Submitted by Prior to the general availability of AIX 6, IBM intends to make a pre-release version of AIX 6 available in an open beta. “AIX is an open, standards-based UNIX operating system that provides the enterprise information technology infrastructure for thousands of clients around the world. IBM intends to take the next step in the evolution of the AIX operating system with the release of the AIX V6.1 OS.”
I understand the “standards-based” part. What does the “open” refer to?
p.s. This is an honest question, not a sarcastic one.
The Open part could be referring to Linux affinity, the “L” part used by IBM for AIX 5L. This would include adding parts of the GNU tool chain, the Gnome GUI and the ability to run Linux in a LPAR on a pSeries system.
If you had read the article (last section), you would have seen that the open is not for “AIX 6 Open” but for “Open Beta”. Which itself is refined as open for all their customers and not open for the world.
“The open beta program for AIX 6 will allow clients to download, install and run an early version of AIX 6 on suitable IBM systems in a non-production, non-supported mode.”
Read the first paragraph:
“AIX(R) is an open, standards-based UNIX operating system (OS) that provides the enterprise information technology infrastructure for thousands of clients around the world. IBM intends to take the next step in the evolution of the AIX operating system with the release of the AIX V6.1 OS.”
That is what fretinator is refering to and my response is based on IBM’s literature and some experience with AIX 5L 5.2.
Actually, no.
Here it is from Wikipedia:
“Open systems are computer systems that provide some combination of interoperability, portability, and open software standards. (It can also mean systems configured to allow unrestricted access by people and/or other computers; this article only discusses the first meaning.)
The term originated in the late 1970s and early 1980s, mainly to describe systems based on Unix, especially in contrast to the more entrenched mainframes and minicomputers in use at that time. Unlike older legacy systems, the newer generation of Unix systems featured standardized programming interfaces and peripheral interconnects; third party development of hardware and software was encouraged, a significant departure from the norm of the time, which saw companies such as Amdahl and Hitachi going to court for the right to sell systems and peripherals that were compatible with IBM’s mainframes.
The definition of “open system” can be said to have become more formalized in the 1990s with the emergence of independently administered software standards such as The Open Group’s Single UNIX Specification.”
[EDIT: Cleaned up quoting]
Edited 2007-05-22 16:34
Back before the term “open source” was coined, an “open system” was one where the user was allowed to see the documentation and write their own applications.
This might help:
http://www-03.ibm.com/servers/aix/standards/index.html
There seems to be a misplaced comma between “open” and “standards-based.” It should have been :
“AIX is an open standards-based UNIX operating system…”
That means “open standards.”
http://osgeek.blogspot.com
Since IBM keeps preaching to everyone else, are they going to open source AIX?
Don’t bank on it, I’m sure that IBM would have just as many if not more problems open sourcing AIX than Sun did with Solaris.
Yeah, but they’ve been using AIX for quite some time now, and I’m sure that more than one client would shout a few nasty words if IBM dropped AIX support to use GNU/Linux exclusively.
Personally I wish they follow Sun and build some excitement around AIX. There’s no reason to let a great OS stagnate.
There would very little reason to do so. Sun is able to build up hype about Solaris for one simple reason, it runs on x86 machines; cheap and accessible hardware which provides a significantly low enough bar for contributors to get involved.
If I were IBM, I would probably work with Sun and bring Solaris to POWER, phase out AIX and instead work on the middleware – given that the money is made off the middleware, and next to nothing (possibly even a loss) is made on their hardware and operating system, it would be a good way to cut costs and share the development costs with those outside IBM.
What I do find disappointing is the lack of a affordable POWER (or atleast some sort of cut down) workstation priced along the lines of Sun’s old Blade 100/150 (NZ$2600-NZ$3500 range) to encourage greater use of their processors outside the highend – which is what IBM is trying to accomplish OpenPOWER and blades, without much success.
IBM does sell POWER workstations close to that range (roughly 7400NZD for a tower, 4700 for a rackmount). Not quite as cheap as you want, but still affordable.
