Slashdot Mirror

This is why the greenies roll their eyes when the nukies say "Trust us, we know what we're doing!"

And the rest of us roll their eyes when the greenies expect us to roll back ~100+ years of progress because nuclear accidents have happened.

Nuclear power has the lowest carbon output per megawatt of ANY base load power supply. Full stop.

This is a chart of deaths per TwH of power:
http://www-958.ibm.com/software/data/cognos/manyeyes/visualizations/2e5d4dcc4fb511e0ae0c000255111976/comments/2e70ae944fb511e0ae0c000255111976

Nuclear? 0.04. Coal? *161*

Wow, great, we've had Chernobyl and Fukushima as major incidents. You know how many people die every year because of coal-fired generation? Hundreds of thousands. Greenies can fuck off.

Several years ago I read that IBM set up a team on researching Metal Air Battery ... lemme search the link ... ah, found it

http://researcher.watson.ibm.com/researcher/view_project.php?id=3203

The project started around 2009

Unfortunately there is no news on the Metal Air Battery project from IBM

If you have any info regarding the latest development(s), would you kindly share with us here?

Thanks !!

A link to another startup that is researching Metal Air battery --- http://gigaom.com/2013/03/01/fluidic-shows-a-peek-of-its-metal-air-batteries-for-off-and-on-the-grid/

POWER != PowerPC

Both POWER (all-caps) and PowerPC refer both to instruction set architectures and brand names used on processors that implemented them.

The PowerPC ISA took the POWER ISA, added some stuff such as general-register-based multiply and divide instructions, and removed a few instructions (and didn't add in the ones used in the POWER2 processor).

POWER3 was a 64-bit processor that implemented the union of 64-bit PowerPC and POWER; I don't know whether any subsequent POWERn processors implemented the POWER ISA-only instructions or just the current version of the PowerPC/Power (not all-caps) ISA.

For some reason I can't find the paper linked in that article. But accessing 64k devices is basically nothing, especially if they are a bunch of virtual mod3s and mod9's like just about every mainframe config I've seen. And really its less than 64k because your terminals, tapes, etc are also in that address space. Its like saying I have 4 Gigabytes of IO space and each adapter only takes 16 bytes of it. Useless BS information. Really, I was thinking about the address space issues on the mainframe last year, and came to exactly the opposite conclusion. Its a limitation, especially since the address space is shared between LPARs. AKA if you have 20 LPARs, with independent devices its actually only 4k devices per LPAR, which could be really limiting, especially if you can't run EAVs and the max capacity per volume is 54GB.

But the real problem is that even if you attached 256 devices on multiple 8Gbit ficon channels, the CEC to IO drawer interconnect is going to be the bottleneck. Thats assuming you can start enough ficon commands to keep each channel busy.

Here are real numbers from IBM on their ficon performance at 8Gbit (still current speed).

ftp://public.dhe.ibm.com/common/ssi/ecm/en/zsw03127usen/ZSW03127USEN.PDF

Go look at emulex's site. Their numbers are over an order of magnitude better and they offer a 16Gbit board too.

Anyway, did you ask EMC why the numbers were worse on your other platform? I'm betting you can "tweak" it up to speed too. Rarely is the machine the bottleneck when talking on a SAN. Its almost always the disk subsystem. Linux out of the box on nearly every platform needs "tweakage" Many of the fiber channel drivers on linux have very shallow queues and need to be significantly deeper for IOP benchmarks.

From some IBM propaganda

The z196 is the premier high end server and the flagship of the IBM systems portfolio. It contains 96 of the world’s fastest, most powerful microprocessors running at 5.2 GHz and is capable of executing more than 50 billion instructions per second. With up to 80 configurable processors, the z196 can scale to over 52,000 MIPS (Millions of Instructions Per Second) of compute capacity in a single footprint.

Hurray for multiprocessing!

Watson was made up of 90 Power 750 servers (2880 cores, 16 TB memory, etc). Benchmarks for this sort of system are readily available The iPhone offloads its voice recognition to an Apple "mainframe".

...will no longer sell any business machines. Interesting.

