Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
XGrid Officially Dead
- Thread starter FluJunkie
- Start date
- Sort by reaction score
While I personally never used Xgrid or worked anywhere that did, as both a scientist and a Mac user, I find this deeply troubling. Abandoning tech without warning makes Apple unreliable for research. At least we were warned about Classic and Rosetta, both of which were clearly meant as transitional solutions. 
I can see no reason for this abrupt descison, especially when there appears to be no plan for a replacement.
Apple may never win enterprise, but it has a STRONG presence in academia, especially among students. Why not make this an Apple stronghold? Why not sell OS X as a "High-performance *NIX for the rest of us". . . like they used to . . .
Oh, and the lack of updates to Apple's Science Page, which still features XServes and makes no mention of the death of Xgrid, is truly appalling and should be considered false advertising . . .
I can see no reason for this abrupt descison, especially when there appears to be no plan for a replacement.Apple may never win enterprise, but it has a STRONG presence in academia, especially among students. Why not make this an Apple stronghold? Why not sell OS X as a "High-performance *NIX for the rest of us". . . like they used to . . .
Oh, and the lack of updates to Apple's Science Page, which still features XServes and makes no mention of the death of Xgrid, is truly appalling and should be considered false advertising . . .
Last edited:
I don't think you need to worry at all. The mentioned alternatives are very good, maybe even better than Xgrid. Boinc is something that is cross platform (you can use it on Linux, FreeBSD, Windows, OS X, etc.) and is used for quite a lot of scientific programs such as the protein folding projects. Using Boinc makes more sense because you reach a bigger audience and thus have more machines that will do things like folding proteins. It also seems that tools like Apache Hadoop are becoming a standard in grid computing & distributed computing. When I look at what universities use mostly for grid/distributed computing I also see other solutions. Some even use computing power from a 3rd party or build something together (teamwork is a big thing in the science world).
Yet another step in the transition from "Apple Computers Inc" to "Apple Toy Corp".
No wonder that other solutions become comparable alternatives. That what happens when there is no R&D going into a product for several years.
No wonder that other solutions become comparable alternatives. That what happens when there is no R&D going into a product for several years.
The point is not that there aren't (better) alternatives, which I know there are. The point is that it makes the Mac as a platform look bad. Those people who were using Xgrid now have to scramble to find a compatible alternative, which is disruptive to their workflow. At little advance warning from Apple would have been nice. If it were me and this cost me research time, I might not ever buy a Mac again, at least not for this purpose.
OS X Server is joke and has always been a joke. I'd rather Apple focus on improving Exchange and AD support than continue to waste their time on their own server option. Integrate all the OS X Server features that are useful for home and small-business networking into mainline OS X and call it a day.
Of course, academia is a different story, and I stand by my earlier statement that Apple might find more success there by selling OS X as a "*NIX for the rest of us," competing with Linux-based solutions rather than Windows-based ones. This is something Apple used to do, but, apparently, it has fallen out of vouge since M$ recovered from the Vista fiasco and iOS became Apple's major focus . . .
There are MANY fields in which Windows is the platform of choice, not because there is something special about Windows but because the requisite professional-grade programs in the field are Windows-only. This is what leads to the "Mac as a toy" perception. OS X cannot compete here. If these people want to use Macs, they have to use BootCamp or some virtualization solution to get their work done. I remain convinced that BootCamp is primarily responsible for the increased success of the Mac post-Intel transition.
Last edited:
I'm not so sure about that. OS X is first and foremost a desktop OS. They have never been successful in the server area nor in the hpc area. It is better if they recognised that and create tools that make their great desktop OS work great with all the other tooling and software out there.
