Slashdot Mirror
Red Hat Gets Into The Clustering Biz
Bryan Mattern writes: "An adapted version of Red Hat 6.2 centered around Linux Virtual Server/cold failover is now available. The price is $1995, but includes a year of support for a node." This is also known as Red Hat High Availability Server 1.0.
I think we need more information to see whether this signals a lessening of RedHat's commitment to an open source corporate strategy, or whether they are executing the "traditional" open source strategy of selling support and configuration, or whether this is something yet different. It's definitely something to be looked at carefully, but I wouldn't get overly concerned until more facts are in.
No, not a flame, but I don't 100% agree it's the truth...
I feel that as the SysAdmin it's my duty to know everything about everything on my network.
Ideally, that is true. However, there are two factors to consider:
1. Scale. In a small network the "boss man" partitions the raid him/herslef. Sets up Legato all by his/her lonesome yatta yatta. Scale that up by a factor of 3 and the boss man gets a "subordinate" to set up Legato to his or her spec. Scale by another factor of 3 and the boss man gets a subordinate to cook up a 1 page proposal ("Amanda's nice, but we should buy Legato...") and then implement it. The bossman still knows what's going on, but actual implementation is diversified. Clinton doesn't train marines or cut Blue Cross cheques, but he's head of those departments.
This leads to two situations:
a) Bob knows Legato cold, Rachel doesn't. If it bombs on Rachel's shift and Bob is at the cabin/serving his weekend time there's a serious problem.
b) Sysadmin's get paranoid. We all work in an environment of trust and mutual respect for the capabilities of our colleagues (if you don't you should be reading dice.com). But the sysadmins job is on the line in the event of catastrophic failure.
2. Nobody knows the product like the manufacturer. Really. You call tech and they say "oh, that's a common problem." Common? It's the first time _I've_ seen it! Sun sees it 3 times a day. So, while I could use my well-honed trouble-shooting skills to narrow it down to a set of three potential causes and then methodically try the most-likely responses, logging results as I go, Sun knows you just have to wiggle the this and restart the that and you're up and running.
and I'm also getting 60k, then what the whole idea for me being there?
You're there to implement, configure, maintain and fix non-life-threatening problems for dozens and dozens of pieces of software and hardware. And make them all work together. How many servers do you have? How many pieces of high-performance software? Scripts? Kilometers of cat 5? Swiches? Make it work together effectively and reliably for your end users and you've earned your $60K.
2 1337 4 u!
Well, first off, if I wanted to have spares for everything in the shop it would cost more than the service contract. There are two things to consider when you asses your reliability:
a) The likelihood of failure
b) The speed to return to full capacity
Sun stuff is good for a, my story aside. This is the most important consideration seeing as even the fastest return to normal is slower than no failure at all. The drawback is that if you do have a failure, getting it back can be a drag. (Note: if you're not a Mac fan, go open up a G4. It's a tinkerer's wet dream). Try buying a frambuffer for a sparc10 from that internet cafe... hence, the service contract.
2 1337 4 u!
10:35pm cruching away on E3000. I hear a slight "pop" come from the SENA array. Watch in mild fascination/horror as the database drifts away from the front end like an un-hawsered boat from the pier.
10:36pm Panic. Check cables/power. Panic more.
10:38pm init 6 the whole thing. Call the end users (only 2... it's night shift). Act calm.
10:43pm system up. No data.
10:49pm call Sun (finding the card with the service number takes 5 min). Tell them my theory: "it's the gbic card". Sun's response: "we'll be the judge of that."
11:21pm Sun guy comes through the door. He's actually running...
11:58pm Sun guy stops working. Computer starts working. It's not the gbic cards, but he throws in a pair of new ones "just 'coz".
1:00pm I come in to work the next day because my ass is not fired.
The situation as it would be with that "free" rig
note, this is an untrue story...
10:35pm slight pop. no data. some screaming.
10:36pm assume it's the scsi card. remove casing. use compressed air to clear smoke away..
10:38pm scan room for compatible card to replace with. Find two ISA's used to prop up a short leg on a desk and Safeway bag full of SIMMs in the coat closet. screaming throughout.