Someday that might be true, but, for now, they're selling these and these business machines.

...will no longer sell any business machines. Interesting.

Someday that might be true, but, for now, they're selling these and these business machines.

They've still got System Z mainframe line, and I can't see them selling that business unit off

...and they also still have the IBM Power Systems line (Power Architecture boxes running IBM i, AIX, and Linux).

What nonsense. Every hear of IBM's Remote Supervisor Adapter?
    http://www-947.ibm.com/support/entry/portal/docdisplay?lndocid=MIGR-50116

It gives you a virtual serial console and much more. Simple serial port access is handy but the x86 vendors aren't so stupid that they make you attach physical keyboards or even KVM concentrators. I've worked in environments where dozens of xSeries servers are provisioned daily without a single keyboard or monitor anywhere in the datacenter.

According to IBM's website, $45k will get you:

4x E5-4650 @ 2.7GHz (32 total cores)
192GB RAM
0 HDD
2 1 Gbps NICs

At the $53k price point the SPARC box also includes a pair of boot drives and 4 integrated 10 Gbps NICs and an extra 64 GB of RAM. More PCIe 3.0 slots. Not to mention the 16x SMT that was already pointed out in another reply. I will note that I did erroneously double the number of cores in the SPARC system. There are only 16 cores.

It doesn't seem to be that the pricing is all that bad comparatively.

why bother? It's still sold and you can run the current version on MS Windows

http://www-01.ibm.com/software/lotus/products/smartsuite/

"AI" has always been that which AI can't do. Here are several activities that once were considered sci-fi-level AI but are no longer considered AI in a broad sense because we know how to do them more-or-less:
* Looking stuff up for us (Google);
http://www.google.com/
* Inferring questions from examples and answering questions posed in natural language (IBM's Watson);
http://en.wikipedia.org/wiki/Watson_(computer)
* Generating hypotheses and doing hands/grippers-on scientific experiments (Adam);
http://en.wikipedia.org/wiki/Robot_Scientist
* Reading text in multiple fonts reliably and quickly and cheaply;
http://en.wikipedia.org/wiki/Optical_character_recognition
* translating one human language to another on the fly;
http://domino.watson.ibm.com/comm/research.nsf/pages/r.uit.innovation.html/
http://www.gizmag.com/go/1833/
* reading and translating signs;
http://questvisual.com/us/
* Making portraits;
http://www.slate.com/articles/technology/future_tense/2012/11/tresset_robot_artist_artist_engineers_robots_to_make_art_and_save_his_own.single.html
* Playing the piano including from sheet music;
http://www.synthgear.com/2009/music-misc/synth-playing-robot/
http://gizmodo.com/5963137/watch-this-adorable-horde-of-intelligent-swarm-robots-play-piano
* Driving a car in busy traffic (Google, Stanford, CMU, others);
http://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2007_Urban_Challenge
* Winning chess games (IBM's Deep Blue and pretty much any PC now against a mid-level player);
http://en.wikipedia.org/wiki/Computer_chess
* Image recognition for quality control in factories;
http://www.general-vision.com/products/mtvs.php
* Recognizing faces;
http://en.wikipedia.org/wiki/Facial_recognition_system
* Figuring out the name of a musical composition from a few notes as well as making new compositions and dynamic accompaniments;
http://www.wikihow.com/Identify-Songs-Using-Melody
http://en.wikipedia.org/wiki/Music_and_artificial_intelligence
* The diagnostic aspect of being a doctor (Watson again);
http://www.wired.co.uk/news/archive/2013-02/11/ibm-watson-medical-doctor
* Investing in volatile financial markets;
http://en.wikipedia.org/wiki/Program_trading
* Serving as a sentry with a machine gun;
http://www.youtube.com/watch?v=v5YftEAbmMQ
* Twirling a cell phone;
http://www.hizook.com/blog/2009/08/03/high-speed-robot-hand-demonstrates-dexterity-and-skillful-manipulation
* Identifying things by smell;

No production JVM is real time.

