What is Digg?50 Comments
- ThinkBox, on 10/12/2007, -0/+6THe same price as 1% downtime
- swaxhog, on 10/12/2007, -1/+6Here's a little cheat sheet for uptimes:
90% - 876 hours (36.5 days)
95% - 438 hours (18.25 days)
99% - 87.6 hours (3.65 days)
99.9% - 8.76 hours
99.99% -52.56 minutes
("five nines")
99.999% - 5.256 minutes
Finally,
99.9999% - 31.536 seconds - randylovin, on 10/12/2007, -0/+1Can someone quote the part of the article where it said how exactly Walmart upgraded it's downtime, other than the fact that "it did" ?
- nickster, on 10/12/2007, -0/+1Wal-Marts Network is impressive, I know the US government has gone to their central mainframe just to see how they have their security setup.
- inactive, on 10/12/2007, -1/+2"I think Wal Mart needs to upgrade how they treat their employees so I don't have to pay for the food stamps of their full time employees."
I've known three different Wal Mart employees, and they were paid good money and treated just like anyone else I know. - BlueLaser, on 10/12/2007, -0/+1While socket may have achieved 100% uptime, his "shoe string" must be made of gold. There are known figures for the average cost of increasing the availability of a system from 5 nines to 6 nines to 100%...and those numbers are in the millions.
There is no such thing as cheap 100% uptime. Those two concepts do not belong in the same sentence. - rigorious, on 12/10/2008, -0/+1it means:
"percentage of uptime - downtime" - CorpT, on 10/12/2007, -0/+1I also did this on a shoe string budget != It's called multiple geographically diverse sites with internal network with no single points of failure. It's called multiple load balanced fully redunant layer 7 switches. With, again, grographically diverse round robin DNS. Then add in 5-6 tier one network providers at each site.
- BlueLaser, on 10/12/2007, -0/+1I don't think it is accurate to say "no one" has 5 9s. You can certainly achieve 5 9s, it will just cost a lot! You can have redundancy at every point, patch in shifts, shadow copy your data, etc.
Is it hard? Does it cost millions of extra dollars? Yes and yes. Impossible? No. - socket, on 10/12/2007, -0/+1I've ran plenty of operations for years at a time with 100% network uptime. This also included any maintenance. I also did this on a shoe string budget.
- Nightfall, on 10/12/2007, -0/+0At some places, 99% uptime isn't good enough. With redundant servers and systems in place, the best I have done is 99.9% in a year, but that was for a hospital. Man, did they spend a lot of money on their IT department. :)
- bigred, on 10/12/2007, -0/+0Finally,
99.9999% - 31.536 seconds
WTF does this mean?
--
It is how many minutes or seconds of downtime per year you would have to be under to achieve those uptime percentages. So for 99.9% uptime, you could be down for 8.76 hours a year. - CaughtThinking, on 10/12/2007, -0/+0Not a very impressive article. I was expecting a detailed analysis with regard to cost and scale, and I got "oh wallmart is pretty big!" no digg.
- TwoDotOh, on 10/12/2007, -0/+0"Alistair Cockburn."
I will be chuckling like a junior highschooler for hours. - spadin, on 10/12/2007, -0/+0I believe the chart means the amount of downtime allowed in order to be able to say, "We have 99% uptime" or 99.999%... whatever.
You can see that 10% of 365 days is 36.5 days which means out of one year the network would be down for more than a month. That's not good. - Double-Z, on 10/12/2007, -0/+0socket, no need to insult me. You seem to know what you are doing, so I admit you may have achieved it. It's rare, but if you say so, I'm happy to believe you.
No need to resort to insults dude. :) - inactive, on 10/12/2007, -0/+0i think we've always had "virtual" 100% uptime by just keeping several smaller sites-often just a good dsl... and doing load balancing. its not impossible for all of them to be down at the same time but its fairly unlikely. might get slow for a while. then just get heavy duty but consumer level batteries to smooth out the bumps and keep the servers online for short outages... its not expensive compared to a single point of failure architecture.
- RHamel, on 10/12/2007, -0/+0Does Walmart need that sort of QOS? Do they lose the sale, or does the customer come back later to make the same purchase? I've worked at companies that require this type of uptime, airline reservations. The secret is redundancy and quality of staff. A good place to start is to find good management. Not easy to do, most are morons.