10:41pm call Harold. He has lots of stuff. Mom answers and takes message. Find out later she relays the words "networth is down: need scooby cart".
10:51pm Debate "calling the boss man".
10:57pm call the boss man. "why don't you use the ones under the desk leg?" inform him their ISA. screaming from both parties throughout.
11:16pm have abandoned post. At "java bytes" internet cafe on 12th St. Offer pimply-faced teen large sum of cash for card from beaten Gateway in corner.
11:32 $200 poorer. case open. card in.
11:33pm it's the bus.
11:36pm surfing dice.com.
The bottom line is, if you think you're smart enough that you don't need that expensive contract you're too dumb to admin a network.
flame at will.
2 1337 4 u!
I was under the impression that you did in fact have to recode to make a process run accross clusters simultaneously.
Also, nt4/Win2k sever only supports two nodes, Win2 advanced server supports 4, Win2k data center server may support 8 (when Win2kDCS finally gets out). For the price of Win2k, you would expect much higher levels of clustering.
Also, check the cost. Win2kDCS costs more without support than RedHat's offer with support. Next check the machines. One of the nice points of clustering is avoiding large super expensive boxes. Win2kDCS has much higher minimum requirements than other Win2k/nt4 versions. While Linux lets you decide what to run things on. Scalability goes in both directions.
-- soldack
Because you're a big company sold on Office? Desktop reliability bites. Couldn't a contolled, monitored installation of selected apps work better when centralized? Hey - isn't the network the computer? I'm not trying to troll here - one can reasonably state that (a) Sun has a valid model and (b) M$ productivity software is hard to beat.
I wouldn't bet the company on that, either. But neither vendor is explicitly wrong. They both have long histories they're both betting on.
Of course, neither of them is RH, whom we're supposed to be discussing, and that's my fault. But where do you think they're headed with this?
Wherever the market takes them. (And where are the market leaders?) -j
I forget what 8 was for.
I believe the GPL makes source availability compulsory, not binaries. As long as the source for all GPL-ed stuff is available in some form, that ok. They aren't (AFAIK) required to provide a fully working binary installation set for you to download. Yes, its a shame but I think its legal..
I have seen a lot of posts on here about how it's free software and shouldn't be charged for. The GPL never says this. Instead it says you can charge whatever you want for it, however if you distribute a binary you have to make the source available for no more cost then that of the media to distribute and your time (basically.)
If I put together a distribution I can charge anything for it, say $100,000, but what the GPL protects is anybody from saying here is a bunch of binaries, if you want the code it is $100,000.
Just thought it should be mentioned.
This product is for people that can't, don't want to, or would rather spend their time doing business than recompiling their kernels.
$2000 for a support contract is nothing to a business betting the farm on a high-availability server. This product -- and, more importantly, this service RedHat is offering -- is for them.
Stop complaining about the price. As far as support contracts go, I can tell you this is pretty damned cheap.
- A.P.
--
"One World, one Web, one Program" - Microsoft promotional ad
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
So companies which sell products are to blaim when complete idiots go out and buy these products even when they could use something else (maybe cheaper) instead?
Thats a bunch of crap; no one is forcing these people to buy commercial products and if they are to lazy to even do a bit of orientating then they got no one to kick but themselves.
So... What you are basicly saying is that when I'm going out to buy a new car, totally unprepared, and end up with a Ferarri because I told the salesman that "I want to drive really fast to work" the salesman is to blaim for selling me this Ferarri? Yeah right...
The thing I like about this product is that it doesn't install any of the extraneous crap that often comes with Redhat. Yes, I know you could always tailor an installation to not include the junk. But this version comes with no extras as default. And that's a good thing. I don't really need all that stuff when I'm trying to run a high uptime, highly secure server.
--------- Beware the dragon, for you are crunchy and good with ketchup.
Businesses are in the business of making money, and as much as you or I might hate that fact, it's the absolute truth.