I'd agree that the standard JVM / JDK is anything but realtime, but Sun (Oracle now of course) created a JSR specifically tackling the shortcomings w.r.t. real time:

• Sun's real time VM implementation
• JamaicaVM
• IBM's real time VM
• Atego's real time VM

It would be nice if we could have the JRE as a completely separate product from the plugin. I could happily live without the plugin (and do!) but the JRE itself is useful for other apps.

After this horrible sequence of 0-day exploits, I've finally disabled the Java plugin in ALL my browsers. There you are, instructions for removal of the Sun (or IBM) Java browser plugin on Windows, without removing the JRE. :)

That is your experience and here is mine, I work for a very large hospital in the US (4500+) providers are employed all over the state it is in. The hospital has 2 "Big Data" warehouses (Netezza High Capacity Appliances http://www-142.ibm.com/software/products/us/en/ibmnetehighcapaappl/). My current job is working in the IS Analytics department where we are in the process of consolidating/porting thousands of legacy COBOL programs to SAS http://www.sas.com/. Along with this we also do ad-hoc reporting for providers/users. The amount of data collected at the hospital from other systems and then funneled into these machines is staggering and runs 18/5 right now. If you ever had to deal with QM (Quality Measures) you know the kinds of data that is collected and stored.

Based on the fact that the primary (and really, ONLY) interface to Google glass is voice recognition, and given my experiences with voice recognition using the latest (or at lesast recent, Android 4.1) technology Google has for voice recognition, Google Glass is their Apple Newton.

The tech, it just ain't ready yet. I carefully enunciate: "Send Text to Kathy (pause) I think the problem is Becky, who wants to cancel Robert's plan"

A few beeps later...

"Sending text to Becky, The problem is Becky who wants to cancel Robert's plan".

Yeah, the example sorta sucks, but this pretty much happened to me when I decided to trust the text to speech for texting. It was almost a complete interpersonal disaster. It's good, but it's just not good enough. And given that text to speech has been "almost" good enough for at least 20 years, I'm not expecting it to improve any time soon until semantic understanding is part of the mix. (Watson: I'm looking at you....)

In response I like to send random sounding texts to family members like "Happy birth tazer ahh" just to see the response, to which I can reply: "Stupid voice to text, happy birthday Sarah!"..

I second that. Unfortunately, the originals are pretty rare by now, but these guys make a creditable replica.

Had one for years -- but only at home; you're right about the noise. (Don't forget, this is the company that brought you the Selectric and the IBM 29.)

I second that. Unfortunately, the originals are pretty rare by now, but these guys make a creditable replica.

Had one for years -- but only at home; you're right about the noise. (Don't forget, this is the company that brought you the Selectric and the IBM 29.)

Here is a link to IBM's mainframe linux: Linux on System z - Why

About the "system z" thing; IBM calls mainframes "z series", their aix machines "p series" and the intel servers "x series".
I don't know where the "z" came from for mainframe; I'm guessing "p" is for Power chips in the aix boxes, and "x" would be "x86".

Be aware that "mainframe" doesn't always mean IBM.
The most surprising mainframe I encountered last year was a Burroughs mainframe that is still in production for a large social services department (thankfully I just had to read data extracted from it, but the character conversions were sometimes surprising (no, it used Field Data, not EBCDIC nor Ascii in case you were wondering)).

Almost as surprising was a corporation that has a UNISYS mainframe at the center of their IT universe.

Yeah, mainframes are hanging in there.

No, the problem wasn't no specs. The problem was that the system was designed on paper first, without actually building it. Then the specs for the individual pieces were created, and those individual modules were built from the specs. The idea was that then the parts were completed, they would be integrated and work perfectly together. Of course, that never happens because when the pieces come together for the first time, unanticipated problems occur. This is why early integration is a good idea and is part of the philosophy of release early, release often.

Selling it as a phone that combines the security and safety of an enterprise phone with the features an fun of a "home" phone is the right approach. But they're still going to have to prove themselves on both fronts. And the clock is definitely ticking.

I don't think it's too late for them, but it's definitely the 11th hour.

iOS and Android already have this so what's going to make it stand out?

