
David Iain Greig
>
>We're only dumping 100 gigabytes and it's already bigger than one tape will
>hold... but we do have the 100 megabit dedicated network I previously noted.
For 100+ GB we'll direct attach a DLT7000 and put it in the Powderhorn.
Then the robot can change tapes. Make us mad and we'll put you on
the Brocade and use 9840 drives.
--D.
Jeff Wolfe
9840=Eagle, right? You've had problems? (Our powerhorn is about to
land on the loading dock with a bunch of 9840 drives..)
-Jeff
--
Jeff Wolfe College of Earth and Mineral Science - Penn State
David Iain Greig
Tape drives are fine. Brocade == Mildly Evil.
--D.
Russ Allbery
> Russ Allbery <r...@stanford.edu> probably said:
>> ... performance is amazing, ...
> ... until your data set doesn't fit in their cache anymore, then it goes
> through the floor, IIRC.
Well, I've yet to see it happen, and I'm running transit news service
against one at the same time as it's serving out about 15,000 user home
directories....
Russ Allbery
>> Well, I've yet to see it happen, and I'm running transit news service
>> against one at the same time as it's serving out about 15,000 user home
>> directories....
> You've got a suitable application.
> Now, a large database on the other hand ...
Point. Although folks here are doing that too... although come to think
of it, they're having performance problems. Hm.
The transit news feed does churn the write-back cache pretty bad, though,
which is why I mention it. But not bad enough to cause any problems that
I've seen.
Jeff Wolfe
>>9840=Eagle, right? You've had problems? (Our powerhorn is about to
>>land on the loading dock with a bunch of 9840 drives..)
>
>Tape drives are fine. Brocade == Mildly Evil.
Phew.. Anybody know of an ACSLS shim for AMANDA? We have a couple
drives set aside for our AMANDA server, lest we let ADSM and Cray DMF[1]
have it all.. :)
I ordered the ACSLS developers kit, so I suppose I'll have to roll my
own if nothing already exists.
-Jeff
[1] The One True HSM, as far as I'm concerned. It's too bad SGI sucked
what life was left out of Cray and discarded them. I really thought
the T3E was an interesting box.. We might have been looking at one of
those now instead of an SP.
Strange
>> There is ssh. I kid you not.
> Person I used to work with (who may well be reading this) hooked a
> Ricochet modem up to his Palm (larger than the Palm itself) and has I
> believe done AFS volume releases remotely that way.
YES. This is my more-or-less fantasy portable connectivity setup.
Ricochet strapped to a 3x.
But I live in Chicago. Ergo it's still a fantasy portable connectivity
setup.
SSH for pilot port by Ian Goldberg, I believe....
-M
--
Michael Brian Scher (MS683/MS3213)| Anthropologist, Attorney, Policy Analyst
Mainlining Internet Connectivity for Fun and Profit
str...@netural.com li...@foad.org str...@cultural.com str...@ispfh.org
Give me a compiler and a box to run it, and I can move the mail.
Calle Dybedahl
> It will? I suspect that depends on how hard you're pushing it or something.
> We've had a few crashes but they've been on top of them.
I guess it may have got better, it's been a while since I used it. NFS
plus re-export via Samba worked better anyway (for our needs, anyway).
--
Calle Dybedahl, Vasav. 82, S-177 52 Jaerfaella,SWEDEN | ca...@lysator.liu.se
"Such a pretty day for a bloodbath." -- Callisto, "Xena: Warrior Princess"
Bryan C. Andregg
>Although I do like the response from my NetApp supplier here in Oslo,
>aproximately 2 minutes after the disk failed, they called and asked if
>this was an original NetApp disk, and if so, should they send it to
>the office.
Our Sunnyvale office (né Cygnus) has told us that they have the same kind of
service, excpeting that the call from NetApp support includes the tracking
number for the next business morning FedEx package with the drive in it.
--
Bryan C. Andregg * <band...@redhat.com> * Red Hat, Inc.