The people who now use Xgrid can still continue to use it because it still works. The only problem is that there won't be any new version so they have to look into alternatives if they haven't done so already (like I said others are able to do a better job, especially when you have machines with different operating systems; there simply have been better options to Xgrid for quite some time now). It might mean that some will be using the current setup for a prolonged period to transition from the old environment to the new and/or to have more time to move to an alternative. It is definitely not like they have to scramble. Most of these kind of environments are used for 5 or even 10 years. In that time the only thing they do is make sure updates are rolled out. Updates and especially upgrades have a big impact on these kind of setups so they tend to be more careful with rolling those out mostly resulting in not ever doing upgrades, only updates. But it does depend on the setup and who manages it.
The people who now use Xgrid can still continue to use it because it still works. The only problem is that there won't be any new version so they have to look into alternatives if they haven't done so already (like I said others are able to do a better job, especially when you have machines with different operating systems; there simply have been better options to Xgrid for quite some time now). It might mean that some will be using the current setup for a prolonged period to transition from the old environment to the new and/or to have more time to move to an alternative. It is definitely not like they have to scramble. Most of these kind of environments are used for 5 or even 10 years. In that time the only thing they do is make sure updates are rolled out. Updates and especially upgrades have a big impact on these kind of setups so they tend to be more careful with rolling those out mostly resulting in not ever doing upgrades, only updates. But it does depend on the setup and who manages it.
XGRID help...
Hi All,
the term "newbie" does not even start to picture what i am in terms of grid computing! however, i recently started using FDS, a software used for the assessment of smoke flow and hot fluids (see: fire.nist.gov 'Fire Dynamic Simulator aka FDS-SMV).
Having recently come across xgrid, i thought i would try to use the grid to increase my processing power using the aggregate Ghz forces available on my private network.
the problem is, i have no idea how to get this software to start using the xgrid functionalities.
so here we go: is there someone out there that knows about FDS or would know about how to replicate (for lack of a better word) the parallel processing options available for FDS on standard IBM-compatible machines?
Thanks for the help.
you can send me your thoughts both on this forum and on lricher@hotmail.com using xgrid in the subject field to avoid the message being sent to the junk-oblivion folder
Again, thanks so much for the help.
Lau
Hi All,
the term "newbie" does not even start to picture what i am in terms of grid computing! however, i recently started using FDS, a software used for the assessment of smoke flow and hot fluids (see: fire.nist.gov 'Fire Dynamic Simulator aka FDS-SMV).
Having recently come across xgrid, i thought i would try to use the grid to increase my processing power using the aggregate Ghz forces available on my private network.
the problem is, i have no idea how to get this software to start using the xgrid functionalities.
so here we go: is there someone out there that knows about FDS or would know about how to replicate (for lack of a better word) the parallel processing options available for FDS on standard IBM-compatible machines?
Thanks for the help.
you can send me your thoughts both on this forum and on lricher@hotmail.com using xgrid in the subject field to avoid the message being sent to the junk-oblivion folder
Again, thanks so much for the help.
Lau
While I personally never used Xgrid or worked anywhere that did, as both a scientist and a Mac user, I find this deeply troubling. Abandoning tech without warning makes Apple unreliable for research. At least we were warned about Classic and Rosetta, both of which were clearly meant as transitional solutions.
Apple may never win enterprise, but it has a STRONG presence in academia, especially among students. Why not make this an Apple stronghold? Why not sell OS X as a "High-performance *NIX for the rest of us". . . like they used to . . .
Oh, and the lack of updates to Apple's Science Page, which still features XServes and makes no mention of the death of Xgrid, is truly appalling and should be considered false advertising . . .
"Apple finally included what I would call the first fully functional release of Xgrid with Lion server (there had been longstanding bugs since Leopard). ... there was the nagging fear of seeing the word 'deprecated' in some of the key Xgrid systems. But I naively thought it was a mistake ... "
https://forums.macrumors.com/threads/1412689/
So last year Apple came through and marked key elements deprecated and this year it disappears. That is pretty much the standard decommissioning process. If folks need a "longer term" place to figure out where to move next Apple create a slighlty more stable node of Lion to sit in while figure it out.