No...they don't. They really don't. Neither OS was designed with enterprise deployment in mind. The new BYOD model is exactly like herding cats. A quick glance at this chart shows just how much is left untouched.

IBM's Endpoint Manager which wants to be the BES for Apple devices is a royal PITA to use and certainly not as seamless as a BB BES solution. I am praying we ditch it for BES 10 now that it is out and allows BB, Android or iOS devices to be managed. I have been saying it since the first iOS devices started showing up...All it will take to get the entire company back on BlackBerry is for sensitive data to get leaked by one of these devices and it cost us money. Everyone will be mad because they have to give up their precious while the PR and Security teams do damage control.

There are patches for Linux kernel to similar effect in the works and devs claim nice results

According to Eric Brown, an IBM research assistant...

Eric Brown isn't a research assistant, he's a research scientist. More formally, he's a Research Staff Member, which is IBM's title for its research scientists. He hasn't been a research assistant since grad school.

So you're saying, obviously with no evidence or examples whatsoever, that a company like IBM -- with 433,000 employees, filled with engineers, and other highly skilled people -- could not have possibly come up with 6478 patents in a single year. Assuming that each patent has a unique author -- that's 1.5% of their employees.

That seems completely plausible.

Not with Java. Allocation is faster than malloc and deallocation is basically free. http://www.ibm.com/developerworks/java/library/j-jtp09275/index.html

It would be useful for someone to go back and review their previous predictions to see how often they get it right. Any volunteers? http://www.ibm.com/smarterplanet/us/en/ibm_predictions_for_future/examples/index.html

IBM has offered up its annual list of five innovations that will change our lives within five years.

If you look at their own website, their past predictions seem to have come up short! Sounds like a very good job a guess work!

Sadly, because it's from IBM, folks will listen and accordingly provide airtime for what I call a very good marketing job. Go IBM!!

And here's the IBM press release
http://researcher.ibm.com/researcher/view_project.php?id=2757
which has a sidebar that has "links to additional information" with a lot more details.

The article is remarkably lacking in technical details.

This article from two years ago is a little more detailed: http://www.eetimes.com/electronics-news/4211151/IBM-debuts-CMOS-silicon-nanophotonics
or this press release: http://www-03.ibm.com/press/us/en/pressrelease/33115.wss

FreeBSD... actually BSD is very newsworthy as it holds an integral part of computer history. Like other important pieces of history such as VMS, IBM 360, and other gone technologies BSD will be part of it that we owe a gratitude for.

Actually, neither BSD nor VMS are gone - OS X is a BSD-flavored OS at the UNIX layer, and DEC^WCompaq^WHP are still selling VMS IBM haven't made S/360's for a while, but they are making their (64-bit) descendants, which still run a descendant of OS/360, a descendant of DOS/360, a descendant of CP/CMS, and even a descendant of the Airlines Control Program. And, yes, it runs Linux, although I don't know of any BSD ports.

FreeBSD... actually BSD is very newsworthy as it holds an integral part of computer history. Like other important pieces of history such as VMS, IBM 360, and other gone technologies BSD will be part of it that we owe a gratitude for.

Actually, neither BSD nor VMS are gone - OS X is a BSD-flavored OS at the UNIX layer, and DEC^WCompaq^WHP are still selling VMS IBM haven't made S/360's for a while, but they are making their (64-bit) descendants, which still run a descendant of OS/360, a descendant of DOS/360, a descendant of CP/CMS, and even a descendant of the Airlines Control Program. And, yes, it runs Linux, although I don't know of any BSD ports.

FreeBSD... actually BSD is very newsworthy as it holds an integral part of computer history. Like other important pieces of history such as VMS, IBM 360, and other gone technologies BSD will be part of it that we owe a gratitude for.

Actually, neither BSD nor VMS are gone - OS X is a BSD-flavored OS at the UNIX layer, and DEC^WCompaq^WHP are still selling VMS IBM haven't made S/360's for a while, but they are making their (64-bit) descendants, which still run a descendant of OS/360, a descendant of DOS/360, a descendant of CP/CMS, and even a descendant of the Airlines Control Program. And, yes, it runs Linux, although I don't know of any BSD ports.

