107 posts • joined 7 Mar 2008
Cloud hyperscale growth
"The additional growth in ODM Direct sales, spurred most recently by a native data centre expansion by some of the largest device providers, will help the server market reach that tipping point faster,"
If you need 20K identical commodity boxes/year you can just set up your own production than buying it from the likes of Dell.
Re: What the hell did they expect?
"And how did Linux approach this? Linux's approach shows it doesn't matter if MS had performed as well as Linux, since Linux isn't taking over the desktop marketplace."
The challenge is that most people find their current desktop just as good and do not want to upgrade.
This is what happens when software becomes mature. Then you switch to yearly licensing instead of a one time purchase.
Apple Cube 2
Recall the Apple Cube from 2000. Also a very nice design, but the premium price made it fail. There were also some issues about overheating because it was fanless. And the performance was not that good.
I do not care about the design of my box. it is performance and noise that counts.
I like to hide the box in a closet somewhere out of sight.
Re: I keep looking at these NAS devices
If you want to have all the features a typical SMB-level NAS has, that'll be a six hours chargeable time to have it in a turnkey state for the user - maybe two or three if you are only throwing SMB, NFS, SMTP, an intelligent UPS on it, including installation.
Most linux distros have this preconfigured, you only have to turn on some systemd/chkconfig flags and enable some filesystem. On opensuse it takes a few minutes to format disks and configure samba and nfs. What I hate though is their apache configuration with 20+ different files.
What I really like is webyast with makes you configure the entire server via web,
Fortran with openMP
I've not written much fortran code, but I've benchmarked it and parallelized it using openMP.
As long as you could identify the serial and parallell part and assured no variable dependencies outside of the parallel part it was just to use the openmp pragmas. worked like a dream.
!some OMP pragmas I've forgotten:-)
do your stuff
!$OMP end parallel
Re: @ Eddy Ito
>Burglary 101: leave the car engine running to facilitate a fast getaway.
In Yakutsk, Siberia where the average temperature is -41C in january. On the coldest days
people leave their cards running while they're on work to ensure it is nice and warm for the trip back home and to assure it starts. Bet what that is doing for the climate.
vmware and centos makes my day
CentOS6 is the best host. Linux has way better firewall and routing capabilities with route, openvpn and ipsec. I've had all problems with routing when using windows as host. And it is 100% compliant with RHEL so all commercial linux software supports it. It also do not have the dreaded systemd (yet).
I develop commercial software used by banks/processors requiring me to log into their systems via all kinds of vpn software and citrix. I mostly use windows xp as the guest for it. some of the guests have failed when running on a windows host, but never on a linux host.
Our software is cluster based so I have lots of centos5/6 guests with dhcp along with database servers (DB2/Oracle express ) and websphere MQ guests for delivery. I use the express editions because they are compliant with the enterprise editions but I do not have to care about the licensing and lower requirements, so copying vm-images around has never been easier. I keep separate guest-vms for each customer emulating their setup. this way I can build rpms that require just a stop of application, rpm -Uvh new package and start. (usually 20sec) in production.
When it comes to the virtualization I use vmware workstation9, it works fine with the rhel6 kernel and everything works. even hwrendering (I use the binary nvidia drivers on the host) of windows games like skyrim on win7 guests works excellent.
Years ago I used the vmware server v2, and afterwards I was forced to use Xen and virtual box. (didn't want to pay for linux software) they work fine for stuff that do not requite performance. Our software is 25%-50% faster on a vmware guest compared to virtual box guest. So it is a very easy decision to use $179 for a vmware license. vmware handles io _WAY_ better than xen and virtual box.
Personally I love opensuse, but their upgrade policy makes me have to reinstall every time there is a new relase, so I have some opensuse12.3 and 13.1 guests I build and run our software on to make sure it works. with CentOS it is just yum update(ex: to upgrade from 6.4 to 6.5)
and all your binaries still works.
I've created startup scripts which starts my database and websphere MQ guests automatically on reboot. I have a samba share on the host I share among all the guest.
Here I have the source build and release directories and a webserver so I can get hold of all the software nomatter if the guests are down and I'm on a customer vpn guest and need to install software built on an other guest that now is down.