1024/625FA2C5 F5 F3 DC 2E 8E AF 26 B0 2C 31 78 C2 6C FB 02 77
1024/0x46E7A8A2 46EB 61B1 71BD 2960 723C 38B6 21E4 23CC 46E7 A8A2
nic...@esperi.org
Rune Bakken <ru...@bakken.no> wrote:
>Simon Burr <si...@bpfh.net> writes:
>
>> Now, if it support ssh..... :)
>
>They do, although they wont export, it contains cryptography ... [1]
>
>>>>>>>>>>>>>>>.rune
>1/ We obtained license to IPSec software from Cisco, but NetApp cant
> find a way to legally export their software [2]
Thus making the whole thing about as useful as a chocolate teapot
for Europe-wide [1] use of their product manufactured in the USA.
Bloody export laws, but also bloody vendors that don't manage to
sell their product everywhere.
Nicolai
[1] Including Russia, where I believe the current status quo is that
cryptography is banned. Joy and delight.
nic...@esperi.org
>>In article <86g4ll$t6n$1...@triad.bpfh.net>, Simon Burr <si...@bpfh.net> wrote:
>>>Another problem I have with NetApps is that unless you put a
>>>tape drive onto them, its kinda hard to back the damn things
>>>up - you are forced into doing the backup over NFS which kinda
>>>sucks.
>>
>>You can use their "dump" program via "rsh".
I suspect this might not play well with the EDM, which wants to
install and run the provided proprietary client software on the
client machine, then talk to it via the proprietary protocol the
EDM uses. Attaching a tape drive and manually swapping tapes would
not go down well with us, since we like the idea of a robot changing
tapes. It's more intelligent than some of our operators :-(
>Dumping the better part of a terabyte over LAN is an exercise
>I'd only do over ATM or gigabit ethernet. Might as well buy a tape.
It's quite feasible given sufficient networking; there was a talk at
LISA 98 (I think) about backing up approximately 5TB across Gigabit
Ethernet, although the single largest data chunk was 1.5TB.
Summary: It works, and goes a lot faster if you use Jumbo Frames
and stream to several tape drives simultaneously.
You do need a tape robot and several drives for this, obviously.
Netapps, on casual inspection, appear not to do multitapes very
well. I may be wrong, in which case a brief reference would be
appreciated [1]. Either way, having bought the EDM, we have to use
it.
Nicolai
[1] Ie, don't toast me for not worshipping your favourite piece
of computing equipment.
Rune Bakken
> Our Sunnyvale office (né Cygnus) has told us that they have the same kind of
> service, excpeting that the call from NetApp support includes the tracking
> number for the next business morning FedEx package with the drive in it.
You mean they trust FedEx???
In my case, the salesdroid delivered it personally[1].
>>>>>>>>>>.rune
1/ No they are not co-located with us.
Zaphod Beeblebrox
>Nothing sucks more than waiting 26 hours to get a new head..
Amen, brother.
Z.
Jonathan H N Chin
>[elided]
IHNJ. I am just astounded that Goh didn't whinge about the headers.
On the subject of filers, what is available for those on a tight budget?
So far I see, in no particular order:
Auspex used to advertise an AS200 model
but now apparently all they sell is the NS2000 family
- NFS, CIFS
- no RAID (?)
- scales
I had a few links to Dell but they have rearranged
their site and now all the old URLs don't work
Connex N3000 $5000-$10,000
- I don't have proper details
- to 90G
Maxtor maxattach $1000-$2000
- I don't have proper details
- to 73G
- IDE disks apparently
NetApp F720 $20,000+(?)
- NFS, CIFS
- RAID (type unspecified)
- scales a bit
- to 464G
- snapshots
Cobalt NASRaQ $1700/$2700
- NFS, CIFS, AFP(?)
- RAID 0 & 1
- no expansion
- 20G/32G
- IDE disks apparently
Quantum SnapServer 2000 $???
- NFS, CIFS, AFP, IPX
- RAID 0 & 1
- 40G
- IDE disks apparently
- more than 15 concurrent users probably not a good idea
- appears not to be available
HP Surestore HD Server 4000 $5000-$9000
- CIFS only
- RAID 5
- modestly expandable
- DDS-4 tape drive included
Veritas File Server Edition $??? + hardware cost
- NFS, CIFS (and presumably whatever other
protocols the host server supports?)