I certainly wouldn’t call that affordable – thats massively priced compared to the UNIX workstation offerings offered by Sun in the form of their Solaris/Opteron (and soon to be Intel as well) offerings.
I’m not just talking about people like me, but businesses in general and the cost of not only acquiring the technology but all the associated costs that go with it; why would I want to contribute to AIX in terms of middleware when the bar is set so high, the market relative to Solaris, is so small, it makes little sense to develop for AIX.
For IBM to develop an ecosystem, they need to lower the bar for developers to enter, they need to expand the marketplace by bringing the cost to enter alot lower by making their workstations affordable, and their developer tools for free.
Sun has realised this; right now from the outside it might seem to be not alot is happening, but like a Greek dance, it starts off slowly and gradually speeds up. Free developer tools, low cost workstations and free download of the operating system.
All of this has created a bigger, stronger community – and there is nothing stopping IBM from doing the same thing; but it always appears to me IBM just doesn’t have the balls to actually come forward and make something descent of their existing assetts.
It was seen with their lack of committment to OS/2, their lack of committment to getting Linux on the desktop – they want to do the easy things rather than growing some ‘balls’ and actually putting out a grand vision and aiming for it.
AIX along with POWER has the ability to grow larger than it is right now, but it require some vision and passion by management at IBM rather than waht I’ve seen today which is ‘going with the flow’.
Edited 2007-05-23 03:59
AIX is not stagnating. On the contrary, it is the focus of huge initiatives in RAS, virtualization, and management. Part of the reason why the Open Beta exists is to show customers how far we’ve come with AIX. Make no mistake: AIX 6 is a massive release, and the POWER6 platform it exploits will cause major shifts in the midrange and high-end of the IT market. The excitement will be in the bottom line. Whether it’s double the performance at the same power or half the power at the same performance, POWER6 will have an impact the economics of the datacenter.
Enterprise IT customers told us that RAS (Reliability, Availability, Serviceability) is the most important characteristic they look for in their mission-critical computing environment. AIX 6 on POWER6 reflects an enormous commitment to RAS. Storage keys is a huge feature that allows user applications as well as the kernel to protect pieces of their memory against flaws in their own code, effectively creating hardware-enforced memory protection within an address space. Solaris has DTrace. AIX has system trace, lightweight memory trace, component trace, and dynamic trace (ProbeVue). AIX 6 has live dump, which allows fine-grained kernel dumps to be taken at runtime without bringing down the system. This joins system dump, minidump, firmware-assisted dump, and full-memory dump. Concurrent update allows kernel updates to be applied to a running system without disruption, sometimes known as the holy grail of RAS. The recovery subsystem can intercept failures and recover the affected component at runtime.
AIX has hardware virtualization and OS virtualization (LPARs and WPARs). Live partition mobility, virtual I/O, multipath I/O, dynamic LPARs, shared LPARs, CPU and memory hotplugging, high-availability clustering via HACMP, and more. System P features PHYP, an advanced hypervisor that forms the backbone of the hardware virtualization environment that supports both AIX and Linux. It also has a dedicated service processor for the firmware and NVRAM for use by the firmware and the OS instances, enhancing data capture beyond the limitations of DRAM and hard disks.
AIX 6 on POWER6 is leading in so many areas that it’s hard not to be excited. These are the kinds of technologies that enterprise customers want to build their businesses around. It’s not going to run on your laptop, and it’s not going to have fancy desktop effects. For the free software community, it’s only interesting in the sense that it shows where the enterprise IT environment is heading. There’s no reason why commodity hardware and software cannot replicate the big iron and mainframe features provided by System P. Think about how to make a mainframe out of a cluster of commodity PCs. Think along the lines of what Google has done with Linux.
NOTE: I’m an AIX developer with IBM, but I do not speak for IBM in any way. See my profile.
That was great information about AIX 6, explaining some of the advanced Enterprise features. After reading that it makes me see the difference in directions AIX and Solaris are going.
AIX is targeting on those really high level Enterprise features, that are linked closely to their hardware. Many of their AIX features are very dependent of their hardware features (like LPARS).