For dedicated clusters that isn't an issue because wouldn't be rushing to install Mountain Lion anyway. For folks "borrowing spare cycles from neighbors" in more informal and/or non dedicated clusters ... well
1. those really weren't your cycles anyway.
2. a more heterogeneous grid system would provide for more neighbors to borrow from.
As to "bad mark" on the platform. It isn't. If XGrid isn't competitive anymore it is actually better to pull it off the table rather than to continue to Balkanize the set of viable products for the market. XGrid has a fatal flaw in that it is Mac OS X only. ( yeah there were some 3rd party efforts to make non OS X agents but that was never absorbed into the proprietary product itself.)
Xgrid's man page hasn't changed since 2007
https://developer.apple.com/library/mac/#documentation/Darwin/Reference/Manpages/man1/xgrid.1.html
It really wasn't evolving for a significant number of years either.
Documentation guide
http://developer.apple.com/library/...gramming_Guide/Introduction/Introduction.html
Last change 2007. Zero revisions since then.
I agree, Apple should get rid of products/services that aren't working. However, Apple should also give proper notification far enough in advance of the discontinuation so that IT departments can start looking for alternatives.
I'm aware Apple likes secrecy and likes building hype so they can better sell their consumer products. However, enterprise IT departments need a more visible road map of what Apple's going to do. They need to be able plan what they need, get the appropriate money budgeted, etc.
Thanks for the info! Definately a case of "did not read all of the OP's post" on my part. I think that counts as a warning. Could've been clearer, but this is the best we get out of Apple.
. . . And their Science Page still needs work . . .
Like I said above, I never used Xgrid myself. I had no idea it was so far behind the times! No revisions since '07! WOW! No wonder they killed it!
All true, but I'm not sure Apple will change for enterprise IT. The iPad and iPhone seem like the only products that can compete there.
Like I said above, I never used Xgrid myself. I had no idea it was so far behind the times! No revisions since '07! WOW! No wonder they killed it!
That's not quite true - the man page hadn't been updated for a while as there had been no new features for a while (there were no new features that would have been useful to us anyway, besides support for other platforms) BUT there were huge improvements in reliability and significant bug fixes with Lion one year ago (coincidentally at the same time as the word 'deprecated' made an appearance buried in the guts of the system - although for the record this was very much not obvious).
I absolutely accept that the vast majority of OS X server is total garbage, but XGrid worked well and was easy enough to set up that we could persuade people within our institution to donate spare cycles to use. And technically you didn't even need server anyway - all it did was add a GUI to the controller commands that we're there in the desktop version! It's going to be far more difficult with any of the alternatives, and I think we will end up with a smaller cluster even though we can add Linux/windows boxes...
Also, I don't accept the argument that we can continue to use Xgrid even though its been axed because people outside our control will eventually update to ML (or buy a new machine with it preinstalled) and drop off our grid - so either we have to hack the required files from Lion to new installs of ML to prologue it's life (which is apparently quite straightforward, at least for now), or watch our cluster gradually loose all the desktop nodes until it is down to just the server core and the one or two machines of people that I can persuade to boycot ML...
Maybe there are better alternatives for some or even most people, but not for us. Migrating away from Xgrid is going to be a headache I could have done without
Xgrid was not present in the Lion developer drops for a long term before the release resulting in speculation it may not be included in Lion.
The flurry of "Lion RIP" messages in 2011.
http://lists.apple.com/archives/xgrid-users/2011/Jun/index.html
Eventually it should up, but this is a least a yellow caution light. Either Apple was on track to make this an optional install ( like Rosetta the iternation before its retirement ) or on a path to be decoupled from the OS ( X11 and Java ). Since it is closed source, the latter was no particularly likely.
I suppose Apple could post a bright blinking neon sign, but that if this isn't a trigger to at least start looking at what the alternatives are, I'm not sure what is.
This repeated again in May 2012
http://lists.apple.com/archives/xgrid-users/2012/May/index.html
Any org with developer access at the first developer seed drop would have seen a missing element twice in 24 months.