- RAID 1 & 5
- Solaris-based, software-only
- snapshots?
SPANStor $??? + hardware cost
- NFS, CIFS
- whatever RAID the hardware supports apparently
- Intel/3Com/Adaptec PC-based, software-only
(My use of acronyms above may be slightly inaccurate. US dollars.)
Can anyone fill in the ???s ?
Recommendations?
Things to stay away from?
What other products are available?
-jonathan
--
Jonathan H N Chin, 1 dan | deputy computer | Newton Institute, Cambridge, UK
<jc...@newton.cam.ac.uk> | systems mangler | tel/fax: +44 1223 335986/330508
"respondeo etsi mutabor" --Rosenstock-Huessy
Andy Isaacson
Jeff Wolfe <wo...@ems.psu.edu> wrote:
>Cray DMF[1]
[snip]
>[1] The One True HSM, as far as I'm concerned. It's too bad SGI sucked
>what life was left out of Cray and discarded them. I really thought
>the T3E was an interesting box.. We might have been looking at one of
>those now instead of an SP.
We're not dead yet! We're feeling better.
I think you might be able to buy a T3E today if you cozy up to your
local Cray saleswonk. At least, I heard somebody mumbling about
thinking about it. Ask (or have the right people ask) your Crayon
for a NDA presentation on the follow-on plans.
-andy (waiting for divestiture)
Jeff Wolfe
Andy Isaacson <adis...@pirx.candyland.cx> wrote:
>In article <86j944$31e$1...@laurasia.ems.psu.edu>,
>Jeff Wolfe <wo...@ems.psu.edu> wrote:
>>Cray DMF[1]
>[snip]
>>[1] The One True HSM, as far as I'm concerned. It's too bad SGI sucked
>>what life was left out of Cray and discarded them. I really thought
>>the T3E was an interesting box.. We might have been looking at one of
>>those now instead of an SP.
>
>We're not dead yet! We're feeling better.
Ahh, Monty Python..
>I think you might be able to buy a T3E today if you cozy up to your
>local Cray saleswonk. At least, I heard somebody mumbling about
>thinking about it. Ask (or have the right people ask) your Crayon
>for a NDA presentation on the follow-on plans.
I'm waiting for the divesture myself. I saw an NDA back in the fall,
in fact, we have an upgrade path for our J90 that's essentially the
same cost as the SP. Unfortunately, with the uncertainty about Cray,
most of the folks in our field are moving to MPI and DSM systems.
We'll see.
-Jeff
Rune Bakken
> I have found the NetApp sales staff to be extremely pushy and bad
> at trying to understand your problem. They seemed unable to understand
> why anyone where I work was questioning their wonderful product
> instead of just signing right there for $hugenum amount of chrome
> toaster.
I have not seen this kind of behaviour from our salesdroids, but I
guess it helps that the salesdroids is (former) technichians
themselves and that they did not approach us, but the opposite.
(We more or less knew most of the stuff they wanted to tell us about anyway)
>>>>>>>>>>>.rune
Andy Isaacson
As are we all.
>I saw an NDA back in the fall,
>in fact, we have an upgrade path for our J90 that's essentially the
>same cost as the SP. Unfortunately, with the uncertainty about Cray,
>most of the folks in our field are moving to MPI and DSM systems.
MPI is good. Can't beat the price/performance of ia32 or Alpha
clusters, but peak performance generally leaves something to be
desired compared to big iron.
Nobody involved is happy about the uncertainty. All I can say is
they tell us they're trying their best.
-andy
Matt McLeod
that nic...@esperi.org did write:
>I have found the NetApp sales staff to be extremely pushy and bad
>at trying to understand your problem. They seemed unable to understand
>why anyone where I work was questioning their wonderful product
>instead of just signing right there for $hugenum amount of chrome
>toaster.
