It isn't really. SDLT replaces DLT but some backwards-compatability
remains. There are so many DLT systems in service that there should
be no problems coming by tapes & parts for the next few years (which
is all you should expect anyway).
That's software & configuration specific except for some LTO's that
can adjust speed to compensate.
Lady said:Can you be specific about LTOs that can adjust speed? Exabyte claims
their VXA drives are the only ones that can adjust speed. For me
that's important. I like the larger capacities of more recent drives,
but tape speed is an issue.
SDLT, DLT 8000, and most LTO drives have variable speed capability. The
only LTO's that don't seem to have it are HP and IBM versions of LTO1.
FWIW, it is German "kaputt", from Yiddish "kaput", which made its way
into American slang.
Just *had* to throw this in...
Curious said:Then why is it only advertized in Seagate LTO-1 and newer HP & IBM
LTO-2's?
I'm not sure what you are implying.
There are these things called Search Engines that help a lot with thisCurious said:That AFAIK advertized specs & manuals don't corroborate that claim for
DLT 8000, SDLT's, & several early model LTO-2's. Would you mind
giving an official reference for either DLT flavor for this feature?
There are these things called Search Engines that help a lot with this
sort of thing ;^):
As far as I know, all SDLT drives after this included this feature,
however, you are correct that it doesn't seem to be in the literature.
I found that SDLT-600 does have it but possibly the SDLT-220 does not.
In any event, whether it is due to better buffering or some sort of
variable speed, I get no shoe-shining when backing up over 100 Mbps
network on a SDLT 220 drive. It may be theoretically possible to get
11MB/sec on a 100 Mbps connection but it is cutting it very close and
the SDLT is doing something to compensate.
Further, I find it odd that
they would have had this feature on the DLT-8000 but not on SDLT-1.
Each vendor calls it something different: FastSense, Digital Speed
Matching (DSM) and Adaptive Tape Speed (ATS) from Certance, IBM and HP
respecitively.
Actually, the HP link indicates that it was available even on LTO-1 so
it looks like IBM LTO-1 was the only version not to have this feature.
Or doing more digging. You're right though to an extent. It took longerCurious said:Naw. Ya think? So the secret to avoiding such criticism is to just
write "Nope" or "Clueless" ; )
Search the document for "variable" and the first mention of it is thatActually here's a link I didn't see earlier in the conversation or the
past:
http://tinyurl.com/9e9ge
According to this SDLT does do it using "Digital Data Rate Agent"
(DDRA). It claims it was first introduced in the SDLT 220 (not the
DLT 8000). I thought the SDLT 220 didn't have it. That appears
wrong.
did that before the last post & other times in the past. Don't see
anything that resembles the feature. The user man refers to an
"adaptive control mechanism" & "Dual Speed Recording" but AFAIK
neither is a "variable speed" feature. I'd love to be proven wrong
and be able to implement a feature I've been neglecting to use from
ignorance.
Could be. It runs pretty smoothly in any case. With TCP overhead, 11SDLT600 is newest/best and not necessarily indicative of the whole
SDLT line (e.g. DLT ICE). Apparently DDRA is part of the whole SDLT
line anyway.
True it isn't really in the literature. even a search of Quantum's
knowledge base of "DDRA," "Digital Data Rate Agent," (or parts of that
term) ,"variable speed" etc, turns up no hits. My searches of manuals
look the same. It makes no sense to me Quantum would want to hide
this feature.
Buffering is a major factor. Without enabling the DDRA it's very
possible to backup over 100Mbs near / just shy of full speed. That
means basically a good stream with some very infrequent stops (perhaps
too infrequent to notice?) but definitely not bad shoe-shining per se.
Just trying to create a complete record.But I'm not convinced it is on the DLT 8000. The manual's timing
specs show fixed rates with no discussion of DDRA or similar. But
then who can trust the manual. : (
I'm not sure I agree it is _so_ unlikely a feature could be present,
dropped, and not picked up again until a generation or two later. It
wouldn't be the first time someone had trouble with a "new" feature
that wasn't playing well with the next new reworked design and it had
to be put on the shelf until the kinks were worked out.
I always agreed Seagate/Certance had this since LTO-1
Here is what it says:Those are recent HP LTO's
Actually it did. From 6 MB/sec to 7.5 MB/sec:HP Ultrium 215 did not have it (LTO-1).
Clueless. (I'm just kidding!) See above.It was introduced in the 230.
I thought there was an IBM LTO-2 without it also, but cannot recall
the model #.
Or doing more digging. You're right though to an extent. It took longer
than I expected to find the references to this ubiquitous feature. Sorry
for the cheap jab.
Search the document for "variable" and the first mention of it is that
they changed the term "Variable Speed Recording" to "Dual Speed
Recording". It doesn't specify the range though so who knows?
Could be. It runs pretty smoothly in any case. With TCP overhead, 11
MB is more like 110 Mb so even with 100% utilization of the the network
channel there would still be a 10% overrun. I can't believe it would
stream this well, even with some buffering, without it changing speed to
match the data coming in.
Just trying to create a complete record.
Here is what it says:
Q1. Does Ultrium 960 FC tape drive support Adaptive Tape Speed (ATS)?
