Run_II_Tapes_for

Tape Technology For Commissioning Run

D. Amidei March 2000

There is a Committee to consider this for Run II

"The Serial Media Working Group"

D. Amidei, M. Diesburg co-chairs

J. Bakken, D. Benjamin, B. Blumenfeld,

K. Chadwick, G. Cooper, S. Fuess, K. Genser,

S. Lammel, L. Lueking, J. Patrick, D. Petravick,

D. Schamberger, H. Schellman, R. Thies,

A.Walters

General Background:

General purpose robot allows (almost) any tape product

Waiting out the market place for next generation
technologies which promise substantial gains in capacity
& cost effectiveness

Focussed on commodity options with good road-maps

SONY Advanced Intelligent Tape:

AIT-1 25GB 3MB/s

AIT-2 50GB 6MB/s shipping "now"

Drive: $4500/3500 (list/us)

Tape: $125/90

AIT-3 100GB 12MB/s "2001"

EXABYTE Mammoth

M-1 20GB 3MB/s

M-2 60GB 12MB/s shipping "now"

Drive: $4000/4000

Tape: $90/80

M-3 120GB 18MB/s "2001"

Extensive testing of M-1 and AIT-1 by S. Lammel, and
Petravick team in CD.
Way better than our memories of 8mm, one of each still
running after 20K mounts, compare to 450 mounts with
Exabyte 8500.
Sense of the committee: Either technology would work,
Mammoth has x2 bandwidth and slight edge in overall
engineering.

Issues

1. AIT II Media Problem

Both technologies use Metal Evaporated Tape

For the 50/60G tapes (230m)

Kevlar-like substrate for AIT has poor yields with the
AME process. Tapes are backordered. Second supplier
due in April. SONY will not engage in frank discussion.

Comparative gentleness of Mammoth tape path allows a
different substrate with good yields in AME. 4 suppliers,
cartridges available in volume.

2. Mammoth II Hardware Problem

At least a quarter late. We have been cutting them slack,
working with betas since September 99.

Very good technical engagement with company, but still
do not have an M-2 that works with FNAL tape code.

3. The Commissioning Run Forces Action

Either use M-1 for commissioning run, count on M-2
coming along.

Or get started with AIT-2, test the media reliability and
availability on small scale.

Make mid-course correction in October if needs be

Tapes for the Commissioning Run

A."20 Hz for 4 weeks" ???

Consensus, but no real knowledge about this number, or
conditions. Hardware triggers? Myron mode?

B.Event data assumptions:

No silicon.
RAW: 200kB/event x 20 Hz = 4 MB/s ?
L3 DST: 100kB/event x 20 Hz = 2 MB/s ?
FARM (PAD) 50kB/event x 20 Hz = 1 MB/s ?

C.Tape writing steps

1. FCC data logger: RAW+L3 DST = 6MB/s
2. Farm input: RAW+L3 DST = 6MB/s
3. Farm output: RAW+L3 DST+PAD =14 MB/s (2x7) ?
4. User jobs: 30 MB/s ???

Total = 56 MB/s

D. Drives required

Assume vendor BW spec, and contingency 1.5 hot spares

AIT-2 56 / 6 = 10 x 1.5 = 15 drives @ $3500 = $52,500

M-1 56/2 = 28 x 1.5 = 42 drives this would be nuts…

E. Media requirements:

Assume users do not write tapes ???

Step 1 + 0.5 x Step 3 = 13 MB/s x 4 weeks = 31 TB (???!!!)

Note: the disk pool will be ~ 7TB

AIT-2 31TB / 50 G B x $90 = $56,000

M-1 31/TB / 20 GB x $60 = $93,000 also unpalatable…

Recommendation 3/23/00

Unless a miracle occurs with Mammoth-2:

The use of Mammoth-1 for the Commissioning Run
seems untenable with the assumptions above, and is
also a poor investment of manpower…

We believe the AIT-2 media problem represents a
smaller overall risk than either of the Mammoth options.

We lean to committing to AIT-2 for the Commissioning
Run. We would purchase minimal number of drives
and cartridges, with contingency to revisit the question
for "Run II" in October, after experience with AIT-2
media, time to see the Mammoth-2 situation evolve.
Note that ultimately we have to converge with D0 as well.

n.b.1 Don’t run out and buy tape drives, unless you
are prepared to repurchase for the real Run II

n.b.2. It would be very useful to know export
needs, if any for the commissioning run. Would
we need to copy to 8500?