Linux·4,557 discussions

Frustrated with hardware vendors trying to attract Linux users with crap

Locked

Recently, Ive been researching using Linux at the center of an IP storage network. Specifically, I researched storage controllers (SATA) from vendors such as Broadcom, 3Ware and SuperMicro. When reviewing available performance data in the form of benchmarks , I normally tried to compare the controller’s performance on Linux against it’s performance on the Windows OS. In almost every benchmark I’ve read and digested, it seems that the controllers do not perform as well on Linux as they do in Windows. An example of this can be found in a white paper read through. That paper can be found here:
http://mgmt.datatag.org/sravot/paper/TCP_WAN_sr081604.pdf

The white paper is a research effort to overcome obstacles when performing data transfers over high speed WANS. It also deals with the last mile of the data transmission, which is to disk.
In one of the testing involving the SuperMicro SATA controller, it shows that the best write on Microsoft Windows is 85% BETTER than the best write on Linux. They attribute this obvious inconsistency to “better drivers and optimization in the Microsoft OS”. This is soooo frustrating. It seems that vendors are trying to attract Linux system builders and admins by developing drivers for their hardware on the Linux platform. This good! We want access to better and more modern hardware to run on our Linux powerhouses. Howerver, the performance on their hardware on Linux is no where near what one would achieve when they take the same card out of the system and stick it in an identical Windows box. Vendors, from my perspective, are not working close enough with the Linux kernel team to develop high quality drivers and software for their hardware. They appear to be just throwing stuff together and giving it to us and saying “Hey, guess what.. we got a driver for our XXX-455 Super SCSI Wonder Card”. Just go to this vendor and pay $1K US. But when you stick that card in your system, load the drivers and run bonnie++, you are very upset when you manage to get 200MB write per second across 12 drives JBOD when you know that you can get 400MB on Windows! I think the linux community needs to develop a standards body that addresses this issue and holds vendors accounting for delivering quality software and drivers for the Linux OS. There needs to be a Hardware quality Labs equivliant to the WHQL (Windows Hardware Quality Labs) that standardizes linux driver development and independantly tests and confirms the operating of hardware and their drivers on the Linux OS and makes vendors work to get a seal of approval. Does anyone have any thoughts?

Errol

Topic

Well instead of trying the free stuff

In reply to Frustrated with hardware vendors trying to attract Linux users with crap

If you really want the latest drivers for the latest hardware try Unix which is another pay for OS like Windows. It has their own developers writing code for the drivers for the latest hardware.

The Linux community on the other hand is only dipping their toes in the water testing these drivers as they are still even under a Windows platform slow compared to the tried and trusted SCSI controllers and drives that have been around for so long.

If you are really thinking HIGH END Servers you just would never consider SATA as it is so slow compared to SCSI. The only advantage is price and that shouldn’t be an issue with the high end of the market.

Col

SATA

In reply to Frustrated with hardware vendors trying to attract Linux users with crap

You are going down the wrong road when attempting to implement an IP storage network using SATA. SCSI is the tool that you need and it’s well supported under Linux…..I’m assuming that you are working at the hobby level on this.

You’re definitely right…

In reply to Frustrated with hardware vendors trying to attract Linux users with crap

… and the two guys that posted before me completely missed the point.

The point is not that you’re trying to build an IP storage solution using SATA and not SCSI controllers, but the fact that using the same controller under Linux results in much lower performance than under Windows.

The idea you suggested in the end sounds good, but I don’t know if it has chances to become reality. The main disadvantage of Linux OS before Windows in this respect is the competition on the same hardware platform, the PC, and the lack of a central decisional body that approves or denies an OEM driver implementation as being “compliant” for the platform. This is the reason I think your idea sounds really good, a “driver approval team” working in cooperation with, or responding to the kernel team makes a lot of sense. Another option would be, of course, that the major distribution vendors like RedHat, Novell, Mandrake, etc have an “approval” process of a certain driver for their particular distribution.

Yes I understood that but

In reply to You’re definitely right…

Why is this happening.

You have a limited number of SATA drives available on any M’Board and unless you are only talking a very small storage platform why use SATA.

I’ve used these drives under Windows and I really can not see any real improvement over the IDE drives that are currently available and even under Windows the SATA are slower than a ATA drive.

Sorry but I can not see any advantage of using SATA drives for a storage solution. Even the network storage devices only support IDE or SCSI depending on your budget.

They are more expensive than a better IDE drive and far lass expensive compared to the much better SCSI drives. The only real advantage of using SCSI is the fact that with a good controller you have the ability to plug in 30 Drives which is something that just will never happen with either IDE or SATA.