Not talking about future products is a deeply seated core corporate policy. That is a deeply flawed tactic to take to try to rewrite corporate policy from the outside.
It is not a "road map". Wrapping future products "road map" up into the same discussion as "support lifecycle" is why wailing about this to Apple will likely almost always fail. An infinitely more productive discusssion with Apple would be about "where they have been", not "where they are going". Apple has an openly stated hardware desupport schedule.
Vintage and Obsolete
http://support.apple.com/kb/HT1752
That isn't talking about future products. So it doesn't come into inherent conflict with the corporate policy. What is missing is something similar to
http://www-01.ibm.com/software/support/lifecycle/
Where Apple would post in a somewhat consolidated fashion a place where deprecated frameworks and software .
The man page for 'sed' hasn't changed since 1995.
Xgrid wasn't so much behind the times as calcified. Any software of significant size always as bugs to fix whether internal defects or cracks in the interfaces to external services or components. All true, but I'm not sure Apple will change for enterprise IT. The iPad and iPhone seem like the only products that can compete there.
Macs are growing substantively faster than the rest of the classic personal computer market. As a platform it is healthy inside of Apple. It isn't the "only child" anymore, but Apple can walk and chew gum at the same time.
Even at reduced nodes it will be useful over the immediate future. However, this risk is not "new". If any of those folks had flipped to a Windows or Linux box they would drop off the network also. The grid was always at risk because it is built on an assumption that all the grid users will stay on OS X and that they'll upgrade roughly in lock step. That risk was always there. It is only exposed at this point.
What you are implicitly looking for here is a way to continue forever using XGrid. That isn't really the issue. The problem at this point is to make a transition to something else. If you need time to make the transition there are ways to prop up XGrid temporarily. Apple doing that substantial bug fix at the Lion transition more completely enables that.
Pooch is largely at the same level of ease of use as XGrid. That's probably why it got highlighted by the Apple product manager as an alternative. The only thing difficult about it is that it costs money. It comes do to something similar to "XGrid is great .... as long as I don't have to pay money for it or install it."
Perhaps with an increase in the number of XGrid refugees the Pooch folks can lower the price. It is kind of high for what it does. I'm pretty confident that is because they don't sell enough copies. It may be a 'chicken versus egg' .
I think the major problem with Pooch is that it scales poorly in terms of price for what XGrid is really good at - taking the Macs that are just sitting around and turning them into useful nodes.
For example, even with the academic discount, each Pooch node is $125. If you're talking about a badass cluster node, sure, that's not that big of a deal. But if you're talking about turning a departmental lab of 30 iMacs into a grid, you're talking $3750.
XGrid was outstanding for adding value to an existing Mac purchase with no additional outlay. Pooch is good, but makes that a little tougher.
That's the thing, I don't think losing XGrid will kill Macs in science. But the bad website, obsolete Mac Pros, XGrid...it's starting to feel like death by a thousand cuts.
Actually not. Past the intial node. it is $100/node. So $3,025. That seems large until look at 30 x $1,199 ==> $35,970. So that is about 8.5% of the cost of the cluster. It is a bit high but in the single digit percentage of the "cost" of the cluster. If get more CPU cycles out of the part time Pooch grid than a dedicated $4,000 server running full time then it is a good value. If there are so few "excess" cycles that can't outcompute a $4,000 server then not.
As I said before, this is far more about getting something for nothing ("free beer" and "off the IT radar" software acquisition) than about usability or features/value.
I think the folks selling Pooch would do more revenue if they charged something like $10-25/node/year for the license. So a 30 node iMac cluster would be $300-750/year and the profs using the cluster would have to pony up just "hundreds" of dollars per year.
$25/year would be about $2/month. That's pretty cheap. That's cheaper than what Amazon would charge for a single EC2 Linux node for 8 CPU days/mo.
I suspect Dauger Research charges so much for Pooch because folks buy it and then disappear for many years. So the number of paying customer per year is low.... so the unit price goes up. That generates a bad positive feedback loop pushing the unit costs up (or at very least stay constant over time).
