Letters to the editor

The Unfortunate state of evms.

Dear sir,

LVM has been available for 2.4 for some time. It works
but is very painful to use.

The IBM sponsored EVMS subsystem ( http://evms.sf.net ) has changed
that.

Evms provides a wonderful front end to a very powerful kernel
back end. At last tasks such as:

Expanding volumes and filesystems while mounted.
Shrinking volumes and filesystems.
Raid levels
Snapshotting

and more are not only possible, but positively joyous to perform.

Evms is standard in the gentoo distribution but not mandatory.

As the 2.5 feature freeze approached, tension mounted.
There were many things queued for Linus, evms was one.

It didn't make it, device mapper 2 (DM2) was merged instead.
The reasons are technical ( ie. I don't understand). However, 
evms replaced rather than worked with existing kernel subsystems.
That didn't go down too well.

The evms people took stock of the situation and decided to
change direction. Instead of replacing DM2, they would use DM2
to implement evms functionality. 

From the users point of view, its the user interface of evms thats
so great. The internals are irrelevant.

I congratulate the evms team on their coolly competent analysis
and response to what must have been a body blow.

This leaves us users in a quandry. I have a neat system with evms.
But I cannot use the 2.5.X kernels with it! The latest patch was against
2.5.46 and with the new direction, future compatibility may not be
possible.

Evms is stunning, but until we are ready to make the move to 2.6.0
I cannot recommend using it.

I would like to suggest to the editor that monitoring the development
of evms would be a valuable service.

I think everyone should use evms, but not perhaps, just yet.

Yours,

John Huttley
New Zealand

Comments (2 posted)

Public relations and journalists, again

  Hi all,

  It's the second time in a few days that I've seen an article
speaking about Gforge as being the only real Free clone of
Sourceforge.  This one was on LWN[1].  While I appreciate the
publicity that makes for free software, it's twice in a few days I've
seen incorrect info posted on a news site.

  I have therefore written a rant^Wcomment on the latest one, where
the status of the forks is explained.  So that the next article about
Gforge or Berlios (or whatever the next big one is) turns out not to
forget the people who have worked for long hours.  I tried to keep the
tone cool, but I was rather angry when I first read it.  Maybe I
expect too much about journalists or reporters, but checking their
info before posting would seem natural to me.  I understand that Tim
reviving the code that his own company made proprietary made for a
good news article, but if you're going to mention others (which you
should), just get your info accurate.

  Let me restate quite explicitly: there is no competition between the
different forks.  The Savannah people wanted something adapted to the
GNU/FSF servers, we Debian-SF people want something that Just Works,
the Gforge people (well, Tim actually ;-) wanted to clean up the code.
Fine.  Gforge and Debian-SF are trying to merge, we might get some
code from Savannah, they might take some for ours.  That's the rules
of the game, we accept it quite gladly.

  End of rant from my part.  Please read my comments on the LWN
article[1].  For your penitence, Steve and Rick, you're condemned to
wait for a month or three and then write an article about how Gforge
and Debian-SF have merged and are now the best thing since sliced
bread, and I'm Cc:ing LWN so that they can link to this forthcoming
article when it's out :-)

Roland.

[1] http://lwn.net/Articles/17369/
-- 
Roland Mas

Late frost burns the bloom / Would a fool not let the belt / Restrain the body?
  -- in Good Omens (Terry Pratchett and Neil Gaiman)

Comments (none posted)

Re: Public relations and journalists, again

Roland, since I don't (yet) have a subscription to LWN (I know, I
know!), I haven't seen what they wrote.  I _do_ know what I sent in to
LWN's letters column, and I did pretty clearly state (I think) that
Debian-SF and GForge are aiming towards merger.  I _didn't_ in any way
claim GForge is the only Free clone, and I _did_ attempt to list (but
not denigrate) the numerous forks that people have worked so hard to
keep moving forward.