-=God Hates Me=-
I don't know about the others, but while SuSE only provides an ISO for the first CD, their FTP distribution includes almost everything on the 6 CD's except for the "pay" section (which includes mostly trial versions of proprietary software.) The non-US FTP version includes a lot of encryption software too. They release the full FTP version about a month after the CD version. Here is their mirror list.
It's for me. Just spec'd out a system that uses this, because it made my IT guys happy, since they're not big on the Linux thing. $2k for a year of support is nothing. Literally, nothing.
The servers that are in the cluster all cost > $40k, and the importance of uptime is critical. So the IT guys said 'could we have something in a shiny box' and I said 'hell yeah'. After all, if something fails at 3am, I sure as hell don't want them calling ME.
Besides, unless I could finish the whole design/code/debug process by myself, without QA, in under a week, it'll cost us more than $2k anyway just in my salary.
----------------------------
Pirhana is part of the standard RH 6.2 distribution. What you're getting here is a year of support; but more key to this, you're also getting consulting services: For your $2000, they will provide, via phone and email, the exact configuration information necessary to set up a node's worth of clustering on your network.
Comfortable with documentation and understand clustering already? Then get the $150 SSL-enhanced version of 6.2 and set it up yourself. Everything you need is there.
Don't need an RSA-licensed Apache-SSL in the node? Then grab Redhat 6.2 for free and configure Pirhana (which is included) at no cost.
Consulting costs money. If you don't need consulting, good for you. High-availability clustering is free.
Dunno about that point, I've got a copy of SuSE loaded on my system, and they have their own package that consists of all the mods they do to the kernel, mostly cool stuff like backporting 2.3's USB support, and putting in the Reiser filesystem support. All of their modifications are available in the source tree, so you can tweak their hacks to your hearts content. I'll have to check RH 6.2, Slack, and Mandrake (hey, I'm a distro whore, what can I say?)
Besides, from the other comments posted, it seems they're charging mostly for the right to blame someone else, aka tech support.
Marxism is the opiate of dumbasses
Actually, they are under no GPL obligation to make their software available on the net at all. The GPL only says that they have to make the source available anyone who gets the binaries, and that anyone who gets the source and binaries is free to redistribute them. IOW, you're free to buy the full price distribution and make all of the GPLed software available for free download, but you have no right to demand they make it available for free download.
FWIW, this is exactly what the Free Software Foundation advocates. To quote (from Selling Free Software on the FSF web site.):
(Emphasis is theirs.) IOW, the people who wrote the GPL don't just accept the idea of selling GPLed software for what the market will bear, they actually advocate it. Of course the fact that anyone who can get a copy of GPL software is free to redistribute it inherently limits the price anyone can charge, but that's for the market to decide.
There's no point in questioning authority if you aren't going to listen to the answers.
Picture this: on-site support, preferably by a brunette sitting on your lap and fixing the cluster... =)
.. you guessed it .. Natalie Portman! That's right! Natalie Portman supporting your Beowulf cluster! My, wouldn't that just be odd?
Picture this: on-site support from
But a $50 copy of Suse or CorelLinux retail will not get you a Red Hat technician on site for all support to guarantee 24/7 avalability for an entire year.
Regardless of whether the software is open-source or not, someone has to physically spend time to install, customise, tune and upgrade it otherwise the software suffers from bit rot. It then comes down to do you hire that person directly or outsource it? Either way you have to spend the money (not to mention usual overheads). Now if the profit margins for consultancy (*cough* Oracle *cough*) are high, then it makes sense to acquire such expertise in-house. However, if it is *NOT* a core competency, *AND* the risk/cost is lowered by hiring specialists (the whole point of a free market in services IMHO) then it makes sense to pay for the external support. It's the same reasoning that we don't all become in-house corporate doctors, the time it takes to learn the stuff and the frequency we use it doesn't make it worthwhile. Instead we pay for health insurance and scream for emergency help when disaster strikes. Even if you can view the base software as "free" (once development costs are written off), support costs is really a form of insurance that you will have some competent help (*cough* MSCE *cough*) around *WHEN* things go wrong (and I guarentee that error-free software has not been invented yet apart from 1-line hello worlds).