Had a chat with "our" NetApp rep yesterday. Seemed OK, if a little
more pushy than our long-suffering Sun guy.
But there are a few things counting against them for the
purchase:
* he lied about CIFS -- claimed it was rock-solid, which according
to most who've mentioned it here is not entirely the case
* the price! close on AU$1k per 1G usable disk. That is a little
bit on the excessive side (they quoted us ~AU$81k for a low-end
box with 7x18G disks, and that's with the $VBC discount)
* the quote -- it was sent as a sodding Word document, despite me
going to great pains to make them understand that we're a Sun shop.
On the plus side, though:
* he claimed that they've just about got 'em certified by Rational
for handling all of the storage pools, rather than just some
* offered to give us one for a week or two to try it out
I suspect we'll just be going for an E450. It's cheaper,
more versatile, and adequate as a file server.
--
"Life. Loathe it or ignore it, you can't like it."
-- Marvin
Rune Bakken
> * the price! close on AU$1k per 1G usable disk. That is a little
> bit on the excessive side (they quoted us ~AU$81k for a low-end
> box with 7x18G disks, and that's with the $VBC discount)
I suspect you were quoted a F720, but that was a rather steep price,
even here in Norway[1] they sell for about NOK360k (~AU$68k).
Additional shelves with (7x18G) sells here for ~NOK110k (~AU$21k).
>>>>>>>>>>>>.rune
1/ A very expensive country to live in.
Matt McLeod
>use...@wibble.cx (Matt McLeod) writes:
>
>
>> * the price! close on AU$1k per 1G usable disk. That is a little
>> bit on the excessive side (they quoted us ~AU$81k for a low-end
>> box with 7x18G disks, and that's with the $VBC discount)
>
>I suspect you were quoted a F720, but that was a rather steep price,
>even here in Norway[1] they sell for about NOK360k (~AU$68k).
>Additional shelves with (7x18G) sells here for ~NOK110k (~AU$21k).
Yes, it was an F720. Was that NOK360k for just the base unit,
or for the base plus a shelf with 7x18G?
Thing is, a reasonably-specced E450 with ~100G of disk in it
is only ~AU$60k.
>1/ A very expensive country to live in.
You haven't tried purchasing kit in .au, have you? :-(
--
"Don't tell my momma I'm a sysadmin, she thinks I play piano in a whorehouse."
-- Alan J Rosenthal, ASR
Stig Sandbeck Mathisen
> * offered to give us one for a week or two to try it out
By all means, do. You can always say "no" after that time.
--
SSM - Stig Sandbeck Mathisen
Trust the Computer, the Computer is your Friend
Matt McLeod
>* Matt McLeod <use...@wibble.cx>
>> * offered to give us one for a week or two to try it out
>
>By all means, do. You can always say "no" after that time.
I know.
At present I'm trying to find a way of taking them up on
the offer without feeling guilty about it (don't really
possess the Bastard nature -- I imagine He would accept
the offer then find some way to remove all evidence that
the equipment had ever been delivered).
As it happens, we've got to do some longer-term planning
in the nearish future, and a Filer could very well be
part of that. Or so goes my rationalizaton, anyway...
--
/earth is 98% full ... please delete anyone you can.
Matt McLeod
>In article <slrn892r1s...@enzo.netizen.com.au>,
>Matt McLeod <use...@wibble.cx> wrote:
>>* he lied about CIFS -- claimed it was rock-solid, which according
>> to most who've mentioned it here is not entirely the case
>
>with NT.
Our CIFS solution right now is Samba. Works well enough, and
doesn't cost us anything. The company does run a literal
imperial arseload of NT swervers at the main sites, but we
refuse to have anything to do with them.
>>* the price! close on AU$1k per 1G usable disk. That is a little
>> bit on the excessive side (they quoted us ~AU$81k for a low-end
>> box with 7x18G disks, and that's with the $VBC discount)
>
>They're proud of their kit.
There's such a thing as being *too* proud. AU$1k per G is
a bit on the excessive side IMO.
>>I suspect we'll just be going for an E450. It's cheaper,
>>more versatile, and adequate as a file server.