I was of course trying to be brief (this being a letters column, not an
article), and summarise information from diverse sources before LWN's
then-pending publication deadline, in a hurry.  I _hope_ my comments
were generally inaccurate, and tried to ensure that.  If I've
inadvertantly given offence nonetheless, my apologies, as I was trying
carefully to avoid giving any.

-- 
Cheers,                            "To summarize the summary of the summary:
Rick Moen                           People are a problem."
rick@linuxmafia.com                                       -- Douglas Adams

Comments (none posted)

Scalability testing is not just a kernel issue

Scalability testing, especially scalability to the large, is not just a kernel
issue. Suppose a kernel patch reduces the latency of a TCP connection to under
a microsecond and achieves 95% of the theoretical bandwidth using TCP---very
unlikely in my opinion. This is purely accademic if an absraction layer eats
10ms per packet. Some popular implementations of MPI are known to be more
expensive than the current linux TCP implementation.

As the author of the mpkern task parallel programming library it would be nice
to be able to see how the scalability changes as the library evolves and
continued access to a cluster would obviously be useful for this testing. Maybe
STP is not the appropiate vendue but I would be interested in somewhere to do
this on an occasional basis. At present the mpkern library has only been tested
in anger on a linux cluster and a code freeze has been declared.

P.S. The mpkern announcement on comp.os.linux.announce of a few days ago failed
to mention that you can obtained mpkern at http://www.sourceforge.net/projects/
mpkern. For more information see the announcement.
-- 
Duncan (-:
"software industry, the: unique industry where selling substandard goods is
legal and you can charge extra for fixing the problems."

Comments (none posted)

Re: Think of Our Kin Overseas

	I am disabled yet I had to get a part-time job because my disability
retirement isn't enough to live on. I'm also a single parent of a young
child. We make it one paycheck at a time.

	I say this because, while money is extremely tight, I believe in
Software Libre and the work that LWN is doing so much that I will take
Mr. Myers up on his suggestion and offer to pay for subscription(s) for
some non-US techies who would like one. Email me if you are interested.

-- 
"Khaaaaamaaayyyy, Haaaaamaaaayyyy, HAAAAAAAAA!!!!!"
                -- Goku, 'Dragon Ball'

Comments (1 posted)

[Kevin McIsaac] the true value of Linux

  http://techupdate.zdnet.com/techupdate/stories/main/0,14179,2901102,00.html

This article is full of furphies and `damnings with faint praise.' I'd be 
unsurprised to see faux pas in the talkbacks, but in the article itself it's 
obvious that Kevin is a stranger to Linux and writing `at arm's length'.

> This is based on the flawed assumption that because Linux is "free" it
> will reduce TCO.

That assumption is not flawed. Linux really can be had for $0 (try
http://www.linux-mandrake.com/en/ftp.php3 or http://www.debian.org/distrib/) 
and this does indeed reduce the TCO.

> On closer inspection, it appears the recommendation is more an
> emotionally driven reaction against Microsoft than a factual case for
> Linux.

Study after study based on real researched facts and not opinionated 
pontification or financial incentives draws the conclusion that it's 
typically a financially driven reaction against Microsoft and taken after 
much hang-wringing and planning.

The second-teir method of Linux introduction is technicians who've had it up 
to the eyebrows with fancy and pretty systems which - for a variety of 
reasons - fail constantly or can't reasonably be made to do the assigned task 
in the first place.

> Astute IT organizations will recognize that Linux's true value is
> derived more from the price/performance of the commodity Intel
> hardware it enables than from its open source characteristics.

Astute IT organisations won't rely on opinion and unsupported projections, 
they'll either do their own research or look for original research which 
includes hard figures. Having done that, they'll notice a few crucial things 
which go totally unmentioned in your article, and presumably the Meta study.

One of those things is that a Linux sysadmin will typically shepherd at least 
four boxes for every one a Windows sysadmin shepherds, function for function. 
IDC missed that one, http://www.ibm.com/linux/RFG-LinuxTCO-vFINAL-Jul2002.pdf 
shows that RFG didn't. Needless to say, it makes a complete mockery of your 
figures if you don't factor it in.

