Enterprise Unix Roundup: Reliability Measured
Digg
DZone
Reddit
Slashdot
StumbleUpon
del.icio.us
Facebook
FriendFeed
Furl
| Main | In Other News |
  |
We're fond of the phrase, "lies, damn lies, and statistics." We tend to be cynical about data in general, but especially data that comes from sources we've long considered suspect. But deep down we're gearheads, and are lured to numbers like magnet to metal.
So with some trepidation we approached the "2006 Global Server Reliability Survey," which the Yankee Group released this week.
The survey compares various Linux, Unix, and Windows operating systems in areas of reliability, downtime, and recovery. We'll lay it out there and bluntly say that recent Yankee Group's Linux-Unix-Windows surveys, especially involving analyst Laura DiDio, have been suspect for their pro-Windows bend. This is largely due to Microsoft backing some of "studies" that proffered a favorable opinion of the Windows operating system.
Collectively, according to Yankee analyst DiDio, "individual corporate Linux, Windows and Unix servers experience three to five failures per server per year, resulting in 10.0 to 19.5 hours of annual downtime for each server." This is a marked improvement from five years ago, she added, crediting hardware improvements as much as software for this.
Beyond this, the results were revealing! shocking! earth-shaking! Er. Not really.
Unix, not surprisingly, was found to be the most reliable. Data centers running operating systems like Solaris, AIX, and HP-UX reported 10 hours of outage per server, for the year. Reliability and stability are the main selling points of these operating systems. The workloads they're processing require minimal downtime, and the 99.984 uptime they achieved is certainly laudable.
Assuming it's a 24/7/365 world, that is less than .032 percent downtime, or as we prefer to look at it 99.968 percent uptime -- not the sought-after five-nines of reliability, but not too shabby, especially when you factor in for backup systems that any enterprise worth its salt should have.
So all good, so far.
Windows Server 2003, to the operating system's credit, exhibited the highest reliability gains, with uptime increasing 20 percent from Windows 2000 Server to Windows Server 2003.
This, we feel safe saying, has as much to do with Windows shops being more on the ball and attune to viruses as it does to improvements in the operating system. Acceptance of autopatching has no doubt helped tremendously as well. (According to survey results from a similar study Yankee Group/Sunbelt Software performed earlier this year, 32 percent of Windows 2000 Server IT managers and 38 percent of Windows Server 2003 managers use automated Group Policy to apply their patches.)
Increased reliability and performance is always a good thing, so you'll get no complaints from us regarding Window's improvements.
What does raise our hackles quite a bit is how the Linux results have been parsed. Linux servers saw more outages in the past year. Why? Not for performance or reliability reasons, according to DiDio, who notes, "The one random element I can't emphasize enough is I think some of the disparity we see between Red Hat Linux down time and Windows and Unix comes not so much from any inherent flaws in the Linux core kernel, but the unfamiliarity of some of the network administrators with Linux."
Really? In 2006? We're not buying that. It doesn't jibe with Linux deployments growing like wildfire. In first quarter 2006, alone, Linux servers grew 17 percent and account for 12.2 percent of all server revenue, according to research firm IDC.
We do see a correlation between increased deployments and outages, however, and that is the fact that there are so many new deployments in server rooms and infrastructures still getting used to them. This is far different from a skill set issue.
We also think where these systems are deployed is playing a role. Linux servers are still by and large at the network edge. They process high volumes of traffic. They are clustered. They are in places where servers are particularly vulnerable to going down. It is not an apples-to-apples comparison.
And if lack of training is the culprit, so what? It's a problem more easily fixed than inherent operating system flaws. In this case, the-devil-you-know-argument is certainly not justification for sticking with an operating system not ideal for your infrastructure.
We hear the outrage, and while we're inclined to agree with much of it, we find the conclusions too faulty to be worthy of outrage.
» It cannot be easy being Novell these days. Here it is, trying its best to reshape its entire business model around an open source product line, and it just can't seem to catch a break.
On June 1, the Waltham, Mass. company reported its second-quarter financials, which were decidedly not good. Revenue was down to $278 million, from $297 million in the same quarter last year. Actually, this trended with market analysts' expectations for the company. What really got the market's knickers in a twist were the projected third quarter numbers, which Novell expects to be even weaker.
Based on the whack the company's stock price has taken since then, investors were not happy.