Every guests except the database ones I keep under 20GB in size, that way makes it very easy to backup and move around to other vmware instances like my laptop, home computer or to ESX/vsphere servers. I have base guest images with the base software requirements so designing for a new scenario is just a clone away (with SSD cloning 20GB takes 20sec).
Earlier I used to have several noisy boxes under my desk. they are now all in a single almost silent workstation. I have around 30 different virtual machines where up to 8 running at the same time on a 32GB Intel i74770 512GB SSD with a /shared sata disk of 2TB.
>Brit upstart ARM for a pittance, for example.
AFAIK ARM has been around since 1985, A very slow upstart to say the least. I believe they have licensed IP cores since atleas 1992.
We develop on CentOS, customers buy RHEL
All our customers run RHEL but we develop and build rpms on CentOS. It is binary compatible no need for compilers etc. Many critical systems are not allowed to have compilers installed. This works like a dream and we do not have to buy expensive RHEL licenses.
Another issue is that Opensuse and Fedora changes way too often. CentOS 5.0 was released in 2000 and I've still just 'Yum update' it up to CentOS5.10 without any problems. That is almost 13 years of patching the same binary compatible linux system. Impressive.
I've always hoped someone would do the same with Suse Enterprise. Which I consider being superior to RHEL. But their licensing is 60 day trial. No access to the patch and update service after that. It is just too much job when you have 30-40 vms lying around. Running all kinds of strange instances you might need.
Old legacy software You may suddely be paid to update and then you have to reinstall just to get the latest patches.
"bare sorgen" is a norwegian saying meaning everything is just shit and cant be worse.
I bet the hardware vendors are happy - not
Revenue but not much margin:
I knew IBM had problems with their SmartCloud when they bought SoftLayer. Amazons S3 is way much more advanced and feature rich it is like vmware ESX against first version of Oracles virtual-box. IBM is probably the vendor with the most to loose by moving to the cloud. In the cloud everything needs to be commoditized and streamlined in order to save cost and manpower. IBM wants to deal with your large problems and make a premium on it.
This is because of the Tegra4, I bet lots of other tablet will arrive using that chipset also, then they all will have this resolution.
Just first iteration
The first generation of a new technology will always have shortcommings. If we see the same advances in head up googles than in mobile phone screens like in the last 10 years. I bet the future is looking so bright you have to wear shades. As long as the initial interest is above the investment threshold, there will be money for inventing the technology further. I bet both Apple and Samsung are putting lots of money into this and just talking the technology down until it matures into consumer quality and pricerange. then it will really take off.
sysadmin that monster
I'm unable to understand the size and complexity of those monsters, but I really like to know how they sysadmin those. How they provision and load balance applications and users. How long does it take from initial load until execution when you request 10000cores. What kind of checkpoint/restart software do they use.
I once wrote an MPI ping which I frequently ran on all large clusters and shmems I was able to get my hands on (I used to work with SGI HPC 10 years ago). And with the increase of cores and nodes the bandwidth and latency did suffer. On large systems it was very easy to just congest IO unless there were strict limitations on my user. Broadcast pings where out of the question, closely build up a 1to any core ping and build carefully selected latency and bandwith maps for each core to any other core in the system. And then try out 1to2, 1to3, 1to4 etc until peak bandwith was reached, then 2to2,2to3 etc and so on then at the end lookup the ones with the lowest overall latency and highest bandwidth combinations to find the system bandwidth.
AMD not to push Fusion Android tablets
I want an android tablet with a fusion gpu/cpu. Rumours say AMD is not interested in android and is all behind win8. IMHO a terrible mistake: win8 on netbook/tablet is sinking like titanic. AMD have sacked all their linux developers. And Why not a Fusion APU in some of the 150mill smartphones that ship with Android each quarter. So they are effectively abandoning the by far largest growing market where they may have a superior product. Something is fishy, either M$ has done some behind the scene investments into the empty koffers of AMD or its just incompetent management.
Re: How about a stock buyback??? Seems like smarter move...lots smarter than giving in to Loeb
You cant buy back stocks with money you do not have. A stock buyback is great if You have money in the bank with no idea how to spend (Apple, IBM, Intel etc). But Sony is deep in the red, so they have to wrestle out all the vaules they can find.