Maybe it’s just the way I read it but to me it looks like someone has read a mag article about how much better the SATA drives supposedly are compared to the IDE units and then read something about Linux being a more stable platform so they are trying to marry both into one application and expecting better results than Windows could provide as far as data transfer goes.

As I currently use all the currently available OS I just can not see why someone would want this setup. It is Bleeding Edge and prone to problems and the person who originally posted this thread seems to want the best of both worlds but is unable to accept the limitations of the current development of Linux and the market domination of M$ and all that goes along with that. What I mean here is the way the M$ actively discourages development of third party drivers for anything but Windows. Also I did get confused with the mention of SATA and SCSI in the same proposal as I know that Linux supports all forms of SCSI much better than Windows ever could.

Sorry but I just can not see the reason at all.

Col

Clarification…

In reply to Yes I understood that but

First of all, I apologize if my post made you or Gary (the other poster) fell uncomfortable. I was not my intention. Unfortunately, I almost always see an interesting subject posted in a thread, but only a few of the replies contributing to the problem, the rest… invariably switching the theme towards unrelated issues.

As for your comments, I certainly agree: preformance == SCSI, and definitely almost any *nix has better SCSI performance than Windogs.

In what regards SATA though, I think you are a bit unfair to the initial poster, since I don’t think you can easily find a tech site or publication that recommends IDE before SATA, and I think that they’re probably right (I’m not an IT guy, but a developer, so I don’t have a benchmark comparison from my personal experience), since at least theoretically, this has to be true, at least the interface was designed to have a better throughput than IDE. The reason it isn’t… well, that might just be faulty manufacturer drivers, or just immature technology or device implementation.

All in all, I think the following problem still remains for, probably, a fairly large number of devices: they are supported better by Windows that Linux… and this should definitely change for Linux to be able to consolidate and grow on the server side and start penetrating better on the desktop.

Again, the main problem of Linux is the competition on the same hardware platform as Windows. This is the main reason I guess, companies like Apple running a BSD-based OS hang so desperately to their proprietary hardware platform, for which, most definitely, they have the best drivers in the world! :o)

But, I may add that they don’t represent a threat to Windows neither on servers nor on desktops… ;o)

No problem from this end

In reply to Clarification…

I was only trying to point of the disadvantages of trying to use SATA drives on any storage system.

Also with IDE drive with the 8 MEG buffer of Cache built in now days they perform better than the SATA drives do but that again is purely hardware related. People see figures like 150 MEG per Second and think it’s great but the reality of the position is that unless you are coping consecutive data across you will never see speeds approaching this. The same applies to both SCSI and SATA as well. While the SCSI’s do have a definite speed advantage it does come at a cost which every business is willing to bear but not many hobby people can afford.

Currently a 146 GIG SCSI is about 5 times the price of a 250 GIG IDE and the cost factor starts coming in a big way. The SATA’s are only marginally more expensive than the IDE’s but really do not offer the supposed speed advantage that the figures quote, which I might also add are always under a Windows benchmark speed test.

Recently I built a complete network with SATA drives and honestly I’ll never use them again as they just where not worth the extra cost. I also latter built exactly the same computer but with IDE drives that was the only difference and it works faster than the ones running SATA drives by a lot. Thankfully it is not in the same building as the others or some questions might be asked.

I always use at least 2 HDD’s per workstation one with the OS on it and the other contains all the data so if Windows goes down the data remains intact as in business it is the most expensive part of any computer/network and the one thing that a price can not be put on.

If this person really wants to use something other than Windows for a SATA array there is Unix which fits the bill nicely and they do have some good SATA drivers written for that OS but because of the current state of play between SCO and IBM Linux is lagging a bit in this race.

Col

I’m frustrated by unreal expectations

In reply to Frustrated with hardware vendors trying to attract Linux users with crap

People don’t understand the entire philosophy behind Linux. That philosophy can be stated as follows: If you want something that isn’t there, such as a hardware driver, then write it yourself. If you want a better implementation of something that is already available then write it yourself.

Linux has always been a showcase for people who wanted to contribute to a real operating system. This fact has been obscured by commercial vendors adding their own contributions, such as installation and configuration software, in order to attempt to make Linux more legitimate.

If you want a system that is managed like a commercial product then get a commercial product. If you want Unix then get real honest-to-goodness Unix. There are many flavors available. HP-UX, AIX, Solaris, Open BSD, Net BSD, Free BSD. All of these products are developed by dedicated teams that perform all of the functions that you said that you want to see in Linux.