Solaris is trying also to bring high level Enterprise features, but to do it at a level that supports both their SPARC and X86 offerings. Because of this I doubt they will be able to offer as many special Enterprise features that AIX does because of the limitations that X86 inforces on them. For example the SPARC version of Solaris has better “self healing” because their hardware has more probes, verus their AMD offerings. Hot swappable CPUs would be another feature known in SPARC, but not sure if they have available for the AMD equipment. But they are trying to bring their X86 offerings up to bar with the SPARC.
Personally I like the direction Sun is taking, but like that IBM with AIX is adding some really high level features for that small niche market.
The Power6 will keep the future going for IBM and their machines. Sun is not standing still, coming soon the Rock processor, which has been in the works for over 10 years. Will be interesting to see how both the Power6 and Rock CPU’s match up. The future looks bright in the big iron unix.
Personally I like the direction Sun is taking, but like that IBM with AIX is adding some really high level features for that small niche market.
I question how niche these features might be today and in the near future. I think the midrange of the corporate IT market is coming out a period where they thought they could do IT on the cheap with a bunch of distributed infrastructure. But what they found was that IT gets really unmanageable, unreliable, expensive as that infrastructure grows out of control. The shops that bought a bunch of Dell equipment found that they weren’t running a datacenter, they were babysitting a ticking timebomb and putting out fires.
I think its clear to most CIOs that they need to consolidate workloads onto serious platforms designed to handle the rigors of continuous availability. You can certainly build solutions from commodity hardware and software that fulfill these requirements, but only if your business is IT. This works great for Google, but it’s not feasible for Wal-Mart. Businesses that need serious IT without making IT their business are the target market for big iron and mainframes. That’s no niche. That’s most medium and large businesses, as well as many growing small businesses.
At this point, it’s not so unfair to forget about HP and consider the ballgame being Sun vs. IBM. Sun has a lot of things going for them, especially in the way they’ve positioned themselves as an OSS platform with rapid growth potential. But that’s mostly forward-looking stuff. Businesses that are dying to consolidate and simplify their IT are looking mostly at the here and now. In this sense, there’s no question that IBM has the lead in all areas except for possibly initial cost. They’re got better performance, power efficiency, RAS, virtualization, management, and service offerings.
The question is whether the long-term benefits justify the potentially higher up-front cost. Given their past experiences with IT on the cheap, I bet that many CIOs will seriously consider the TCO advantages possible with IBM, whether it be on X, I, P, or even Z. Sun just doesn’t have the virtualization capabilities to consolidate and isolate workloads, which is a large part of the TCO equation. They don’t have the RAS, and their hardware is only moderately more robust than an ordinary Dell server. They might be in exactly the right place on the price/performance/capability curve, or they could be in no-man’s land.
This week’s product refresh announcement certainly didn’t help Sun any. Rock has some mighty big shoes to fill, and Solaris is years behind where it would need to be to compete with AIX.
Edited 2007-05-23 07:20
“Sun just doesn’t have the virtualization capabilities to consolidate and isolate workloads, which is a large part of the TCO equation.”
That’s interesting – ‘coz I’ve consolidated some workloads with Sun’s virtualization and I know other people who have done so also. And it works great.
In most cases buying AMD/Intel or Niagara servers and consolidating on them is much more cheaper then going with AIX/Power – not to mention that with AIX/Power you are basically lock-in. I do have many servers from different vendors like IBM with Solaris on top of them. Even IBM’s blade servers. I can’t do that with AIX.
“They don’t have the RAS, and their hardware is only moderately more robust than an ordinary Dell server.”
I haven’t been working with AIX/Power lately much (just barely) but when I did I haven’t noticed any better RAS than other vendors. Now when you consider SMF, PSH and ZFS in Solaris this increases RAS – I can see that in a next AIX version you’ll get some subset of that features and probably not that mature (those features are in Solaris for years now).
“They might be in exactly the right place on the price/performance/capability curve, or they could be in no-man’s land.”
It is about being in that place. It’s exactly what changes most data centers in last years. Sun didn’t get it and finaly they do. IBM still doesn’t get it (ok, not entirely – they do offer quite nice x86 hardware).