- Double-Z, on 10/12/2007, -0/+0[rant]
What I hate is when you prove to your host that they haven't achieved the uptime they promise, that there is some clause in the agreement that gets them out of thier promised uptime.
[/rant] - rooke, on 10/12/2007, -0/+0I'm with randy. I went to read about how wal-mart upgraded its up time.
- Double-Z, on 10/12/2007, -0/+0You are fooling yourself. No ISP even has 100%. Your server may have been up, but I bet your net connection went down a few times.
(umless you mean on a closed network, where this is possible) - ender78, on 10/12/2007, -0/+0Lets remember that all of these numbers often exclude regularily scheduled maintenance. Equipment does need maintenance and will have to be taken down from time to time. When you have complex apps or services that rely on dozens or hundreds of components, 5 nines is difficult to achieve. Such uptimes has so many clauses that 99.999% uptime is rarely achieved.
- GMTao, on 10/12/2007, -0/+0Blasted article doesn't say squat about how much Wal-Mart has spent on their network for their uptime. - No Digg
- vdxc, on 09/29/2008, -0/+0gherikil, those are all lengths of time that are of downtime. so eg. a web host promises 99.999% uptime, that means they must have less than 5.256 minutes of downtime per year
- ender78, on 10/12/2007, -0/+0I don't believe that people understand what 100% uptime means. That means that a process never fails, software never crashed, hardware never fails. Assuming that you have a server with two NICs is connected to two switches, NIC A sends a frame to Switch A, while the frame is still on the wire, SWITCH A is reloaded, since that packet or next set of packets is dropped, you've just blown your 100% uptime. Pseudo 100% uptime, where the application appears to be up to the use base 100% of the time, is still not 100% uptime. All clustering and fail-over takes time [SONNET has a 50ms fail-over time], 50ms may not be much but it is an outage.
- CorpT, on 10/12/2007, -0/+0Obviously socket is the coolest guy here. His Network >>>>>>>>>>>>>>>>>> Everyone Else's network. Oh, and socket >>>>>>>>>>>>>>>> Kevin Rose.
- randylovin, on 10/12/2007, -0/+0gmtao said "Blasted article doesn't say squat about how much Wal-Mart has spent on their network for their uptime. - No Digg"
Finally, someone who agrees with me, along with rooke.
The OP's editorial sucks. - marillion, on 10/12/2007, -0/+0According to Walmart's 2005 AR, they brought in $285 Billion - With a "B".
This is $781 million per day. $32.4 million per hour. $9 thousand dollars per second! Settling for 99.999% uptime means giving up $16 million dollars. - CorpT, on 10/12/2007, -0/+0Ender, I don't think you get it. Socket can get 100% uptime because he is a networking god. In fact, he is so good, he actually changes the laws of physics so that the frames travel instantaneously so that packets are never dropped because of a switch reload. Geez. Why don't you believe he has 100% uptime?
- socket, on 10/12/2007, -0/+0You can reach 100% uptime if you leave NOTHING to chance and have a good plan. You also have to define uptime very specifically. Like if a routing bump cases 10-20% of your customers to not have access for 3-4 minutes (while a BGP adjustment is made), I don't consider that down time.
- socket, on 10/12/2007, -0/+0CorpT: If you knew how much that would normally cost and how much I did it for you would agree I was on a shoe string. Not to mention how much the company spent on a year on advertising vs. IT budget. Next Please.
- socket, on 10/12/2007, -0/+0And yes the Digg sucked. Wow thanks for letting me know the world wont end if I can get to boing boing. Not surprised to see a tool like Kevin Rose dugg this.
- Double-Z, on 10/12/2007, -0/+0socket, my company serves educational content to over 1000 schools and colleges in the US. We have multiple servers at several locations. We serve several gigs per hour, but even though we pay over 100k per year we still have outages due to the ***** UK internet backbone (thanks, BT). Hence my previous comments.
Still, no need to insult me. - OutKast, on 10/12/2007, -0/+0"I also did this on a shoe string budget != It's called multiple geographically diverse sites with internal network with no single points of failure. It's called multiple load balanced fully redunant layer 7 switches. With, again, grographically diverse round robin DNS. Then add in 5-6 tier one network providers at each site.
ROFLMAO!!!!! - enzomedici, on 10/12/2007, -0/+0
Did you ever upgrade the software, the OS or the database? Doens't matter if the hardware has no single point of failure or if you have 5 colocation sites. At some point you need to upgrade or patch the OS, application and database. Even if you switch between colo sites, you need to switch an ip address somewhere. You cannot round robin to a site that's down.