>
>We went through the purchase process twice. The first time we made the same
>decision, and got a general purpose box. We did need the general purpose box,
>but it had limitations as a file server. The second time we sprung for the
>NetApp. It makes a big difference, especially when it comes to providing for
>our Windows users.
That's roughly how I'm thinking at the moment. We could make
good use of the E450 to split the ClearCase stuff up, not to
mention providing a stable user-free machine for important stuff.
But we do have to do something longer-term, and a Filer may
well fit into that rather nicely.
--
So little time, so little to do.
-- Oscar Levant
Rune Bakken
> Yes, it was an F720. Was that NOK360k for just the base unit,
> or for the base plus a shelf with 7x18G?
If you mean base as just the 'head', the price was for the base and
one shelf with 7x18G
> Thing is, a reasonably-specced E450 with ~100G of disk in it
> is only ~AU$60k.
But then again, what du you do when you run out of that disk space?
And you will without the slightest doubt do, soon - very soon.
>>>>>>>>>>.rune
Matt McLeod
>use...@wibble.cx (Matt McLeod) writes:
>
>> Yes, it was an F720. Was that NOK360k for just the base unit,
>> or for the base plus a shelf with 7x18G?
>
>If you mean base as just the 'head', the price was for the base and
>one shelf with 7x18G
OK. Then the prices are incredibly high here. Especially
as the price we were quoted was discounted.
>> Thing is, a reasonably-specced E450 with ~100G of disk in it
>> is only ~AU$60k.
>
>But then again, what du you do when you run out of that disk space?
Add another 100G. It has room for 20 disks internally.
>And you will without the slightest doubt do, soon - very soon.
Oh, certainly. Eventually. Thing is, we don't buy, we
do 3-year leases tops. And the current system (an E4k with
~50G of disk in an SSA) will come due for replacement in 12
months, at which point getting a proper file server will
become a lot simpler.
--
Mistakes are often the stepping stones to utter failure.
Rune Bakken
> Oh, certainly. Eventually. Thing is, we don't buy, we
> do 3-year leases tops. And the current system (an E4k with
> ~50G of disk in an SSA) will come due for replacement in 12
> months, at which point getting a proper file server will
> become a lot simpler.
I am not sure if your NetApp provider is as easily turned as the ones in
Italy, as they originally quoted similar prices - upon my Italian
co-workers told them "We want the price they pay in Norway!!" and they
did.
Did I see a mention of an SSA above? We do have a few of them, and I am
trying to figure a way to crash them or have them abolished from our
systems. They may have been OK once, but the world does evolve, I
hope.
Matt McLeod
that Rune Bakken did write:
>> Oh, certainly. Eventually. Thing is, we don't buy, we
>> do 3-year leases tops. And the current system (an E4k with
>> ~50G of disk in an SSA) will come due for replacement in 12
>> months, at which point getting a proper file server will
>> become a lot simpler.
>
>I am not sure if your NetApp provider is as easily turned as the ones in
>Italy, as they originally quoted similar prices - upon my Italian
>co-workers told them "We want the price they pay in Norway!!" and they
>did.
I'm thinking that the trick to use would be to find out how
much the company is paying for them elsewhere -- I know we
have a massive installation of them in Stockholm, and I wouldn't
be surprised if the Texan site(s) had some too.
>Did I see a mention of an SSA above? We do have a few of them, and I am
>trying to figure a way to crash them or have them abolished from our
>systems. They may have been OK once, but the world does evolve, I
>hope.
There's a very simple way of getting rid of them. Sun won't
support them with disks larger than 4G. So they've got a fairly
limited capacity.
At least this is true for the model we've got -- SSA100,
from memory.
This is the biggest advantage of leasing IMO. The finance
side of it doesn't make too much sense to me (I can understand
spreading the capital expense over the life of the equipment, but
you could do the same by borrowing instead, with the result of
owning the kit at the end), but the very well-defined lifespan
of the kit can come in handy...