FreeBSD... actually BSD is very newsworthy as it holds an integral part of computer history. Like other important pieces of history such as VMS, IBM 360, and other gone technologies BSD will be part of it that we owe a gratitude for.

Actually, neither BSD nor VMS are gone - OS X is a BSD-flavored OS at the UNIX layer, and DEC^WCompaq^WHP are still selling VMS IBM haven't made S/360's for a while, but they are making their (64-bit) descendants, which still run a descendant of OS/360, a descendant of DOS/360, a descendant of CP/CMS, and even a descendant of the Airlines Control Program. And, yes, it runs Linux, although I don't know of any BSD ports.

FreeBSD... actually BSD is very newsworthy as it holds an integral part of computer history. Like other important pieces of history such as VMS, IBM 360, and other gone technologies BSD will be part of it that we owe a gratitude for.

Actually, neither BSD nor VMS are gone - OS X is a BSD-flavored OS at the UNIX layer, and DEC^WCompaq^WHP are still selling VMS IBM haven't made S/360's for a while, but they are making their (64-bit) descendants, which still run a descendant of OS/360, a descendant of DOS/360, a descendant of CP/CMS, and even a descendant of the Airlines Control Program. And, yes, it runs Linux, although I don't know of any BSD ports.

FreeBSD... actually BSD is very newsworthy as it holds an integral part of computer history. Like other important pieces of history such as VMS, IBM 360, and other gone technologies BSD will be part of it that we owe a gratitude for.

Actually, neither BSD nor VMS are gone - OS X is a BSD-flavored OS at the UNIX layer, and DEC^WCompaq^WHP are still selling VMS IBM haven't made S/360's for a while, but they are making their (64-bit) descendants, which still run a descendant of OS/360, a descendant of DOS/360, a descendant of CP/CMS, and even a descendant of the Airlines Control Program. And, yes, it runs Linux, although I don't know of any BSD ports.

Condition handling and Conditions. Old school, but does the job without too much clutter!

Condition handling and Conditions. Old school, but does the job without too much clutter!

https://en.wikipedia.org/wiki/Microsoft_Research
http://researcher.ibm.com/researcher/search.php?sn=1

And, he got it in one. And it's not really all that niche an idea; even TED has presented talks on the subject. And it's not just the US any more; Foxconn is getting in on the action and replacing workers with a million robots. (Incidentally, this may cause the American lead in precision manufacturing to narrow, or vanish entirely)

But what happens when everyone who is manufacturing anything automates the entire process? Our agricultural system is already hugely centralized and automated. White collar "knowledge work" is on the chopping block next. Why? IBM wants it to happen, and has the ability to make it happen.

We're also making more and more people; nobody really wants to stop screwing. So we've got more people and less work - how do we allocate enough purchasing power so they can pay for food, medicine, and housing? Two ideas spring to mind: A command economy, which is a terrible idea, and make-work, which is merely bad. Bad because it could either be terrible, or if we get really ambitious, we could make work space exploration, asteroid mining, building orbital elevators and ending our dependence on fossil fuels. So, not necessarily terrible. Except some significant portion of that work is going to be automated.

I don't know the answer, but at least I can direct your attention to the problem.

I'm going to pretend I didn't read your offensive tone and instead you wrote it in a civil manner along the lines of "Why do you think 'push' has to mean OEM agreements, adverts and boxes in stores?".

And I would concede that's a very reasonable question. Let's be honest here, the big businesses out there aren't nipping down to their local branch of PC World and picking up a hundred copies of Office.

That doesn't mean advertising is unimportant, but it does mean you need to push the product some other way. Typically, IBM sells to you via a partner - an independent business that has a number of IBM-trained staff on hand and specialises in selling & supporting IBM products. So let's find some of their partners, see if the partner's website is pushing Symphony.

IBM's Partner Locator for the UK is here: https://www-304.ibm.com/partnerworld/wps/bplocator/search.jsp

And pulling the first three partners I find, we get:

Applicable: No obvious mention of "Symphony" anywhere. Try typing "Symphony" into the search box, we find it's mentioned in passing but doesn't seem to have a product page on its own.