Re: Why so long?
After installing the code I see it waits 10 sec between each shoot so everything becomes static, that is why it takes so long 10 sec every move, would have been nice this was shortened down. but AI need to be smarter then to shorten, not much brute force then.
Re: Why so long?
Or is it trying the level multiple times to see what works and what doesn't? If it is doing this why does the article not mention it?
It tries shot after shot and sees how well it goes, sets angle and power and fires away.
This computation they have shortened down from 10sec to 1/10 because they have probably taken away the presentation and go directly onto the physics computation of the impact.
The nice thing is that they do not bend the rules just try and try, although it is not much AI in this because it is easy to compute the successrate for each shot. Try 10 shots then take the one that destroys most objects and go to next shot ets. it is static and not more than 5-6 birds. This is no chess solver like deep blue.
But it is cool and fun to watch.
Re: Linux as a desktop is a failure, its time to move on
The only program I've missed from windows was Visio. Now with libreoffice4 we now have a much better draw package. And with Steam now accelerating on linux. I would say the future has never looked brighter for the linux desktop.
Re: Is this some new Microshill tactic?
suse have suse studio, I use it to set up rpms for redhat because my big do it all rpm of our commercial app spec file just fails on RHEL6. After making the spec file through suse studio, it worked. So I develop all my company commercial code in open suse, then when it is finished I verify the code with suse studio and build for centos5/6. Very nice piece of software.
>But I'm fond of jfs as a file system, which for some reason Suse stopped supprting some time back.
JFS is in the 12.3 filesystem list, it is not the default though which is ext4, I myself always uses XFS.
I've had lots of cheap disks failing, but I've always managed to get xfs_repair to restore the disk into a readable condition. much faster than Spinrite.
Thank heaven for virtualization
I used to have something similar, but now I have lots of vmware images on one quad core 24GB workstation.
disk-share on host, and all kinds of linux (opensuse 11.0,11.4,12.1,12.2,12.3,centos5,6), winxp for all kinds of vpns for work and a win7 for gaming, different work images of our software so I can test at home. etc all in all I have 23 images on 2x2TB disks all running on vmware workstation v8.0.6. Fortunately I do not need to run more than max 2-3 vms at once.
I used to hate the frequent upgrades with OpenSuse, because I have a rather tricky wikimedia setup with some heavy configured plugins and getting this to work on a new system take a lot of time. Now I Just clone the image and start upgrading while the old one is running until it is ready. if it stalls I just start on a new clone.
Gaming in win7 vm on a linux host works fine. The most demaning game I've played have been Skyrim.
I played to level 67 and put 135 hours into it, so it was a fine experience.
But You need to have a nvidia card, I tried an AMD card and then it just froze, vmware complains saying there are issues with this card/driver on a linux host when you start the vm in accelerated mode.
I consider the $179 vmware license well spent money and well within a home budget.
I've used Xen and Virtual box also, but they are not in the same ballpark. especially considering performance and hw-accelerated gfx on guest vms.
HP get over it
Itanic is dead and has been for years. How long can you beat a dead horse?
The revenue stream from itanics dwindle day by day. There cant be much left.
if they jumped ship when this was announced they would have been in much better shape now.
There is a marked out there for linuxboxes with the same feature set as itanic superdomes,
Couldn't be that hard. SGI moved from itanium to Xeons in 2 years on their Altix line.
Nobody is missing the itanics. it was a bad design that lived way too long.
The slow single thread performance and poor intel compilers will not be missed.
Re: No love for Kubuntu . . .
> KDE 3.5 was a fantastic desktop. I still use it in places.
> Then KDE 4.0 came along. And it was *terrible*.
KDE4.9 is fantastic. I've been using it ~8h/day since beginning of September and it has not crashed once.
Now you can move all icons around from anywhere to everywhere. The shortcuts are stable and always working, which was not the case earlier. Now I'm watining for KDE4.10
I rather use Samba
Why do you get only 35MByte/s write performance? Your network must be saturated.
I do not see the point in testing a NAS array when the network is the limitation.
If the Drobo is the limitation it is a bad selling point.