The value added resellers of Linux such as Red Hat and SuSE and Debian are trying to take what was originally a hobbyist system and make it into a commercial enterprise product. Frankly they’re doing a pretty good job. The Linux 2.6 kernel is a pretty good performer for single and multiple CPU systems. The differences in the performance of the different distributions of Linux suggest that you should try several of them to see which ones do the best job. However, I would expect that one or more of the commercial Unixes would perform better than any distribution of Linux.

I think that your complaints about Linux are like asking a cat to bark. It’s just not made that way. I use SuSE Linux at home and I enjoy it. At work I would look to Solaris x86 and any of the BSD variants before I looked at Linux to implement some enterprise function.

Frustrated by other reasons…

In reply to I’m frustrated by unreal expectations

I think you are doing a big mistake here. Linux is way beyond the hobbyist OS, well, at least some of its distributions.

As for proprietary Unix variants, yes, probably they perform better in many respects, especially on multiprocessor systems, but at what cost? If it wouldn’t have been for Linux, or other free *nix for that matter, probably Windows would have won the servers running these OSes today.

As for Open BSD, Net BSD and Free BSD – these are also open source and free, not proprietary. They are more robust probably because they’re older than Linux, but that’s not the only reason, since the development and contributions to the source code is also different. As for Solaris, Sun open sourced it it a couple of days ago if I remember correctly…

I think that you misunderstand some terms

In reply to Frustrated by other reasons…

It’s clear that you didn’t understand what I said.

Your original post complained about some characteristics of Linux that are derived from it’s history and it’s current management. Linux development has always depended upon the interests of the people who contributed code to it. Linux does not have a dedicated team of full time developers creating a development road map for Linux. All of the other OSes that I mentioned do have developers working full time to manage the development of their respective projects. Therefore they enjoy the benefits of being more fully integrated than Linux.

I also said that some value added resellers have done a pretty fair job of making Linux a legitimate product. However, their efforts do not compensate for the lack of having a full time development team creating an integrated development road map for Linux.

The various BSD OSes that I mentioned have been available for as long as Linux has been around, but they have people employed full time developing their projects. They enjoy the benefit of having long term plans for future development, integrated installation software, full time people employed on proactively looking for and fixing vulnerabilities and bugs. Linux does not enjoy any of these benefits.

The various BSD OSes have been free for as long as I can remember.

Proprietary Unixes from IBM, HP, and Sun are expensive. You can get Solaris for free for personal use and for evaluation, but Sun still charges for licenses to use Solaris in a commercial environment. So your one valid point is that these OSes have always been too expensive for the system administrators and hobbyists to have at home.

Solaris is NOT open source. Sun is expected to formally announce an open source fork of Solaris called Open Solaris. This is NOT going to happen to the main Solaris development branch.

http://www.eweek.com/article2/0,1759,1699436,00.asp

So my point is and was that when someone complains about something in Linux being less polished or less efficient these are the result of Linux’s method of development. Linux still depends entirely on new code being developed by anyone that is interested in seeing something added or improved in the Linux system. All of the other OSes that I mentioned have development teams and project managers and development road maps. Linux has none of these. Therefore, and my point was, don’t complain about something in Linux that isn’t as efficient or as fully integrated as similar functions on the alternative systems. The lack of integration and polish are the result of the way Linux is and always was developed. Value added resellers have made Linux an easier product to use with their own contributions of installation and configuration software but it is still a hodge podge of independent contributions from many different people.

You really should drop in to the Debian forum

In reply to Frustrated by other reasons…

And see exactly how a lot of people are getting on with both the Stable Woody version and the Beta version Sarge.

There are constant questions on how to get things working and a multitude of responses for every question asked. Granted these are mostly for Sarge which is really a development version and they are all hobbyists who are asking and suggesting answers.

Woody just seems to work as the rock solid system that it was always designed to be and there have been no questions from business users about these systems.

So from what I’ve seen on the Debian Forum group it is still very much a hobbyist play toy despite what some mags might like to say. A lot of people are trying it out and because it is so different from Windows which is mostly the only OS they have ever worked with it is treated as a learning experience rather than what the mag writers are claiming.

While most of the *nix’s are newer than Windows because of the advances in hardware they still remain relatively old in comparison to the current hardware available no doubt the new PCI E video cards would suffer the same fate as the SATA HDD.

Col

Say What? Windows older than *nix?

