For the last several years, Red Hat has been pushing forward the development of real-time enhancements for Linux. Yet the company has made no formal product announcement of how it would attempt to productize its real-time Linux innovations. That changed today, with the announcement of the Red Hat MRG (Messaging, Real Time and Grid) platform. The product is expected to be available as a public beta this month, with a generally available release set for early 2008.
Novell came out with this a week ago touting how they were the first to do it. Now i read 80% of it was wrote by redhat employees the rest by wisconsin students. Not only that but Red Hat is releasing it as beta. They’d know its not ready since they’re the ones who wrote the majority of it.
Novell just keeps sinking further and further with me. Not only will i no longer recommend SuSe I wont even talk about them as a viable alternitive. I mean pulling that move of first to have it is one thing, but releasing beta code to enterprise customers just for news headlines? not a very safe company to invest your business in is it?
Suse has some stories of releasing beta code as “stable” just to look more atractive. They started using 2.6 before red hat did. They have been know to include cool external patches gratuitously just to look cooler. Fe, they included the CKRM resource management in suse enterprise server 9 back in 2004, because it was supposed to become the standard resource management infrastructure for linux… but 3 years later, only linux 2.6.24 is including a ckrm equivalent, because as many other patches, ckrm was considered a crap design and was rejected, and it was alpha/beta code so it was buggy anyway. But all that didn’t stopped suse.
Of course, it’s not entirely a bad thing. Red Hat is (or has been historically) the “big and serious” linux vendor. Suse/Novell has to attract customers aswell, so they try to release some features before any other important linux vendor. It’s good for open source – it helps to stabilize things.
Also, Novell’s support is (i think) cheaper than red hat’s. It’s your choice. It’s something that you can’t do with propietary software – choose between different vendors.
Edited 2007-12-04 22:13
Your comment reminds me of the “Pirates of Silicon Valley”
Steve Jobs: We’re better than you are! We have better stuff.
Bill Gates: You don’t get it, Steve. That doesn’t matter! [We got there first]
PS: more at http://www.imdb.com/title/tt0168122/quotes
http://www.techworld.com/opsys/news/index.cfm?newsid=10828&pagtype=…
“They haven’t contributed a line of code”, he said.
It’s kinda cheap for Red Hat to say you guys didn’t do anything to help just reselling our beta code but its also cheap of Novell to pull a snake move like that first of all. Red Hat put the money into developing this software so everyone can use it. least you can do is give them credit and not jump out first like hey look what we got!
I’d take it with a grain of salt. If Novell didn’t adopt the RH kernel extensions and used their previous RT kernel (remember, they’ve sold real time linux since last year) then they’d be accused of not supporting the community and doing their own thing. Their previous version of SLERT was a corroborative effort with another embedded linux company.
The interesting thing is that the head of Novell’s RT effort is the former founder of defunct Mission Critical Linux, who’s former CTO is now the CTO of Red Hat.
> a corroborative effort with another embedded linux company.
It’s “hard” realtime. It’s not the same realtime as Red Hat MRG provide.
The stock Linux already have some (hard) realtime feature.
Sometimes it seems like the commercial Linux vendors have tunnel vision that prevents them from thinking about the parts of the computing industry that lie outside the enterprise IT space. That’s where most of the real-time market lives, but RH and Novell want to shoe-horn it into their familiar environments.
The financial sector is a somewhat viable market for a real-time server platform. But the telcos will want to start with something lean and mean, more of an embedded deployment style. Grid computing is pretty much the opposite of real-time, the antithesis of deterministic performance, a trade-off of throughput over latency.
Of course, real-time is an incredibly important field of computer science, and Red Hat is the leading developer and integrator of real-time functionality in the Linux ecosystem. But it seems like the marketing department leaded some more meetings with the engineers to figure out how to package and market this technology.
Someone should have put it to them like this: Real-time means you get stuff done at very consistent rates, but only with well-behaved tasks running on under-utilized machines.
It won’t process stock trades at guaranteed rates if the transactions are sensitive to cache misses and/or page faults, nor will it do so if the machine has too many trades to execute. If either of these conditions are common, then the system will probably achieve higher overall performance (albeit non-deterministic) without real-time scheduling.
In my opinion, the embedded space is a much more robust market for real-time. Network and storage equipment, consumer appliances, manufacturing automation, industrial equipment, vehicles, etc. This is a good market to go after because each design win can drive huge revenues.
Another potentially compelling market for real-time Linux is multimedia production, but even given the huge strides in projects like Ardour, the availability of industry-standard applications like Pro Tools on Linux is a barrier.
I was at an expo last year and at the Novell stand they had a company called Concurrent http://www.ccur.com/
The company seems to have been around for some time. I understand that they are now owned by or work closely with Novell.
I don’t know about hard RT or anything like that but I they did demo their extensions to Novell’s Linux. My understanding at that time was that they did not actually rewrite or implement large portions of the kernel. The chief scientist in fact told me that most of the infrastructure was already part of Linux and they were merely providing nicer userland tools.
For example CPU interrupt shielding. This is a nice easy win on multi-core systems. You shield one or more cores from system interrupts so that these cores can get on with the primary task you set them (you can enable specific interrupts to these cores if they need to see them). They provide tools which makes configuring this easy – but he said the ability was in the kernel already.
The above may be a little faulty but I’m reporting what I remembered of my conversations with them.
The launch is not real-time. The Linux is. That makes it real-time Linux launched with a real-time-Linux launch.
From the title, I started thinking of a distro booted from the network and transparently up-to-date, zero-conf style. In the future, please refrain from dashing my dreams like this. Thanks.