A1. Yes, the HP StorageWorks Ultrium 960 FC tape drives available
throughout the entire ESL portfolio do carry forward the ATS technology
from the HP StorageWorks Ultrium 460 and 960 tape drives. The range for
the second generation of drives was from 10-30 MB/sec; for the HP
StorageWorks Ultrium 960 FC tape drive the range is 27-80 MB/sec.
So, the Ultrium 460 (LTO-2) has a range of 10 MB/sec to 30 MB/sec.
Actually it did. From 6 MB/sec to 7.5 MB/sec:
http://h18006.www1.hp.com/products/quickspecs/12128_div/12128_div.html
Clueless. (I'm just kidding!) See above.
Bottom line: DLT-8000, SDLT, and LTO drives have this feature except
the IBM LTO-1.
I haven't any experience with the Tandberg versions of
these drives. I'm guessing they have this feature too from what little
I've seen of their products.
I agree.Curious said:I still don't understand, and don't believe we've successfully
answered, why such a useful feature would be so obscured to even
require "more digging." That _should_ be a significant selling point.
Unless there is a conspiracy to promote LTO. 8-P
The only "special" things I do in the backup to the SDLT 220 are:it isn't usually so smooth. not out of the box anyway. Unless I've
just been masking a broken/misconfigured SDLT220 with tweaks or
expecting a wider range of variable speeds than exists. It doesn't
sound like you did anything special to enable "variable speed/dual
speed recording." Is that true?
The only "special" things I do in the backup to the SDLT 220 are:
Set compression on:
mt setdensity 145
use a larger block size than the default tar block size (512 KB instead
of 10KB):
tar cvbf 1024 /dev/tape files
This really helped speed things up a lot for me. It helped alot when I
was using a DLT 7000 too.
Since I do set hardware compression on with setdensity, this makes
another case that it has variable speed capabilities since with
compression, it wants up to 22 MB/sec.
11MB/sec only when data is very compressible.
Correct. However, with compression on and with very compressible data,Curious said:Thanks for sharing that. But I don't quite follow what compression
has to do with variable speed. 22MB/sec should not mean the tape
moves 2x as fast.
Correct. However, with compression on and with very compressible data,
the drive will require 22 MB/s to keep streaming. The tape speed stays
the same but now there is more demand from the network interface, and
SCSI bus for that matter. You are now trying to get somewhere around
220 Mb/s (raw ethernet) from your 100Mb/s network interface and
something has to give, so the tendency to "shoe-shine" is increased.
Rewording it: the tape drive now needs 22MB/sec to stream but the
network card can only deliver ~10MB/sec so the tape has to stop and wait
for the network card, unless it can vary the tape speed to slow down to
10MB/sec. The problem is just amplified when compression is used.
22MB/sec from a 10MB/sec pipe vs. 11MB/sec from the same 10 MB/sec pipe.
Yes. I never meant to imply that compression indicates variable tapeCurious said:But that still does not mean compression capabilities 'indicate' that
the drive 'has the ability to slow up automagically' when there are
throughput problems (i.e. variable speed).
Exactly. Since I was seeing no shoe-shining even with compression on (ieI agree, though, that compression can make it easier for there to be a
scenario where demand exceeds throughput, and _if_ the tape has the
ability to slow it will. But if it doesn't it won't; it'll just
shoeshine.
But that still does not mean compression capabilities 'indicate' that
the drive 'has the ability to slow up automagically' when there are
throughput problems (i.e. variable speed).
I agree, though, that compression can make it easier for there to be a
scenario where demand exceeds throughput, and _if_ the tape has the
ability to slow it will. But if it doesn't it won't; it'll just
shoeshine.
Yes. I never meant to imply that compression indicates variable tape
speed. What I said was:
"Since I do set hardware compression on with setdensity, this makes
another case that it has variable speed capabilities since with
compression, it wants up to 22 MB/sec."
Exactly. Since I was seeing no shoe-shining even with compression on (ie
somewhere between 11 and 22 MB/sec) this indicates that the drive is
matching the speed to the incoming data.
Curious George,
Your last paragraph has defined my question exactly. (Ignoring
compression). Which technologies or vendors do in fact adjust tape
speed? Exabyte claims that their VXA drives are the only ones that do
in fact adjust tape speed.
I've seen terms like "adjustable" or "dynamic" on other vendors'
sites, but it is possible that these vendors are conflating the change
in incoming data rate as a result of changes in compression ratio.
Or, it's possible that some drives don't do compression if the
incoming data isn't arriving fast enough. It's not possible to tell
from the descriptions on the web sites.
--Lady Thatcher--
That's dishonest. Others do as reported here by Steve Cousins.
Unless they mean the only "consumer tape drive." In which case it's
misleading but correct AFAIK.
Not everything is equally compressible. The compression algorithm
handles that in firmware or software outside of that other feature.
The other feature you are looking for is basically a drive not waiting
for a total and expected buffer underrun and slowing the tape to keep
the buffer filled. Two separate ideas. Multiple marketing names
Some drives aren't capable of compression (rare & unusually bad/old)
or speed compensation (now common but not totally ubiquitous). Some
that are don't advertize the fact they do in the most clear way. The
short answer is basically all the DLT's & LTO's currently being made
or recently retired do this & are worth a look.