Matt McLeod
>>>>>> "usenet" == Matt McLeod <use...@wibble.cx> writes:
>
>usenet> There's such a thing as being *too* proud. AU$1k per G is a
>usenet> bit on the excessive side IMO.
>
>*cough*EMC*cough*
I take it they're even worse?
What I'm saying, I guess, is that if the price were more
reasonable they'd probably have a sale. But at the price
they're asking management would simply refuse to sign off
on it -- I haven't had time to try digging up any dirt
on any of them yet. ;->
--
"I'm not sure if this is a good or a bad thing.
Probably a bad thing; most things are bad things."
-- Nile Evil Bastard
Russ Allbery
> Kendall Libby did write:
>> *cough*EMC*cough*
> I take it they're even worse?
One of the prerequisites for purchasing EMC is that you're a large enough
shop that none of the technical people ever see the actual invoice.
(We're a large enough shop. I love the things. I don't want to know
exactly how much we paid; I have a good enough idea already.)
Matt McLeod
>buttload of niggling shortcomings.
Fair enough. We only use it to share a few things with the
management, usability, and admin types who insist on using
Windows PCs for some reason. So it's not exactly a big deal
to us.
--
"Yup, men are from Macs,
women are from VMS."
-- Erwin, User Friendly by Illiad
Nicest of the Damned
< use...@wibble.cx (Matt McLeod) writes:
< > Thing is, a reasonably-specced E450 with ~100G of disk in it
< > is only ~AU$60k.
<
< But then again, what du you do when you run out of that disk space?
Trim down the MP3 library a bit?
-jer
--
j...@sidehack.gweep.net Programmer / Sysadmin / Gweep
mpy...@gnu.org -><- Author of Xtacy, an X11 Graphics Hack
jer_j...@real.life.edu http://www.gweep.net/~jer/index.html
Josh Brandt
Nicest of the Damned <j...@sidehack.sat.gweep.net> wrote:
>< But then again, what du you do when you run out of that disk space?
>
>Trim down the MP3 library a bit?
AH ha ha ha ha ha!
No, seriously. 8)
Josh
--
I don't wanna ride the piggy.
J. Brandt / mu...@sidehack.gweep.net
Rune Bakken
> Trim down the MP3 library a bit?
Go wash your mouth, and rewrite your priority list!
Actually, a co-worker at my pervious employer (the backup manager)
scanned the backupdb for new mp3's on the public login-server we were
admining - officially in order to find and remove them if the
world/group read bit was set - ie considered distributing a
copyrighted material - a big no-no.
Matt McLeod
>j...@sidehack.sat.gweep.net (Nicest of the Damned) writes:
>
>> Trim down the MP3 library a bit?
>
>Go wash your mouth, and rewrite your priority list!
Fortunately this is not a useful option at ork -- our archive
is highly distributed -- a bit here, a bit there... Including
a rather large bit at another site where they've got something
of an excess of disk.
>Actually, a co-worker at my pervious employer (the backup manager)
>scanned the backupdb for new mp3's on the public login-server we were
>admining - officially in order to find and remove them if the
>world/group read bit was set - ie considered distributing a
>copyrighted material - a big no-no.
Wish I'd thought of that earlier. Good thing I turned indexing
on for the home fs dumps...
Hanno Wagner
Am 28 Jan 2000 12:27:44 GMT meinte Peter da Silva:
>It's been pretty solid for us... it's certainly "rock solid" when compared
>with NT.
Well - we use it as NFS-Server and CIFS. Sometimes CIFS crashes the
system if the NTs are excessive in requesting stuff. But all in all
- its good and even if it ever needs a fsck (I had this some weeks
ago) there was no data loss. I like that stuff really - if it
detects any problems, it simply mails. and shuts itself out in time.
>>* the price! close on AU$1k per 1G usable disk. That is a little
>> bit on the excessive side (they quoted us ~AU$81k for a low-end
>> box with 7x18G disks, and that's with the $VBC discount)
>
>They're proud of their kit.
IMHO it's worth its price. Too bad their highest stuff is 1 1/2 TB...