> nor have many clients embarked on major Linux projects outside of Web
> server farms, appliances (network-attached storage), or general
> infrastructure servers (e.g., DNS and DHCP).

Um, factor in email service and (a key and common task for Linux servers which 
you seem to have missed; rolling it into `general infrastructure' doesn't 
seem appropriate) and about the only major sectors you've really got left are 
databases, groupware or application servers.

Oracle is working on the first. `Less money on OS == more left for Oracle,' a 
fairly straightforward equation; plus `more reliability == Oracle looks 
better' gives you about all the motivation you'd need if you were Oracle.

Meanwhile, back that the fairly lengthy list of functions Kevin drew, the 
reason that Linux is being used in those areas is because you can just plug 
it in and forget it. As people try this out and see it for themselves, 
they'll also trust it for their databases and other traditionally `big iron' 
applications.

> The Linux OS license is "free," but that does not ensure that total
> cost of ownership will be reduced.

Yes, it does. The point you should be making is that this reduction is not the 
be-all and end-all of TCO. But at each component of TCO you examine, it gets 
better for Linux.

> Even if all other Linux costs were the same,

But they're not. They're *all* lower.

> It is only when other significant pieces of software can be licensed at
> little or no cost (e.g., office suite, e-mail, and DBMS) that TCO
> reduction is at a level significant enough to merit the additional
> complexity, risks, and potential cost overruns of Linux.

This one really gets under my skin. Why are you citing `complexity, risks, and 
potential cost overruns of Linux'? The potential for cost overruns exists 
with every OS, and in particular Windows is well known for doing the 
unexpected. Singling out Linux for mention in association with `cost 
overruns' is a cowardly way of talking it down.

Be a man, explain why Linux in particular should be especially susceptible, or 
print a prompt retraction!

Meanwhile, the office suites (plural), email and databases that you will find 
on your $0 copy of Linux (see above) are all $0 themselves. Linux may *look* 
complex to someone who sees a shell promtp and wets his pants, but the design 
is more orthogonal, more systematic and more predictable than Windows.

You can also overlay it with a GUI and WYSIWYG management tools that are far 
prettier and more consistent than Windows, thereby keeping your pants dry.

> The key attractions to Linux are:
> Royalty-free distribution

Again, you significantly undersell the point. No more licence tracking, no 
more BSA nightmares, no more worryig about what employees take home and 
install (or upload) using work's activation keys.

> Access to source code:
> All versions include source code, making Linux compelling for
> technical staff.

And again, you significant undersell by limiting the appeal to techies only. 
Users and management are often overjoyed that their techies can quickly 
tailor their $0 software to exactly suit their needs.

> High levels of reliability:

Give with one hand...

> Although this was compelling compared to NT 4, increased stability
> of Windows 2000 has narrowed this gap, making this less of an
> advantage.

...take back with the other. Linux is still an order of magnitude less flakey 
than either Windows 2000 or Windows XP, especially should you (ghasp) venture 
away from Hardware Compatiblity List gear.

> Linux is still missing native high-availability features such as
> journaling file systems or clustering

Now this, this is a flat lie!

Linux has *four* native journalling filesystems: ext3, XFS, JFS, ReiserFS and 
on top of that can use Windows' own journalling filesystem, NTFS.

Google for the term `Beowulf'. You'll learn a number things. The first is that 
you just used a huge Linux cluster to do your search, the other is that Linux 
clusters are bigger, better and badder than Windows clusters and have been 
for a long time. How many Windows-based supercomputers are there? None. Yet 
the 5th fastest (2nd fastest if you take peak values) computer in the entire 
world is a Linux cluster! Missing clustering?

> Linux has its place in the data center, but it is not a silver bullet
> for Windows.

It's `place' is as a silver bullet, a bundle of oaken stakes and a whole 
coffin full of garlic, Kevin.

> Where should I use Linux?
> In an appliance where the OS is not exposed

The City of Largo has 450+ Linux desktops, Kevin, and a lower IT spend by 60% 
than their municipal neighbours. Is that exposed enough for you?