“[…] and Solaris is years behind where it would need to be to compete with AIX. ”
From my experience I would say exactly the opposite.
AIX will offer in a future similar functionality to Solaris Zones which are there for years and widely used. AIX will offer something like Solaris’s DTrace – it’s not there yet and we’ll see how it compares. And again – DTrace is there for years in production use.
AIX will offer file system encryption – so will ZFS. Now there’s nothing like ZFS in AIX even on a roadmap.
Then you have Predictive Self Healing (it saved my butt more than once), SMF – you’ll teg there with AIX some day.
Then Solaris is free and is Open Sourced. And most importantly it runs on x86/x64 and on several different HW vendors.
AIX is great and just because it’s been running in many critical environments for years (like HP-UX or Solaris) it will be there for many years to come. But AIX is becoming more and more of niche product every year struggling to catchup while Linux and Solaris are getting more and more ahead (with Linux still being behind AIX and Solaris in enterprise environments).
What I see is that if you already are a AIX/Power shop it could make a sense to still stay with it. If you’re not in most cases it doesn’t makes a sense at all. Solaris and/or Linux are much better choices.
I’ve consolidated some workloads with Sun’s virtualization and I know other people who have done so also. And it works great.
Solaris only has OS virtualization. You can’t run Solaris and Linux, or different releases of Solaris, on the same box as far as I know. Maybe with Xen? All Zones share a single point of failure due to the fact that they share a kernel. There’s limits to how differently you can configure and tune each Zone.
WPAR in AIX 6 is IMHO essentially a checkbox to compete with Solaris. The traditional selling point for OS virtualization is that it’s easier to manage than hardware virtualization, but with System P they’re both pretty easy to manage. I doubt that WPARs will be deployed heavily as a workload consolidation solution, but they might be used to facilitate RBAC and partition mobility. For most consolidation challenges, LPARs are the better answer.
Sun didn’t get it and finaly they do. IBM still doesn’t get it (ok, not entirely – they do offer quite nice x86 hardware).
I don’t understand how developing non-x86 hardware for big UNIX workloads is tantamount to not getting it. For those that want to go the x86 route, there’s System X, which runs Linux or Windows. That’s more of a direct competitor to Sun’s midrange. System P isn’t really competing directly against Sun’s high-end. It’s a performance-oriented quasi-mainframe. There’s nothing else quite like it, but I think it addresses a sweet spot in the tradeoff between ultimate performance and ultimate reliability. I guess my point is that Sun’s high-end and System P are often seen as competitors, but they’re really operating on different planes in the market. The decision becomes whether Sun’s high-end is good enough, or if System P is a better investment.
There’s nothing for IBM to get. AIX won’t run on x86. Linux already has that market covered, and I think that the market will eventually bear that out with respect to Solaris. Sun will exhaust its resources trying to match pace with Linux on x86. Customers will prefer Linux on x86 due to more flexible and competitive support options. AIX and Linux are not competitors. Solaris and Linux are direct competitors on x86. I cannot understand why Sun wants to pump money into two competing operating systems. It’s like funding both sides of an arms race.
Yes you are correct, Solaris has OS virtualization (zones, which is like BSD jails) and hardware partitioning. They are working on getting Xen working in a high quality way on Solaris X86, which would provide X86 virtualization support (for Linux, Windows, etc).
The thing with LPARS is it tied to the IBM hardware hyper visor. As far as I know it runs AIX and Linux, but not Windows (like Xen can). Benchmarks out their show Xen is actually very efficient, and close in many areas to native speed.
Solaris zones also support Linux binary compatibility. So you can run Linux applications right on the Solaris systems without recompile. So basically by using Solaris zones on X86 you can run Solaris applications and Linux applications in virtual zones. How is IBM’s hardware solution that runs Linux and AIX tht much better. Does it justify spending that much more $ for their hardware hyper visor? I have used Solaris system with many zones running on them and I can say it is very efficent and very stable.