Ciao, Hanno
--
void
>
>RAID-5 is striped in chunks (64k, 128k). IIRC, RAID-3 is striped by the
>bit, byte, or sector (or something like that). It's supposed to be good
>for large files - it spreads reads and writes out over many devices
>simultaneously, rather than by chunks.
RAID-3 is closer to RAID-4 than it is to RAID-5. Specificly, a level 3
RAID stores parity all on one disk, like a level 4 RAID.
The difference between RAID-3 and RAID-4 is in the granularity of the
chunks of data, IIRC. *handwave*
--
Ben
"Cleverness is serviceable for everything, sufficient for nothing."
-- A fortune cookie, lunch, 25 December 1999
p...@black.plague.apana.org.au
While catching up on some news I saw this gem:
On 28 Jan 2000 12:56:29 GMT, Matt McLeod <use...@wibble.cx> wrote:
: [...] we've got to do some longer-term planning
: in the nearish future, [...]
I love it. Even better than "we'll get those backups done Real Soon Now"
-Pik.
--
$f='|2 /kip~/ua.gro.anapa.retnuh.www//:ptth1241213151|5 ua.gro.anapa.eugalp@'
.'kip:otliam1551113151|6273|3 rekcaHrJ lreP / nimdAsyS13 911130o2121|3 gni'
.'rekciP giarC - KiP23912173';$f=~y/|/\n/;while($f){$c=chop($f);if($c=~/\d/){
$x=(($x ne "#")?"#":" ");for(1 .. $c){print "$x";}}else{print $c;$x="";}}#EOF
Tony Finch
>
>RAID-3 is closer to RAID-4 than it is to RAID-5. Specificly, a level 3
>RAID stores parity all on one disk, like a level 4 RAID.
>
>The difference between RAID-3 and RAID-4 is in the granularity of the
>chunks of data, IIRC. *handwave*
Yeah. The RAID levels were defined ages ago, when intelligent drive
hardware was hard to come by: RAID2 is the same as RAID3, but instead
of a parity disk, it has a number of ECC disks. This is useless
nowadays when drives can do their own ECC. The difference between
RAID3 and RAID4 is that if you read a disk block from RAID3 it comes
from all of the disks at the same time, but if it's RAID4 it only
comes from one of the disks, hence you can parallelize IO from a RAID4
array.
Tony.
--
** **
*** *** *** **** *** ******* *** *** **** *** ****
*** *** * *** *** *** * ***
void
>
>The difference between RAID3 and RAID4 is that if you read a disk block
>from RAID3 it comes from all of the disks at the same time
So RAID-3 implies synched spindles?
--
Ben
`~~"~"`"'""''"`""~`"~~``'`""``""`"'`'~`~"~"'~~`~`'"''`"`"'~`""``"~''~"""~''`""``
``"~""`"`'~`'''`'~~'~~'`'~''"~~``"'~'~~~'``~"~'``"~`"'"'"`~"`~~"~"~""~~``''`'`""
~"`"'```''~~"`~~~~"~`"`"`~~~`"~`"`~`''"``~'~''~""~"`~`~'``''`'~~~`''"`~~''`~``"`
'"''~"~`~'~``~'``~~``~'~""~``~"~"`'"`"`''`""'``~'``~'~~'~`"''"'`~~`'~"~'`~'"'"~~
Matt McLeod
>
>While catching up on some news I saw this gem:
>
>On 28 Jan 2000 12:56:29 GMT, Matt McLeod <use...@wibble.cx> wrote:
>
>: [...] we've got to do some longer-term planning
>: in the nearish future, [...]
>
>I love it. Even better than "we'll get those backups done Real Soon Now"
Of course, it gets even better as part of the longer-term
planning involves finding alternate employment.
But there is actually a good reason for the lack of planning
so far -- corporate politics.
Tony Finch
It'd be helpful, but I don't thing the RAID definitions have anything
to say about how to make an efficient array.
Jasper Janssen
>You misspelled "Mariana Trench"[1]
Are you absolutely sure he did not misspell "Marianas trench"?
Jasper