So, how much are you willing to pay for peace of mind?
LL
Count me one of those "stupid" people.
As a pointy-haired manager, I have to allocate resources as best I can. The most expensive resource I have to manage is labor, not machines or software. Worse, labor is not a situation where I can get all the labor I want at a given price.
I have a support staff of about 25 people, most of which are involved in end-user support. I have four full-time systems people with two vacancies I've been trying to fill for several months. To fill the gap, we had to hire an on site contractor from a consulting firm for $42/hour (dirt cheap actually) to do some of the chump work for the systems people so they could do more high-end stuff.
Now, with that in mind, and with the fact that we have several projects that are behind in, including purchase and deployment of a new SAN, roll out of Windows 2000, AutoCAD 2000, etc, etc, let me think.
If I need to deploy a Linux cluster, do I blow two grand to save time? Remember, two grand is not much more than a week of tech time on staff. Tech time that I can't afford to give up. So, $2K is really nothing.
Now, let me play the other end. To attract and keep the best talent, I have to keep them involved in what interests them. In this case, I'd ask my systems people if they want to do this on their own or to go the quick route. If they want to get down and dirty configuring it themselves, I'd actually prefer it, because that's the best way to truly understand a product. So in that case, something else has to give elsewhere, so I have to find out how to cover that.
(This is why I have the outside contractor help doing low-end stuff. If I hired an outside firm to do our higher-end stuff and stuck my loyal employees on mundane every-day shit, they'd all quit...)
There are so many variables to consider when making decisions on how to best manage resources. I certainly appreciate that Red Hat gives pointy-haired (actually, my hair is 3/4 down my back!) bosses like me choices like this, and they get to make money at it too!
Capitalism is the best!: )
I come from a Digital UNIX background [*], and I've been fortunate enough to use DEC's TruCluster product. True, it does require "better" hardware, and more support dollars - but you're not about to set up a production failover environment in your bedroom to "play" with it either.
Linux HA is still missing some of the major features of commercial clustering packages. These can be (and are, for me) showstoppers to using Linux in an HA environment.
The most notable, in my mind, is:
Shared disk. This doesn't mean shared filesystem (although TruCluster v5 is apparently approaching this). This means that all cluster nodes have a scsi (or fibre channel) controller on a shared bus. This bus is used to access the filesystems, but also for a non-network based inter-node communication method. When I see a TruCluster node boot, along with all the disks, I see "processor at id 2", "processor at id 3", corresponding to other nodes. In this manner, nodes know each other are up, even if the network "blips".
...and to a lesser extent (ie: I run without this now)
Shared System.TruCluster v5 introduces to the UNIX world what VMS has had for over a decade - the shared system disk. Each system uses the same system disk. I don't mean an identical copy, I mean the *same* disk. (Where disk can be, of course [and should be], a hardware mirror/RAID set). Node-dependent data is kept separate by maintaing private node config directories, which are referenced using "context dependant symlinks". CDSL's are essentially symlinks with variables in them.
..anyways, I've said enough. If you're truly interested in this sort of thing, I invite you to check out: http://www.unix.digital.com/cluster/in dex.html, which is the TruCluster site.
This is the sort of stuff I'd like to see on Linux. I'd help code it but, unfortunately, I have neither the time nor the skill (yet).
I am not, nor have I ever been, an employee of DEC/Compaq - but I have used their products in an enterprise class environment for both user (www, mail, ldap, etc) and database (oracle) services.
-Jeff
<ducks in the corner as the flames rise>
[*]: But I have been using Linux on my personal system since v1.2.8 (walnut creek slack aug 1995)
I also assume that they have *tested* the stuff on the CD so you won't do an install and find out you have to upgrade 10 different aps.
Go Badgers! -- #include "std/disclaimer.h"
Damn, that must be the narrowest definition of "us" I've seen in a public forum. Is that "myself and the other 11 superelite uebermenschliche sys admins with 30 years experience scattered through the world in our mountaintop fortresses"?
