There was an embarassingly bad TCP window size scaling bug in RHEL 5.4. It
wasn't acknowledged in any way by RedHat, until late in 5.5, and wasn't
fixed until 5.6.
I faced long and continued skepticism from the network people, and the
Linux admins, that such a bug could exist in a RHEL release, that I would
be able to discern such a bug, or that it wasn't a TSM problem. (It took
about six months to resolve that particular problem, and would've taken
even longer if we'd have started looking at it sooner.)
Only a small percentage of Linux boxes get their network cards barraged
with heavy-heavy receive traffic all night every night. So bugs like that
don't get discovered quickly, or fixed quickly, or documented quickly.
(This one affected other brands of backup software too.)
Also, there is at least one brand of inexpensive 10GBE cards that are
complete garbage. They overheat under heavy use, and typically cause the
whole system to crash. (For political reasons, we abandoned 10GBE on the
backup servers.)
These cards are sold under three or four different brand names, and were
the ones sold by the the top 3 big Intel box makers, because they are the
cheapest cards available.
I'll stop here, before I get into editorializing.
[RC]
shawn.drew AT AMERICAS.BNPPARIBAS DOT COM
Sent by: ADSM-L AT VM.MARIST DOT EDU
04/20/2012 09:38 AM
Please respond to
ADSM-L AT VM.MARIST DOT EDU
To
ADSM-L AT VM.MARIST DOT EDU
cc
Subject
[ADSM-L] AIX vs Linux 2012
I know this has been discussed in various forms over the years, but I'm
specifically wondering about the current state of hardware
I have a long history with TSM on AIX. It's stable, familiar and an I/O
powerhouse. Our Unix admins also favor AIX for serious, heavy-duty
workloads.
We are looking at refreshing our largest P570 now. I discussed this with
our unix admin, who also has a very high opinion of IBM. He said that
current SandyBridge implementations can really make Linux a contender in
terms of I/O and CPU performance. And at about 1/7th the cost.
I normally dismiss Linux because I was under the impression that you would
need many inexpensive servers to equal one P-series for I/O. It wouldn't
be worth it with the added management of dealing with multiple TSM
Servers. Now with DB2, TSM seems to be going more towards the monolithic
direction if anything.
But If I can get a single 32-core, 128GB ram intel server that can
actually push multiple 10gbe and 8gb FC interfaces I am finding Linux a
little more attractive.
Does anyone have any stories, gotchas, or opinions with replacing a
P-series host with a modern Intel system 1 for 1?
Regards,
Shawn
________________________________________________
Shawn Drew
This message and any attachments (the "message") is intended solely for
the addressees and is confidential. If you receive this message in error,
please delete it and immediately notify the sender. Any use not in accord
with its purpose, any dissemination or disclosure, either whole or
partial,
is prohibited except formal approval. The internet can not guarantee the
integrity of this message. BNP PARIBAS (and its subsidiaries) shall (will)
not therefore be liable for the message if modified. Please note that
certain
functions and services for BNP Paribas may be performed by BNP Paribas
RCC, Inc.
If you are not the intended addressee, please inform us immediately that you
have received this e-mail in error, and delete it. We thank you for your
cooperation.
|