5 9's uptime of 99.999% or about 5 minutes is a myth. Show me the company that does and I'll show you one that's not patching their OSes and databases regularily and probably have security holes the size of Texas.
In Oracle you can run RAC and do minor patches on a server at a time and acheive high level of uptime, but you cannot do major versions. You can switch between multiple colocation sites if you want. But, going from Oracle 9i to 10g for example, will still require some downtime, even if it's just 10 minutes. That blows the 5 9's out of the water.
Nobody's got 5 9's, they're full of *****. There's always a software single point of failure. A swap file, a run away process, a bad administrator.
If they are confident of their 5 9's uptime, then they shouldn't have a problem letting me rm * as root on a few of their boxes and letting their failover kick in.
We'll see how they do. - jasqwerty, on 10/12/2007, -0/+0LOL @ enzomedici
You must be using VOIP dude?
By LAW landline phone companies must guarantee 5 9s uptime, barring wildly unforeseen circumstances, such as nuclear annihilation. The phrase 5 9s even originated from this issue. - MikeMacMan, on 10/12/2007, -0/+0Misleading title... no digg
- nonchallant0819, on 03/28/2008, -0/+0This is a great story... found this one through http://www.google.com
___________________________________
http://www.TopNotchCarpentry.com - enzomedici, on 10/12/2007, -0/+0>By LAW landline phone companies must guarantee 5 9s uptime, barring wildly unforeseen circumstances, >such as nuclear annihilation. The phrase 5 9s even originated from this issue.
Right. And when an LA radio station has a contest to giveaway 10 free Rolling Stones tickets to the 49th caller and the circuits max out, that's not 99.999% to me or the other thousands of callers. 9/11 same thing happened. Land lines are certainly not five nines uptime. - wwwdeveloper, on 10/12/2007, -1/+0"Now what if Delicious, Feedster, or Technorati goes down for 30 minutes? How big is the inconvenience of not being able to get to your tagged bookmarks or do yet another ego-search with Feedster or Technorati for 30 minutes? Not that high. The world does not come to an end. Nobody gets fired."
- Ignorance. These may not be ecommerce websites, but they DO lose revenue!! I am sure that being down for 30 minutes and loosing THOUSANDS of hits puts a bruise on the income being generated from ad revenue.
- NO DIGG. Get off your soap box and provide concrete examples of your statements. - socket, on 10/12/2007, -2/+1Double-Z: No ***** you're wrong.
The customer requested 100% uptime so that's what I gave them for over 3 years. It's called multiple geographically diverse sites with internal network with no single points of failure. It's called multiple load balanced fully redunant layer 7 switches. With, again, grographically diverse round robin DNS. Then add in 5-6 tier one network providers at each site.
I never said I had a network that never had a piece of equipment fail you boner. Learn to read child... - lowkey, on 10/12/2007, -1/+0thousands of dollars? Nope. Hundreds maybe. Linux + failover clustering. Use cheap hardware to achieve the same results.
Work smarter not costlier. - Double-Z, on 10/12/2007, -1/+0socket, no need to insult me. You seem to know what you are doing, so I admit you may have achieved it. It's rare, but if you say so, I'm happy to believe you.
No need to resort to insults dude. :) - jackspack, on 10/12/2007, -2/+0and they also fail to mention wal-mart is a microsoft shop
(arh arh! use firefox, opera, it r0xr3s teh!) - joshGwyther, on 10/12/2007, -2/+0Who doesn't know this? I hate stories that state the obvious.
- gherikill, on 10/12/2007, -3/+0Here's a little cheat sheet for uptimes:
90% - 876 hours (36.5 days)
95% - 438 hours (18.25 days)
99% - 87.6 hours (3.65 days)
99.9% - 8.76 hours
99.99% -52.56 minutes
("five nines")
99.999% - 5.256 minutes
Finally,
99.9999% - 31.536 seconds
WTF does this mean? - kevinmoore, on 06/13/2009, -3/+0Wal-Mart: Always Low Wages. Always.
- tomaburque, on 10/12/2007, -3/+0I think Wal Mart needs to upgrade how they treat their employees so I don't have to pay for the food stamps of their full time employees.
Check out the new & improved 
© Digg Inc. 2009
— Content created and posted by Digg users is