Why X86? If you have noticed recently even X86 is moving quickly. In the past you could say Intel/AMD sucked compared to Power6 and SPARC, but X86 is catching up quick. AMD64 chips and the Intel chips are nothing to laugh at. They are taking power requirements serious, they are adding virtual environment extensions, and multiple cores. X86 is getting to a point where it can be used efficiently in the Enterprise (unlike in the past).
I know in a previous post you said IT bought a bunch of Dells/HP and tried to run on that, and found out it was junk. I agree with you totally. But yesterdays X86 hardware is just not the same as today’s X86 hardware (like Suns AMD offerings, Dells DL’s, HP Proliaents). They are compact, they have management cards built in, they have built in diagnostics, they have multiple cpus (with multiple cores), and the big one being they are power efficient. The huge problem companies saw with the past cheap X86 hardware was it broke a lot, and it used ALOT of power (like P4’s). Today, and into the futures trend is a whole new direction for X86.
You use a generic term with Linux. Most serious businesses will use a supported Linux such as Red Hat, Novell SUSE, etc. Have you compared the cost of using Red Hat Enterprise vs Solaris X86, in terms of getting the software and service support? You will be surprised that Solaris X86 is very competitive on price. So I think Solaris X86 can compete very well with supported Linux’s in the business market.
So I think Solaris being in the X86 is a VERY smart move, even in the medium to Enterprise business area.
Solaris 10 x86 used with AMD processors has FMA capabilities and if I am not mistaken has had them since Release 6/06.
Is it me, or are they trying to play catch-up with other OSes ? All that seems pretty much work they could have put into Linux.
Okay I’m not being fair, but for the fun, here is the feature list with some rough guessed comparisons :
* Workload Partitions – Xen
* Live Application Mobility – Xen live migrations
* Role Based Access Control – Windows ACL, …
* Trusted AIX – SELinux
* Encrypting filesystem – linux loopback devices, Vista BitLocker, …
* AIX Security Expert LDAP integration – ActiveDirectory
* Secure by Default installation option – Major trend, just look at the new linux distros and Vista too
* Graphical Installation – L.O.L.
* Network Installation Manager support for NFSv4 – …
* Kernel support for POWER6 Storage Key – ???
* Concurrent AIX kernel update – Linux modules
* Dynamic tracing – DTrace on Solaris
* Enhanced software first failure data capture – better dump
I take it you have never used AIX, so let me help you out on your comparison:
* Workload Partitions – more than likely an extension to IBM’s LPAR technology
* Role Based Access Control – Check out RBAC for Solaris
* Trusted AIX – More like Trusted Solaris
* Graphical Installation – for people who have bitched about how bad it is to install Solaris have never installed AIX (I have). Install BOS (Base Operating System), want X, load a CD. You want man pages, load another CD, wash, rinse, repeat until AIX is installed with what you want or need.
A graphical installation would only be the beginning of the improvements IBM could add to the AIX installation routine.
Actually, using the installp interface through smit is pretty easy to use, if you know what packages you want. Having the DVD instead of the 7 or 8 CDs (depending on version) helps for convenience too.
Is it me, or are don’t even have a clue about AIX ?
From what I saw on my P5 570s running SAP, linux should try to catch with AIX, not the reverse kid.
AIX will always be playng catch-up on feature checklists, because AIX is about reliability and performance, not features. If you want the latest and greatest features, you can always install Linux.
Depending on what kind of feature we’re talking about. When it comes to RAS feature, AIX has a decisive lead over Linux.
>* Workload Partitions – Xen
More like SUN containers I believe. Except intended for single applications I think.
>* Live Application Mobility – Xen live migrations
Sure. Or VMWARE.
>* AIX Security Expert LDAP integration – ActiveDirectory
No. SE is a tool that can (as a small part) distribute security profiles. Putting profiles in LDAP seems sensible to me.
>* Graphical Installation – L.O.L.
I’ll do with the textbased one. This is one for the featurelist only, not for real admins.
>* Kernel support for POWER6 Storage Key – ???
