I am still trying to collect contacts for some of the STK sites; I am aware
of about six customers switching "TO" 9940 (with SN6000) from something
else - one is replacing their 3590 with 9940's (not LTO).
Changing from 3590 to LTO is definitely a step "backward" in reliability;
performance is also 8-10 times slower... the basic idea behind LTO was to
provide a cost-effecting, competing product against DLT -- which is exactly
what it is.
LTO is 8-10 faster (and more reliable) than DLT; all this, at a price point
competitive with DLT. 3590 is 8-10 times faster and worlds more reliable
than LTO (and 9840/9940 which is comparable with 3590). I think of
3590/9940 as the industrial-strength answer for large data centers -- large
to me means moving over 1.5 TB per day, storing over 20 TB of data in their
silo; smaller sites, more price conscious, willing to "tolerate" slower
restore speeds are the ideal target for LTO. Most shops know that DLT is
slow; if we can approximate 10-15 GB/hour in restoring a file server, they
know that is a good number -- and I have demo'ed that with Dell PowerVault
130 using DLT7000. Large db's can be restored at 20-36 GB per hour; DLT is
at the low end of that, LTO is at the high end -- you just about saturate
the 100 Mbps wire at 30 GB/hr.
LTO, like DLT, does not like alot of "back-hitch" operations; so, I would
minimize the amount of collocation on LTO. Every account I know that has
LTO is ecstatic about the performance; the LTO design was based on 3590
technology, but (to cut costs) some reliability factors were sacrificed...
this is not a drive you want to run without a service contract, especially
if you're gonna beat it with the "back-hitch" action that's required to
back-space from eot, find last tape mark, start writing next data block from
end if the inter-record gap of the last one, etc.
One person's opinion, stretched over a dozen or more customer accounts.
Don France
Technical Architect -- Tivoli Certified Consultant
Tivoli Storage Manager, WinNT/2K, AIX/Unix, OS/390
San Jose, Ca
(408) 257-3037
mailto:don_france AT att DOT net
Professional Association of Contract Employees
(P.A.C.E. -- www.pacepros.com)
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU]On Behalf Of
Joni Moyer
Sent: Thursday, August 15, 2002 6:30 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: IBM vs. Storage Tek
Hello,
I know I've asked about this before, but now I have more information so I
hope someone out there has done this. Here is my environment for TSM.
Right now it is on the mainframe and we are using 3590 Magstars. We have a
production and a test TSM server and each has about 13 drives and a total
of 5,500 tapes used for onsite and offsite tape pools between the 2
systems. Two scenarios are being considered (either way TSM is being moved
onto our SAN environment with approximately 20 SAN backup clients and 250
LAN backup clients and will be on SUN servers instead of the mainframe)
Here is what I estimated I would need for tapes:
3590 9840 9940A LTO
10 GB 20 GB 60 GB 100 GB
Production
Onsite 1375 689 231 140
Offsite 1600 800 268 161
Total 2975 1489 499 301
Test
Onsite 963 483 163 101
Offsite 1324 664 223 135
Total 2287 1147 386 236
Grand
Total 5262 2636 885 537
1. IBM's solution is to give us a 3584 library with 3 frames and use LTO
tape drives. This only holds 880 tapes and from my calculations I will
need about 600 tapes plus enough tapes for a scratch pool. My concern is
that LTO cannot handle what our environment needs. LTO holds 100 GB
(native), but when a tape is damaged or needs to be reclaimed the time it
takes to do either process would take quite some time in my estimation.
Also, I was told that LTO is good for full volume backups and restores, but
that it has a decrease in performance when doing file restores, archives
and starting and stopping of sessions, which is a majority of what our
company does with TSM. Has anyone gone from a 3590 tape to LTO? Isn't
this going backwards in performance and reliability? Also, with
collocation, isn't a lot of tape space wasted because you can only put one
server per volume?
2. STK 9840B midpoint load(20 GB) or 9940A(60 GB) in our Powderhorn silo
that would be directly attached to the SAN. From what I gather, these
tapes are very robust like the 3590's, but the cost for this solution is
double IBM's LTO. We would also need Gresham licenses for all of the SAN
backed up clients(20).
Does anyone know of any sites/contacts that could tell me the
advantages/disadvantages of either solution? Any opinions would be greatly
appreciated.
Thanks!!!!
Joni Moyer
Associate Systems Programmer
joni.moyer AT highmark DOT com
(717)975-8338
|