I tested my samba share from my fanless low end linux box exporting a cheap WD 2TB green disk and get 70MBytes/s write and 85MBytes/s read from a win7 client.
I only have a D-Link dir-825 wireless/gigabit router in between.
IMHO the samba swat web interface is just as easy as the Drobo.
Nice tech, but what about linux?, should be easier to create linux runtime.
In Norway we turned of analogue May 20, 2008. I do not miss the snowy pictures for sure.
Appstore > 500K, Android Play > 500K, Nokia MarketPlace ~ 80K, WP marketplace ~ 100K
Is Nokia symbian apps? or just a clone of WP marketplace?
Why should anyone develop for Nokia and WP when they only hold around 2% of the market.
This is already a lost battle. RIP Nokia(^D^D^D^D^D Microsoft Phone Subsidiary)
Oracle Linux and RedHat
"We think Sparc will become clearly the best chip for running Oracle software. At that point we'd be nuts not to move Oracle Enterprise Linux there. We're a ways away, but I think that's definitely going to happen," Ellison said.
OEL is just a rebranded, recompiled RHEL with some kernel patches and settings suited for oracle software.
Nice try getting that onto SPARC without RedHats Approval. They can make their own distro and not scavenge upon others. Or do they actually send a tiny fraction of all that OEL support income to RedHat, I do not believe so.
I'm not allowed to use it
Installed amazon appstore, but since I'm in a backwater 3rd world country (Norway), I do not have a valid purchase country. I can shop in the Kindle Store on the same device though via android kindle app on same account. Maybe Amazon assumes all appstore customers use Kindle Fire.
Re: Will the Iranians get the fixes?
Even though Siemens is not directly involved with the Iranians, they surely have a subcontracted subcontractor somewhere. All so obscured no one can find the connection.
We all know the 10x boost comes from replacing sparc with x86_64.
I'm still a linux netbook fan
I'm still perfectly happy with my EEE PC 1018 which cost me around £210 with 2GB ram and a cheap 80GB SSD.
Battery lasts on average 11hours. it boots in 4sec from hibernation, instant from suspend.
I can have it in suspend for weeks.
And it runs IMHO the best OS in the world OpenSuse 11.4. (not 12.1)
The screen is not good compared to the IPAD3 though, but it browses much faster, maybe
because the 2GB mem is much faster than whats in the ipad.
it runs citrix and vmware so I've access to all corporate software and everything I need to do my job developing software and do presentations in case of emergency.
Oracle on SPARC/SOLARIS is very competetive
From customers I've heard that Oracle is pushing SPARCS to run their Enterprise Database.
Although it cant match Power6+ or latest X86_64 on single thread speed, the Sparc architecture still offers great benefits when it comes to hardware redundancy, storage management and a rock solid OS in the bottom with some nice features you will never see on RHEL (DTrace, ZFS, Fault manager, Service Manager to name a few).
They discount the oracle core licenses so they become competetively regarding price/performance.
The hardware price is nothing compared to the $47.500 per processor list price, and AFAIK no other processor beats the 64threads on T4 or 128threads on T3.
Add up the discounts and the most attactive system to run a large oracle database is on Sparc/Solaris.
The next SPARC T5 will have 128cores and 16K threads and IO backbone to sustain it.
Singlethread will be mediocre but no one else
I work with databases that grows with >10mill rows each day, and I've never seen any need for faster cpus, it is mostly IO and memory. If a single thread is slow, just run them in parallel and you get linear speedup.
When you run queries you need lots of indexes to back you up which makes you IO bound.
Why do not HP just rebuild their superdome hardware to use xeons, like SGI did with Altix?
HP should have ported HPUX to X86_64 years ago, or switched to linux.
The magins can't have been that good?
Re: Poor contrast on e-readers
I have a kindle4 and it is as crisp as it can be, no problems there. what I miss is the backlight of a tablet so you can read in the dark.
SGI buildings 40-43
The cafeteria looked better when it was part of MIPS/SGI campus and was called Cafe Ozone.
It cost almost nothing to eat and the quality of the food was exceptionally.
SGI sold the buildings to a joint venture firm for around $800mill and leased it back. It financed the company deficit for years.