Dividing a monolithic all-access kernel into “compartments” with access restrictions. AFAIK. 100% dependent on new CPU hw (don’t know if some other chips like Itanium might have similar).
>* Concurrent AIX kernel update – Linux modules
I’m betting on it being rather better and more deep-down guts replumbing. Another 24/7 feature you want.
>* Enhanced software first failure data capture – better dump
Nah, this is about watching hw fail. Or hopefully telling before it totally does. More 24/7 features.
I’m not a big fan of OS-automobile analogies, but you’re effectively criticizing a bulldozer for its lack of lack of air conditioning or leather interior.
I rescued an old RS/6000 F50 from my previous employer. Dual 332MHz 603e with 1GB of RAM and a SSA RAID of 16 9GB hard drives.
I have AIX 5.3, but I haven’t had time or the energy to play around with it. Now I get to play with beta software. How fun!
That depends on whether IBM will release a 32-bit variant of AIX 6 and if I am not mistaken, your hardware is 32-bit.
Don’t count on it unless you currently have a support contract with IBM or your employer is one of IBMs larger customers. If you read the article:
“Notices will be also be sent to IBM sales representatives and IBM Business Partners and notices will be posted on the AIX and Linux(R) on POWER community pages at http://www.ibm.com/systems/p/community/index.html“
Don’t call us, we’ll call you
No support contract will be required, but to make sure that as many current clients as possible have the opportunity to participate, they will be using as many communications paths as possibile. Watch ibm.com/aix for the availability announcement.
Sorry stack – F50s have 32bit processors. The beta will support Power4 and later systems but the F50 is much older.
Unlike Solaris, which has had a mostly Unix-free codebase since at least Solaris 8, AIX (reputedly) still contains rather vital parts of the closed source “official” Unix codebase. Thus, Sun could open source Solaris, because it was 100% it’s own code, but IBM would need to rewrite large hunks of AIX to open source it.
In some ways, actually, IBM has open sourced as much of AIX as they legally could. The parts of the Linux kernel that SCO is suing IBM over once resided in AIX.
I must admit that I am biased toward AIX. I worked for IBM on the R/S6000 line. To compare linux to AIX is very unusual. One might suggest that linux is Unix like or has features of Unix or may even be claimed to be as stable as Unix. I would trust AIX on a critical system, such as landing of jetliners. I would not trust Linux for any critical task. I wouldn’t trust linux for a demanding task that would involve billions of dollars. I wouldn’t trust linux to protect various explosive things. AIX does all that and seems to have proven it’s worth over the many years of up time.
After reading that web page about new features it sounds like they looked at Solaris 10 (the other big iron Unix).
* Workload Partitions – ZONES
* Live Application Mobility – Would like to see live migrations for ZONES.
* Role Based Access Control – RBAC for Solaris
* Trusted AIX – Trusted Extensions in Solaris 10
* Encrypting filesystem – ZFS needs this and is being worked on. But ZFS has many other features over the current JFS.
* Secure by Default installation option – This is in the latest Solaris update
* Graphical Installation – Solaris has this, and working to improve this (try to get to Linux level).
* Network Installation Manager support for NFSv4 – Of course NFSv4 is in Solaris!
* Dynamic tracing – DTrace on Solaris
* Enhanced software first failure data capture – Solaris 10 Self Healing features, and how it links error codes to specific issues.
So if you really want AIX 6 features now (and production quality, which is a big thing), and want it for both SPARC (as compared to Power CPU), and X86, then I would suggest Solaris 10. AIX is playing catchup.
But don’t get me wrong, Solaris 10 is not perfect. Three areas Sun really needs to improve on are:
1) Patching – Too many reboots required when patching, and patching breaks, is slow, etc etc. I can go on and on about how bad Solaris 10 patching and package management system is. Now in days you would think someone could patch a majority of an OS without rebooting. I like that IBM is trying to add something to help this. Sun needs to do this also!
2) Improve the GUI tools for the Linux and Windows users. Linux install GUI’s and management tools are still easier to use.
3) Continue to make the X86 version more compatible with devices out there.