If people are primarily scavenging for cycles they are likely relatively broke. Which makes it just about costs and not features. Those folks tend to like payment plans.
If primarily scavenging for cycles then something heterogeneous is better. If so broke can't pay for CPU cycles it is distinct disadvantage if can only get a small subset of the free cycles out there in the world. If there is a Windows PC CAD cluster down the hall XGrid does nothing with that.
I suspect someone will either do some work to add Condor or Open Grid to macports (i.e., make it so there is an "easy default config install" ) or that some other either "free beer" project will pop up or a commercial one.
There is always flux in applications the die off or come to market. The "bad website" is marginally better than no website. If folks keep bugging Apple about it seems likely they just take the whole set of pages down.
At this point, if Apple releases new Mac Pros it will only be the "obsolete" ones that can run XGrid. If that is such a high value add then not so "obsolete".
As for "obsolete" go here
http://i.top500.org/stats
Select "processor generation" (as opposed to vendors) and hit submit. Between the 5500 and 5600 series that is more than 50% of all the top 500 supercomputers in the world. Really the majority of the world's supercomputers are obsolete? Not really.
Not the bleeding edge is different from being obsolete.
I'll admit I was looking at the Pooch Pro license cost, which I think is the appropriate one for the functions that are being talked about.
It might be a good value, but it's a hell of a lot harder sell than "Free with OS X"
I don't think it is, any more than if Terminal.app was depreciated, and your only option was iTerm. It's not "something for nothing" - its something you were getting for free, core OS functionality that you may have built purchasing decisions on (indeed, XGrid was one of the major reasons I was considering a purely-Mac lab...) that went away rather abruptly.
I think dismissing that as "free beer" is disingenuous. It was an extended capability of the OS. It was a strength that made OS X decent for client-nodes. And yes, it was cheap, on an otherwise expensive hardware platform. It was also baked in and easy to use.
The new alternative is inherently not baked in, and generally speaking easy OR free.
Apples to Oranges.
I suspect its because they have very low unit sales and have to pay the bills. To be blunt, Macs are good enough client Machines in a Linux environment that for the most part, anyone seriously looking at doing this needs to be really dedicated not to go with a non-Apple hardware solution. Selling to a few die-hards (myself sometimes included) isn't a high volume business, even if you lower the price.
There's a considerable excluded middle - notably the potential to essentially have dual-use machines that aren't purely cycle scavenging in the background, but aren't a dedicated cluster setup. For example, 4 or 5 Mac Pros in a lab that when the lights go off at 8 turn into a small cluster for longer running overnight jobs.
Macports itself is kind of in shambles, and its neither as easy nor as well supported as XGrid.
That's the thing. There are alternatives. But every time I go look at "alternatives", and ad-hoc projects and poorly supported ports, I start asking myself if the convenience factor of OS X is still there. Eroding convenience while keeping a price premium is, for at least me as a customer, a losing proposition.
And if this was just random noise, it would be one thing. But we haven't had "good news" about scientific computing on a Mac for some time now. At a certain point, data becomes a trend.
No, its not. A website talking about solutions that can no longer be obtained, with incorrect and obsolete documentation that gives the perception this is supported when its not *is worse* than no website at all.
"Don't worry, our ancient products can still use our depreciated software!" isn't a defense. And XGrid was a value add - it wasn't an infinite value add.
A Top 500 cluster is a vast, yawning gulf of infrastructure investment than a workstation, or even a group of workstations. The Mac Pro is obsolete. It just is. It's got an out of date processor, a laughably out of date graphics chip, and lacks essentially every major new I/O option out there (USB3, SATA3, Thunderbolt...) Far superior workstation models can be had from Apple's competitors right now.
Does that make a Top 500 cluster built off a 5500 or 5600 obsolete? No. Does it make it useless? Really, really no.
But I wasn't saying either one of those things.