Re: ARM Emulator
Interesting. They have to do _MUCH_ better than their last effort:
I recall the X86emulator on Itanic. On the first revisions it was made in hardware, later they figured it did perform better in software and saved silicon.
It had one main shortcoming, it was tied to the core frequency of the itanic, and at 600MHz while the pentiums where at 2.4GHz it was way too slow to be usable, even for testing.
AC power and computers
I recall my days at SGI, During 1997-2000, Every year end of May we had to replace 3-4 powersupplies in the University of Bergen HPC. On the 3rd year we found this to be a travelling circus/tivoli connecting to the same power source nearby. It showed up yearly for the national holiday.
Show me your stinking patents
Do anyone have a list of the offending patents. I tried to google and did not find any?.
When I install opensuse I only get one license for truetype, but where are the other 234?
An emulator wont hurt
I do not see why an emulator will hurt them, it would be a very valuable tool for developing and testing software. They should ease the access to mainframes to help porting software.
I've been asked by our IBM outsourced customers to test/port our software to AIX and eventually also Z/OS. IBM provides access to hardware for free. And running AIX/ZOS seem to be cheaper than linux when outsorced to IBM, when you already have those already.
You can get far better IO on most other high end unix boxes, they just do not provide so many levels of fault tollerant microcode as the mainframe. Something IBM touts you head full of.
I've tested hot swappable memory in X3590 x86 servers and it is slow as hell, almost unusable. If you need this feature you have to sacrifice latency or pay the extra for the mainframe.
I use AIR on linux
In order to use WIMP (a music streaming service like spotify). I had to install AIR and it works fairly well. I have to go back to piracy again then. Fat chance Wimp will make a native linux player.
When flash works on linux, there cant be much work supporting AIR, it is just a runtime for flash outside the browser.
simple used to be faster
Legacy Fortran code is fast because the F77 compiler was really simple, no function stack and no dynamic memory allocation. Just fixed arrays and static libraries for everything. basic stuff like recursion was not allowed
Pure C-compilers used to be fast also before all the OOP stuff was added.
And not to mention all the kernel, multitask, GUI, interrupt libraries you have to incorporate.
For many simple tasks todays computers are too complex.
Why all this talk about clouds. I want sunshine and long lazy days on the beach.
Why not try and get it even faster, port it to fortran.
overkill and expensive
I did not want yet another hot noisy harddrive in the bench under the TV, So I purchased an Apacer AL460. Along with an old usb wlan stick, it streams hd from a local server via a samba-share.
It cost me £45 and is completely silent, small and lukewarm. I'm unable to see any difference between this and the PS3 on playback quality and it plays all format and subtitles I have collected during the years.
I only see the need for a mediapc if you do not have any other pc in the house that can act as a server. I rather spend the money on a vanilla server where you can get 5-6 disk drives. You better soon need the space.
Octane3 is rather nice
I think the Octane3 10-node deskside cluster is probably the most sexy box out there. Xeon or atom(?) processors. The original driver only support quadro gfx cards, but I know from reliable sources that geforce 440 fanless works fine.
No it will be Bangalore.... we have a problem.
cheap numas are useless
IBM has Xseries 3850 and 3950 which they claim are numa, but if you connect more than one box together and use memory cards, the memory becomes so slow that my eeepc901 showed better latency. You are then far better off with a cluster of cheaper nodes.
My experience from this, is to run the stream benchmark on all new configs I get to deploy on and then just reject them if memory becomes too slow.
If you need numa you need to buy something that give you almost the same memory latency for all nodes, like the SGI altix. It doesn't cost you so much more than all those crappy numas around.
I have for the last 10 years been using palm devices, now I've unfortunately moved on to Android, but I love using the stylus instead of my butterfingers. I would have loved seeing more stylus apps on the android, and also a place to put it on the device.
7inch is like heaven with a stylus.
And where is that grafitti1
- Vid Hubble 'scope scans 200,000-ton CHUNKY CRUMBLE ENIGMA
- Bugger the jetpack, where's my 21st-century Psion?
- Google offers up its own Googlers in cloud channel chumship trawl
- Interview Global Warming IS REAL, argues sceptic mathematician - it just isn't THERMAGEDDON
- Apple to grieving sons: NO, you cannot have access to your dead mum's iPad