I don’t know about you but when I patch a Solaris machine I usually use a Patch Cluster, so there is only one reboot. And most of the patches that Sun releases do not require a reboot unless the patch modifies the kernel or a device driver. Also package management works fine for me and many other Solaris admins. I suppose you have never used AIX, because their package management is no picnic either, but it works.
Considering a number of Solaris (and AIX) admins control their systems through either a SSH session or through a terminal server, what point is there in making GUI tools? What I think Sun should provide is an equivalent to smitty (AIX), probably the best administrative tool made (next to smit and SAM (HP-UX).
Hardware support. Maybe you (and everybody else) should complain to the vendors, it is just as much their fault as it is the OS vendor.
That is not true. There are many patches that require a reboot that have nothing to do with the kernel. Don’t believe me check out “http://sunsolve.sun.com/search/document.do?assetkey=urn:cds:docid:1…. That is the latest patch for Apache 2, and as the docs say, it requires a reboot (straight from Suns web page). Also that patch breaks all Apache modules because of them compiling large file support. Not a good idea since not all Apache modules can’t be recompiled (because they are older modules, or source was never available). But that is only ONE example, there are many patches just like this one that shouldn’t want you to reboot. A reboot is required for it to install if you use SMPATCH. Call Sun techs, they will tell you to install most patches in single user mode.
Only do patch clusters? Those are only released like every 6 months (not sure exactly, but it is several months). In between the patch clusters there are many IMPORTANT security patches you will probably want if your Sun server is on the Internet. I won’t name them, but some are REALLY bad exploits.
So what do you use to patch? Well you have “patchadd”. If you want a automated tool kind of new in Solaris, you have smpatch, or if you want the GUI updatemanager. Let me tell you I have used BOTH of those, and two of my Sun systems eventually had a completely broken package map. I even called Sun support, both times they told me to RE-INSTALL or UPDATE from disc. One of the systems I tried to use the update disc failed in the middle, then wouldn’t reboot (and the update keep failing). Sun told me to rebuild the system (not fun).
I expressed my concern directly to Sun about this, and they agreed they have issues with their patch system with Solaris 10 and zones.
Why GUI tools? Well not everyone wants to use command line completely, and it makes it easier for Windows or Linux people to use. I personally do everything from the command line, but that doesn’t mean everyone should! Also since you mentioed SSH, tunneling X-windows apps through SSH is very easy, and makes access remotely to GUI’s VERY secure and easy.
I do agree with you about smitty, and having something like that on Solaris would be nice. Smitty is very nice.
I was just saying working with vendors to continue the device support will only help Solaris cause in running on more computers.
In conclusion I admin Solaris systems for a large company, and we have millions of dollars invested in Sun. I work with them daily. I have even talked to Sun Engineers who agreed with me on some of the things that are issues. I will add that most of my issues with the patch system has been with Solaris 10 and zones. But Solaris didn’t get really good until 10, which is when we invested more into Solaris. If you are a Solaris admin then you know the leap Solaris 10 was over previous versions in many areas.
Sorry to disappoint you but Solaris is not perfect. I am a huge Solaris fan, but I won’t try to cover up its flaws.
I don’t know where you get your information about Patch Clusters, but Sun releases them on a schedule of about one every two weeks and have been for years.
I can tell you when Solaris 10 first came out, they were not releasing Patch Clusters every two weeks. I remember looking for them, and the first Patch Cluster took MONTHS to be released. Their push with Solaris 10 was using SMPATCH and UPDATEMANAGER for updates. The problem was these were not production ready and caused problems. Second the patch/package system didn’t work well with Solaris 10 with ZONES running.
But it looks like your right now in days, Solaris 10 patch clusters seem to be updated every two weeks. But this was not always the case (like you said have been the case for years, not true).
I know that in the past Sun has tried to push using updatemanager and smpatch, and a number of people have had major issues with it (myself included), so I went to the the tried and true Patch Cluster method.
My statement is not limited to Solaris 10, all versions of Solaris that Sun supports have Patch Clusters available and are released roughly every two weeks. As far as Sun not making Patch Clusters available in order to force people to use updatemanager and smpatch, I have no idea. When I needed a Patch Cluster for Solaris 10, there was one available and it was produced within the last couple of weeks.