I'm guessing it would take me a month or more to set up a high availability server. The second time I might be able to cut it down to a week. I don't think there are nearly as many gurus out there as you think. The last sys admin where I work now left because he got a better offer, but there are a lot of things in the system (cron jobs that fail every night; a web server named "mail", a mail server named "bkp") that make me he probably wouldn't be able to set up failover very easily.
Most Linux users, even experienced Linux users, don't know any more than how to keep it running, and how to set up what they've set up before. Being a Linux user doesn't mean one is an expert. Considering a graph of skill level for different OSs, Windows would have a nice broad distribution, centered around "knows how to do daily operations, but can't figure out new stuff on their own". Macs would be an even flatter distribution. Both more experts and more clueless newbies. Linux would be a much narrower distribution -- newbies *can't* use Linux, and to become an expert takes significantly longer than on Windows. But the average would be a little higher.
I'm thinking, what percentage of Linux users could do something as simple as have find delete all file in the home directory that hadn't been accessed in more than 90 days? How many would know where to look? How many would be able to understand the find syntax or know that they need to escape the semi colon? Why do you think pico ships with every Linux distro?
If you think about it, though, if you hire someone who could do this right, you'll probably be paying them about $50/hr as an employee, $75+ as a contractor. Figure a week to set it up, and that's $2000-$3000. And no support.
Of course, the disc will probably be on cheap bytes for $0.99 in a few weeks. Use that, and figure it out as you go.
--Kevin
$1995 is chicken feed. It'll cost you at least 10 times that for Sun Cluster - with no support (but including "knowledge transfer" for the initial installation.) Not that that's unusual for Sun, but it still made me choke when I saw the invoice... If RedHat (or whoever) can offer an equivalent product to Sun DiskSuite/Veritas VolumeManager with similar price advantages, Linux servers will be very saleable.
Red Hat also makes PowerTools available, i.e. not just the basic system.
Anyone good any good ideas what you're really paying for in this? $1995 sounds like quite a lot for a year of support... I'd sure hope the support services were somehting VERY extraordinary
Picture this: on-site support, preferably by a brunette sitting on your lap and fixing the cluster... =)
That's nice, but what does it have to do with the OS???
We have Linux running on various Compaq Proliant in our server room. Aside from the fact they have hardware redundancy (which your scenario doesn't really account for), we do have the hardware contracts with Compaq, so yes, we do have Compaq guys running if something goes bad with the hardware.
Again, nothing to do with Linux, nothing to do with RedHat. It's hardware issue! That's why you get away with using Sun in your example, Sun really sells you the hardware (software is a token, they're even considering making it free, if they haven't already).
If the server goes waaay down (assuming you know your stuff - might be a big assumption), there is nothing the tech guy can do that you can't do. All the tech guy will do is get your server "running" again. The data on it, he can't do any miracles either if it's trashed beyong belief. Backups anyone?
alex, rhce (acronyms are cool).
If you don't want to pay that much, just get RedHat 6.2. It still comes with the Clustering software.
As a corporation, I can sell a cluster server with 24/7 support a heck of a lot easier than a cluster server that usenet people will help me support.
To a corporation, $2k is nothing, I spend more than that on support for the server we use for testing patches on.
-- Keith Moore
This sig is the express property of someone.
1: Everybody and their dog knows linux
2: No per-seat licensing (as said above)
3: This is cheap for what you're getting
4: The target customer usually has millions of dollars to spend on IT (i.e. fortune 500, US Govt., etc)
People always try to defend linux when it comes to the whole "Linux has no support", but really - this is bottom line guarantee _we_will_fix_the_problem_. You don't get that through irc or usenet. $2000 for onsite tech support for large systems such as this is beans.
So will hobbiests buy it? No, of course not - they'll figure out how to do it themselves. But enterprise-level networks need this type of stuff, and have the money to burn.
Those who can't do, teach. Those who can't teach either, do tech support.
... to rip the shit out of one-eyed Linux zealots, usually identifiable by phrases like "wind0ze", "m$", etc, BUT Windows 2000 is NOT a mature product.
Open Source. Closed Minds. We are Slashdot.