In reply to You really should drop in to the Debian forum

Unix goes WAY back before Windows; its history starts with the DEC PDP-11 minicomputer. Comparison of disk write speed between Unix and Microsoft Windows fails on several counts, mostly, it is extremely dependent on what you are doing. Unix and its variants have a nifty feature that maps a file to memory; in that mode you don’t really open a file and load it to memory, technically it is “already there” but you treat it as if it is all swapped out to the disk. If you roll through it then it is pretty much the same as a conventional operating system copying a file to memory. BUT if you jump to a location in memory, the operating system will page fault and load a page of memory from disk, just that one page. It is very slick indeed, eliminating a lot of “lseeks” and so forth. Obviously with more than one program doing this, the operating system and disk controllers must be very smart. SCSI controllers are indeed smart and well suited to this sort of thing.

While I fully agree

In reply to Say What? Windows older than *nix?

The original poster was making a big deal about SATA write speeds and actually Linux a derivative of Unix is newer than Windows.

However if you want to go back far enough it was a Unix system developed by Xerox which was the first to use a GUI and when Xerox dropped the whole thing because Personal Computers would never be a big deal they brought in Steve Jobs Bill Gates and several others who wet their pants when they saw the thing and then had to rush out to develop their own system. SO I suspose you could go as far as saying that the GUI came prior to Windows om a Unix network bu tit was far more than just a simple PC in a box it was a complete network where all units where shared and used as required something that no one has managed to do since.

But what I should have said was that because Windows has in excess of 95% of the market most drivers are written for that OS so the ones that are written for other OS’s and actively discouraged by Microsoft are not as highly developed as the Windows ones.

Does that make you happy?

Col

I’m not so sure about that . . .

In reply to You really should drop in to the Debian forum

I don’t know for certain what it is you were trying to say, but it looks like you’re agreeing with “stress junkie” on the idea that Linux is really just a hobbyist’s OS. That’s simply not true.

Most of the aerospace engineering field runs on Linux workstations. Most of the Internet runs on Linux servers with Apache. Linux is a more-professional OS than Windows ever was.

Debian is a community distribution. That is to say that it’s supported by a community, not by a corporate backer. That doesn’t make it a “hobbyist’s” distribution. It is easily one of the most stable distributions available, and quite well suited to use as a production-environment server. In fact, speaking as an IT professional, it’s my first choice for server work.

you’re missing something there

In reply to I’m frustrated by unreal expectations

The “philosophy” of “Linux” isn’t “build it yourself”. The philosophy of Linux, if there is such a central thing, is that of open source software. Specifically, it’s “free as in speech”. You have the right to build or improve parts of your OS environment, to be sure, just as you have the right to benefit from the work of others, the right to give it away or sell it, and so on. Writing code for your operating system isn’t the whole point of the Linux effort. It’s just one of many things you can (relatively easily) do with it.

Telling someone “Hey, if you don’t like the driver, just write a new one, and leave me alone!” isn’t helping anything. It’s especially unhelpful if the person to whom you’re talking isn’t an excellent programmer.

I’m also not sure where you get the absurd notion that the Debian project is a “value added reseller”. Debian GNU/Linux is a community-based distribution. There is no “official” Debian paid-for version of Linux. If you end up paying for a copy of Debian, you’re paying a third party. Ironically, for the competent system admin, it seems to me that Debian is generally the best-suited distribution for work as a server OS in a production environment due to the extreme ease of system management and maintenance, and the stability and leanness of the distribution with an “out of the box” install. Thus, it may well serve very well in commercial uses, but the Debian distribution is not attached to any “value added reseller”.

The fact that you place such faith in non-Linux unices in terms of “performance” in some kind of sweeping generalization doesn’t do much for my impression of your knowledge of the subject. The only reason that non-Linux commercial unices like Solaris and AIX seem more “robust” in commercial use on the whole is that the deployment of such fully-proprietary systems is to a great degree under the direct control of the vendor. It is installed on hardware the vendor provides or approves, it is set up with the vendor’s input and often by the vendor’s employees, and the vendor keeps tabs on the system by way of a service contract in almost every instance of its use. Under such circumstances, Linux enjoys the same reliability. Even Windows almost performs like a server OS should in such circumstances.

Linux is often deployed on left-over, low-cost hardware that was never assembled with Linux in mind. Often, the OS installer does a better job of figuring out what hardware is present than the system admin has any hope of doing. Furthermore, in many cases where someone sees a failure of Linux to live up to their unreasonably high expectations for hardware support, the piece of hardware in question isn’t supported at all in (for instance) Solaris ? particularly since many major commercial unices haven’t even been officially ported to the x86 platform or, if they have, it generally hasn’t been declared ready for production environment deployment.

There’s more hardware support capability in the Debian GNU/Linux distribution than any other OS, as far as I’m aware.

[Author]

Replies