I have been using Solaris 10 since it was released originally. When it was released originally patch clusters were not happening very often. That is why I started using smpatch and updatemanager. I don’t think it was them forcing people, just they didn’t keep up with things for Solaris 10. So I continued to use smpatch, patchadd, and updatemanager. After using it for awhile it eventually put 2 systems into a corrupted state.
But I will agree with you, I suggest anyone doing patching on Solaris stick with the Patch Clusters. As you said you, and I know I, have both had issues with the smpatch and updatemanager tools.
Also I found that some 3rd party packages can corrupt things also. Which is another reason why I would like to see a better package management system.
So your safe bet, use only Sun packages, and use patch clusters. At least until they fix the package management to prevent 3rd party apps from corrupting things, and smaptch/updatemanager.
But it is hard to tell what works or not until you get bitten.
I have only been bitten by Patch Clusters twice, in both cases it was on machines I could afford to “lose”. My policy now is to wait 30 days and use the Patch Cluster I downloaded 30 days ago to see if anyone else has issues.
After spending two years Beta testing Solaris 10, I held off on the idea of actually deploying it until a few releases were issued. It was a lesson I learned with Solaris 8 where we had two Blade 100’s and the RBAC information was different. Eventually we figured out that one of the Blades had Solaris 8 4/01 installed on it, which Sun had made improvements to RBAC.
We standardized on Solaris 10 6/06 for our base installs.
Yes, you’re mostly right.
S10’s release must have had a huge impact on AIX’s roadmap – remember at that time many people, including many in IBM – were talking about replacing AIX with Linux, even though AIX 5.2/5.3 was already(arguably) a better Unix than Solaris 9.
I can do exactly the reverse.
Example:
. Solaris ZFS – try to catch up with the excellent LVM of AIX (migratepv, etc …). And no, Sun LVM is just a joke (and a pain to manage just look at Sun’s presentation of ZFS to see the comparison)
. Solaris Zone, try to catch up with IBM LPAR (but LPAR is a REAL virtualisation solution). With LPAR, I don’t have to reboot all my servers for a patch in the global zone …
. Solaris SMF – try to catch with System Resource Controller
Some things solaris still didn’t catch
– The device management and I mean it: on AIX checking the hardware, which card depend of this controller, etc is a pleasure.
– The error handling (errpt): the single point to check if your host is failing with some increadible description of the problem (instead of checking dozens of .log file with one line of description..)
I would point to this email:
http://www.unixguide.net/cgi-bin/webbbs_config.pl/noframes/read/765
Not that I really like Solaris 10, but saying that IBM is trying to catch up with Solaris is just a joke. Solaris 10 just catch up with AIX, and not in all area today.
Edited 2007-05-25 15:11
I agree with you. At my current admin position I get the pleasure of using Solaris 10 on E2900’s and AIX on P 570s.
As much as I want to say Sun has the better platform, it just isnt the case. AIX/SystemP is a pure joy to work on compared to the Solaris systems. AIX LVM is so elegant compared to clunky Solstace or Veritas. I dont know of a single Sun shop using ZFS in a production capacity either. IBM NIM makes patching/deploying so simple a caveman could do it. LPAR blows Zones totally away.
I used to be a Sun bigot, but as sad as it makes me to say it, they dont have the better technology anymore.
I use ZFS in production and have been for some time, no problems here.
Have you ever used JumpStart and Solaris Flash?
And are you talking about software or hardware based LPARs?
System P + AIX for companies that the core business isn’t IT.
GNU/Linux + x86 (lowest cost possible, probably clustered like google) for IT companies.
It’s all about support. The rest is showing off.
I’m curious, will AIX boot on an Apple G5? AIX supports PPC970 systems, but I’m guessing the Apple lacks certain chipset features required by AIX.
hey, this post is very useful, I really enjoyed reading this, thx
iPhone Video Converter
http://www.iphoneconverter.com/iphone-video-converter/