I have seen a recurring trend among the top linux distributions, which upsets me. Try to download a distribution other than a very basic one, and the companies make it difficult to impossible. Suse lets you get one of the six disks. Redhat gives you only the standard distribution. Turbolinus gives only the workstation install. I am not sure if this is exactly legal. If the distribution is using gpl software, doesn't it hve to be made available? Please respond, I am confused.
-- the computer doesn't want any beer, no matter how much you think it does. NEVER, EVER feed your computer beer.
This story strikes me as having some relevance with the story about Linux in Beijing that was posted last week. On first blush, this certainly seems like a "more choice, victory for Linux" type of situation, but now I'm not quite so sure.
We all know that the Chinese government would not trust any security-critical systems to Windows: the code could have easily been compromised by the NSA or the CIA or some other arm of the US government. This is the commonly-accepted reason why the Chinese government is pushing the acceptance of Linux (in addition to the fact that they like its fundamentally communist development model).
But think a little further: the US is desperately trying to put it regional missile defense into place, and China is ramping up its anti-US sentiments and dissident crackdowns even beyond their usual extremes. All this can only show that there is some incredible US / China tension building behind the scenes, which could easily escalate to war within a few years.
And when this war comes, do we really want China having the benefit of high-availability Linux servers? Do we really want to eliminate the chances of an accidental server crash that could end up saving the Western world? I think not. Letting the forces of communism run rampant on this planet is about the worst fate I could imagine for any of us.
That's bad enough. But now look at, first, Red Hat's timing of this launch (which I've already explained the significance of), and, second, Red Hat's name. RED Hat -- if that's not a sign that they're in league with the PRC, I don't know what is.
So, please - I urge you - for the world's future - boycott this product. When the shooting starts, you'll be glad you did.
Three cheers for open source companies with viable business models.
I hate bigotted people who misunderstand what the word free means in free software
Ahem. FLAME ON!
I know all you moderators out there aren't going to be happy with this, but there are a lot of people out there, including me, who (omigod) think that free-as-in-beer is just as important, if not more, that free-as-in-freedom. I'm not saying that I'm an uberhacker, but I have learned quite a bit in the 10 years that I have been coding. Much of that knowledge has only been made possible through free-as-in-beer software. I 'grew up' as a programmer while in high school and the first years of college, when money was scarce. I didn't exactly have Kdollars lying around to buy operating systems and tools. And then I see a FreeBSD CD-ROM set at Hastings for around $17 and it says something to the effect of "Full-featured UNIX operating system, free compilers, free networking software, 1000's of free applicatons (and I mead beer). Everything you need to start soaking up that knowledge you so greatly desire. $17 for 4 CDs. HELLO!" If that CD had not been there, my life would be totally different at this point.
I'm not saying that the freedom issue isn't important. Sometimes we just get so caught up in the politics of it all that we forget how we got where we are.
Washington, DC: It's like Hollywood for ugly people.
Let me tell you, $1995 is peanuts for failover.
We are in the process of installing a cluster/HA failover solution onto a pair of Sun boxen here. The software alone costs $105,000.
Heck, even the training Sun requires the local admin to have before the installation is $3495.
True, the RH is likely to be very basic compared to what is available for the large enterprise boxen, but I expect that to change rapidly if RH can get the ball rolling.
In the immortal words of Socrates, who said; 'I drank what?'
Actually we DO release it. ;)
We don't believe in proprietary software any more than Debian does.
The clustering tools are released under the terms of the GPL.
This message is provided under the terms outlined at http://www.bero.org/terms.html
Straight from the news item:
Support: A one year support package that includes standard hours installation and configuration assistance and 24x7 server-down support for two Linux Routers of a Piranha cluster, configuration of Piranha and Linux Virtual Server and any services required to run these applications, including httpd and ftpd, and configuration of Piranha and Linux Virtual Server for connectivity to one node behind the Linux Routers for either httpd or ftpd service.
For a business or organization, $2000 really isn't that much, especially when it gives management the right to point the finger at another company when something goes wrong...hence the importance of "Tech Support".
Please, if you don't like Linux, that's fine, but don't try to troll your way to infamy.
-=God Hates Me=-