Novell's problems were perceived as even worse (if such a thing is possible) by the fact that its rival Red Hat's numbers look great. The stock has doubled in value over the last year, and there seems to be no end of good news coming out of North Carolina. This week Red Hat officially assimilated JBoss, with nary a burp. Well, maybe a small hiccup: Red Hat quietly dropped its own home-grown application server last week--something we wondered about back in April when news of the pending JBoss deal went public.
So, Novell can't even say that it's a lack of interest in Linux that's doing it in. But what is? We have often wondered aloud that maybe Novell's problem is that its trying to be too much to too many potential customers, in an effort to clone itself into another Red Hat. We would submit that there is only one Red Hat, and copying it is not the only path to success. Whatever path Novell chooses to try, it had better figure it out fast. The steady decline in business in an increasingly crowded enterprise Linux market (hello, Ubuntu) is a real problem. We have heard Novell sometimes says to potential employees, somewhat arrogantly,it a company full of smart people.
Let's hope it can put their business where their mouth is.
» Speaking of mouths, it is a widely accepted practice in the business world that if you are speaking for your company, it's a darned good idea to get your facts straight before opening said mouth. Based on what seems to have happened with hardware manufacturer Levono this week, that lesson may still need taught.
Last week, Frank Kardonski, Lenovo's worldwide product manager for the Lenovo 3000 product line, had an interview with CRN, in which he was quoted as saying: "We will not have models available for Linux, and we do not have custom order, either. What you see is what you get. And at this point, it's Windows."
Naturally, CRN followed up and confirmed this little bombshell, getting another company spokesman to confirm that this policy included the ThinkPad line of laptops Leveno had acquired from IBM last year. Given IBM's pro-Linux stance, the irony was not lost on observers.
By Monday, Levono was denying this policy, indicating that Kardonski was mistakenly referring to the licensing program. Apparently, Levono does not acquire licenses for its preloaded software; it has the customer do that. Kardonski meant to say that Levono would not supply Linux licenses. The company would still support and pre-install Linux on machines based on customer requests.
Er, right. Given that CRN confirmed this from more than one Lenovo source, we find it a bit difficult to buy into the reason behind this mix-up. Not knowing the difference between "no models for Linux" and "no licenses supplied for Linux machines" is a bit of a stretch. Particularly, the Linux community was very quick to note, when Lenovo just announced last April that it was buying $1.2 billion worth of Microsoft software.
While we don't always buy the booga-booga conspiracy theories about Redmond, this one seems a little too likely. Given that if there's a hint of evidence that Microsoft is blocking opponent's deals by strong-arming other vendors, the U.S. Dept. of Justice might come calling (OK, we realize that's mostly a theoretical exercise right now), it would explain why Lenovo backed off a stance that may have been the actual party line. Instead of a miscommunication.
» Web 2.0, that almost copyrighted buzzword, is used to describe anything new and cool on the Internet these days. That has to include Ajax, the almost mystical combination of asynchronous XML and JavaScript that has captivated the hearts of developers and users alike. And, hey, we're not knocking it, either. When Ajax apps work, they bring happiness to our inner geek.
But one of the things that troubles the movers and shakers in the world of IT is that Ajax is an implementation without a standard. Which essentially is like putting out a full picnic basket in front of a certain hungry bear. A standard-less tool means someone has a shot of getting their version of said tool to be the standard. And if you're wondering who the bear in the tie and hat is, look no further than IBM.
On Tuesday, Big Blue announced would be partnering with the open source Dojo Foundation, providing open source code to the Dojo project — specifically the namespace functionality that IBM already has in its own Ajax toolkit. IBM has it own internal Ajax toolkit, which does have full namespace support. Boloker noted that some of the lessons learned from IBM's development of its own toolkit will be taken and moved into the Dojo project.
This effort is related to the foundation of the Open Ajax initiative in February and the recent inclusion of Ajax framework tools into the Eclipse development toolkit. It seems, BooBoo, the Ajax standards picnic basket is in the bear's sights, and Ranger Smith is nowhere around.
» The winners of ServerWatch's Product Excellence awards were announced this week. Linux-based and open source products fared well. Some highlights:
- JBoss picked up 35 percent of votes in the app server category, taking top spot along with Oracle Application Server
- Jive's Wildfire dominated in the real-time communication category.
- Open-Xchange finished well ahead of competition that included IBM Lotus Domino and Microsoft Exchange Server 2003.
- When it came to open source communities, Jabber and Open Solaris were ServerWatch readers' favorites.
Congratulations to all of the winners!
>> To Main
