In article
<c79fcde2-3668-4d2e-ab76->,
js.b1 <> writes
>Own tests on various machines 6.2-6.7MB/sec using MSFT Backup.
>DAT72 tape on DAT72 USB2 external.
I venture to suggest that, although they may be streaming, the source
cannot supply data fast enough. The only definitive test is to stream a
large fragment from memory (eliminates disk channel bottleneck). The
original MS Backup was OEM from Seagate software and not very quick! I
don't know how much the engine was improved with later releases; I do
know how much time I used to waste trying to get decent performance out
of it! It was remarkable how significant the software component was to
performance, although I haven't run any tests in recent years.
Recent DAT and LTO generations have a 'slow speed mode', IIRC, to
minimise scrubbing with slow hosts, but there are limits!
>p40 - "Prolonging Head Life"
>"...typical head life in excess of 6000 hours".
>"This figure assumes that you use the appropriate tape for
>the drive (such as DDS-4 media with DDS-4 drives) and that
>you follow a typical usage pattern."
>"You may not attain the typical head life when:
>- You use an older tape format, such as DDS-2 tapes with a DDS-3
>drive.
I asked HP yesterday on-line for the current Vol 1. so as to be able to
check their current statements. They only sent me that white paper. I
don't have a current copy, but I do have the same part number - dated
2003 and the DAT72 stuff in it is provisional.
The section in question (p.40) is largely unchanged from the original
information published with DDS-1 (there wasn't an integration guide as
such then, but from memory it is similar to the OEM tech ref manual of
the time). I suspect it was only given a cursory check when it was
updated for the 2003 release (but I still don't know what the latest
numbers are, nor if HP has made them public). The drum, incidentally,
would be replaceable on those units, although I don't know about later
models.
DDS-2 was a horrid tape formulation, incidentally, that really got
through heads quickly.
>- A large proportion of cartridges loaded are being used for the first
>time."
Actually, this gives *worse* reliability than re-using tapes as much as
possible.
The reason is that new tapes tend to have microscopic rubbish all over
them from the manufacturing processes. This is scraped off in the first
few passes through the drive, whereafter the raw error rate decreases to
a reasonably constant level, until the tape finally wears out (100-200
passes, typically). If you always put new tapes in, it can thus increase
the head wear (although since DDS-3 drives have hardware included to
clean head + tape ). Tape brand can also have a significant effect.
>Seems pretty clear to me. Covers to DAT72.
Well, it makes sense, but I think they're being very conservative (and
the MTBF is optimistic!).
>IBM LTO-2.
>Other LTO drives are double that (eg, Dell, HP) and later editions
>have better head cleaning systems.
Bear in mind, in all cases, the head life quoted is (a) conservative,
and (b) the result of statistical analysis of testing. AFAIK, it's not
derived from real life failure data, although can cause MTBF to be
reviewed.
>My first encounter with DAT was HP C1533A in 1994.
I remember it well.
>I recall HP was the most reliable of all the makes available.
>I also recall many installations failed to stream correctly and
>thus both heads and tapes were subject to incorrect usage.
Indeed so. There remains very little awareness of the issue within the
industry, partly because of the imaginary transfer rates appearing on
disk specs. It's still a problem.
>People also dropped the cartridges without regard, without
>doing a retension even. The 4mm housings are not strong.
They'll survive quite a lot of abuse. I once trod on a cartridge at a
trade show: It loaded, but I wouldn't have wanted to put it into any
backup pool! In contrast, the mechanical issues with, say, DLT were
horrid: if you dropped one on a corner from chest height onto a hard
floor, you'd break it. Don't get me started on the leaders...
>Not see DAT 8mm (160/320), there was always a racket
>that the old 8mm tape was superior. Well yes, but it soon
>died a quick death compared to DLT & LTO but they have
>a long evolution from IBM DEC 3480 era - and big $$ spent.
DLT was always a horrible bodge. The LTO designs essentially started by
dissecting DLT and compiling a list of "we're not going to do that!"
items.
>> >DAT160 & DAT320 are very unreliable and nearly uneconomic to repair
>> >compared to DAT72 from three repair shops.
>>
>> Define "very unreliable" and state your source. I know one popular tape
>> technology from the 1990s that had an AFR (annualised failure rate) of
>> around 300%. It wasn'tDAT(and it wasn'tHP'seither).
>
>Ever remember Colarado Jumbo or Travan... /dev/null?
Oh yes. Or perhaps just "Oh!" As you say, many were input-only devices.
The mechanical designs were just nasty.
>By very unreliable I mean significantly higher failure rate when
>subjected
>to full capacity daily backups compared to DAT72. The evolutionary
>fixes
>from DDS1 to DDS5 improved reliability.
I can't speak to the dual standard ones (my experience of DDS is DDS 1
thru 4). If they are unreliable, I'm not surprised, as the geometry
would be asymmetric.
>The source was 1 USA tape repairer and 2 UK tape repairers one
>of which I know well. I still believe people are miss-handling tapes.
Probably. I wouldn't use a dual-gauge format personally. LTO in
comparison is brilliant, but you pay for what you get.
>It is quite possible people are using the wrong tool for the job,
>thrashing DAT160/320 when something like LTO would be more suitable.
It's a difficult one. The issue of not-streaming doesn't have any really
robust solution, although LTO is far better at 'throttling' than
predecessors. There is an argument that it's far better to use a drive
that matches the throughput of the storage subsystem, i.e. slower if
necessary.
>I suspect these could be Dell badged units bundled with servers and
>someone on the purchasing side sees a budget saving over LTO.
Dell OEMs from someone (don't know who at the moment). One of the
biggest issues with all tape drives (except in high-end library
configurations) is environment: dust kills them. If these are small
servers, the chances are they're not used in the cleanest locations.
>LTO wins hands down, it has a long evolution path, it has WORM &
>Encyrption capability - but it is expensive for small business.
I agree on that, however I'd suggest that most small business, even
these days, doesn't put a realistic value on their data. If they did,
LTO would be more prevalent as cost-effective 'insurance'.
>The biggest downside with DAT used to be interchangeability.
>The tolerance on drives was such that two drives could be at opposite
>ends of the spectrum and thus write tapes they could read but others
>could not. That I suspect was a batch of non-HP drives, although a few
>HP drives suffered it - statistical tail end of the distribution as it
>were.
It wasn't manufacturing tolerances as much as the propensity for tape
guides to work loose from factory settings in high-vibration
environments. Data wasn't lost, and the models concerned had the
manufacturing/alignment process altered to fix it. The issue was made
worse because of the relative success of DDS at the time - other
technologies were as bad/considerably worse, but because they weren't as
popular, the issues weren't as generally obvious. All this was around
eighteen years ago, though!
I know interchange tests between DAT manufacturers were taken very
seriously, with specimen tapes being exchanged to confirm they did what
was claimed.
Cheers,
S.
--
SimonM
Similar Threads
Linear Mode