> Intel servers are widely used for scientific computing [...] Although
> it is possible to use Windows in this application, many Unix-centric
> organizations will be more sympathetic to Linux and will find the
> skill transition much simpler.

Kevin, not only is Windows a dead loss in a compute farm, but you just 
contradicted your previous piece of advice! A scientific computing node is an 
appliance; the OS is not exposed to the operator!

> As a general-purpose infrastructure server (e.g., DHCP, DNS, or POP),
> where solid reliability is required but high availability is not.

Ah, that would explain Linux's recent uptake by telcos, then. Sarcasm aside, 
Kevin, how many industries have tougher HA requirements than telcos? 
Military, medical and space. Linux is used by all of those, too.

> On the other hand, Linux should generally be avoided whenever there
> is a requirement for single-image scalability above four CPUs (scale-up)
> or high availability based on OS-level clustering.

Remember that mention of faux pas?

SGI will sell you a single-image 64-CPU Itanium-2 system running Linux 
(http://www.sgi.com/newsroom/press_releases/2002/september/stream.html) if 
you ask them. And we've already been over clustering.

> Can I use Linux to replace Windows for file and print?
> Although this is possible using Samba [...] it is not recommended.

Ditch Active Directory and the specters you raise flee into the night, along 
with a host of other problems. Samba has a number of ways of seamlessly 
integrating with Windows domains. And if you want to keep AD, Samba 3 works 
now.

> A switch to Linux for file and print might lower purchase costs, but
> it would seriously affect the ease with which users can access the
> services as well as increase management complexity, thereby driving
> up the total cost of ownership.

Since in practice the use of Linux dramatically drops the requirements for 
administrator intervention, and contrary to recent rumour integrating it 
seamlessly is a straightforward process, it actually drives *down* the TCO 
significantly.

> Business impact
> Inappropriate use of Linux as a Windows or Unix replacement will weaken
> the IT infrastructure and reduce its business value.

Appropriate use of Linux, which means in just about everything, will on the 
other hand strengthen the IT infrastructure and free IT staff to concentrate 
on more important issues than managing servers which should be acting like 
appliances but aren't.

Inappropriate use of Windows is a rolling disaster.

> Bottom line
> Organizations that allow emotional reactions (e.g., against Microsoft)
> to drive decisions to replace Windows or Unix with Linux will fail to
> achieve anticipated savings, and will end up with an infrastructure that
> is limiting and difficult to manage.

True at face value, and I wouldn't complain except that you added a short 
burst of Latin and two words in English. Specifically, `exempli gratia, 
against Microsoft'.

You see, Kevin, an emotional reaction against Linux, or more pointedly against 
anything *but* Microsoft, is the single most common cause of people 
continuing to use Windows inappropriately throughout their IT structure, in 
places where Linux would be ideal.

The amount of wastage and damage that this causes worldwide could probably 
power several of the smaller African nations if you were able to recover it. 
Perhaps that's why Linux is taking off throughout Africa, as exemplified by 
SchoolNet Namibia (http://www.schoolnet.na/).

Nevertheles, I almost agree with Kevin in one point: cost is a long way from 
the most important reason for adopting Linux and OSS applications in place of 
Windows and lock-in-ware, as he would discover through imbibing some 
real-world experience instead of quizzing a bucketful of Windows admins and 
PHBs. There are many better reasons for rolling out Linux, to do with things 
like stability, flexibility, control, localisation, security, auditability, 
standardisation, manageability, reliability, and a whole host of other 
abilities unique to OSS (some of them unique to Free Software).

Better luck next time.

Cheers; Leon

-- 
http://www.cyberknights.com.au/  Modern tools, traditional dedication
http://slpwa.linux.org.au/       Member, Linux Professionals West Aus
http://conf.linux.org.au/        THE Australian Linux Technical Conf:
                                 22-25 January 2003, Perth: be there!

Comments (2 posted)