Anix: Anix have been taken over and are now part of the Xerox group. Nevertheless, they're still IBM Premier Partners. No mention of Symphony anywhere, punching "Symphony" into the search box gets us precisely zero results.

CSI Ltd: Computer Systems Integration. Nothing noticeable on the website, the search box shows us a link to a company called "The Symphony Group" - not quite what we're looking for.

Never mind, let's try Google. That's usually pretty reliable. Yep, Google finds Symphony. It also finds a Wikipedia article that tells us it's been discontinued.

I maintain my original stance. IBM are not pushing Lotus Symphony.

Integrated authentication can be done with Firefox if you backend application(s) support it:
https://developer.mozilla.org/en-US/docs/Integrated_Authentication

First hit off a quick Google search of "SPNEGO Firefox" - IBM WebSphere Application Server:
http://pic.dhe.ibm.com/infocenter/wasinfo/v6r1/index.jsp?topic=%2Fcom.ibm.websphere.base.doc%2Finfo%2Faes%2Fae%2Ftsec_SPNEGO_config_web.html

.
Your point on group policies is valid .. with a good reason for having it to distribute the whitelists needed for integrated authentication! Of the two features, this should be the more trivial of the two to implement. No idea why they haven't done so yet.

Funny the Robert Morris Worm got to UNIX though eh? See here -> http://en.wikipedia.org/wiki/Morris_worm

(As far back as 1988 no less...)

---

PERTINENT QUOTE/EXCERPT:

"The Morris worm or Internet worm of November 2, 1988 was one of the first computer worms distributed via the Internet... It is considered the first worm and was certainly the first to gain significant mainstream media attention... The Morris worm worked by exploiting known vulnerabilities in Unix sendmail, finger, and rsh/rexec, as well as weak passwords."

---

* NOW, what was that YOU said again about security on UNIX?

"the *nix model is a good one... It was designed for security and it was Unix." - by raymorris (2726007) on Saturday November 10, @12:47AM (#41940085)

(HOW IRONIC! I just noticed YOUR LAST NAME, is morris, upon quoting you... lol!)

It was "good", until the hacker/cracker type started "probing" it & poking around in it, exposing the weaknesses in it & apps that run on it... same will happen to Windows 8!

It's not so much the OS being exploited anymore as it is the apps or middlewares that "ride on it", ala Flash & other "browser plugins & toolbars", Java, Javascript, iframes, ActiveX (IE), etc./et al...

NOW - Personally, I do LIKE some of the new features in Windows 8 though (beneath the foolish tablet/smartphone-based "metro" interface that is):

---

Guard Pages -> http://news.softpedia.com/news/Chris-Valasek-The-Windows-8-Heap-Manager-Is-the-Most-Secure-to-Date-282466.shtml

Chunk Randomization -> (same link as above)

And, my "favorite" (not really security-related as much as the other 2 above): Self-Terminating Services - which means services (like *NIX daemons) finding themselves inactive, "auto-magically" shut themselves DOWN... this saves "tuners/tweakers" a bit of work, since we've been doing services tuning since, forever (myself back into the Windows NT 3.51 days onward to present into Windows 7 64-bit, currently)...

---

(HOWEVER - like I said above, & you guys all pretty much KNOW this too: The apps will be "targetted-for-termination" more than ever! History above even shows us that on UNIX... So, is that a "bad thing"?? Yes, sure, initially... but, not when the bugs & security holes "shake out", & they WILL, eventually!)

APK

P.S.=> Also, again - lastly: Hate to "burst anyone's bubble", but the original UNIXES were NOT designed with security in mind... No more than the internet was!

In fact, & this is some "interesting trivia"?

UNIX was designed initially for TEXT PROCESSING WORK -> http://www.ibm.com/developerworks/aix/library/au-textprocess.html

PERTINENT QUOTE/EXCERPT:

"The origin of UNIX® lies in simple text processing"

"Believe-it-or-not"...

... apk

The IBM research paper is available at http://www.nature.com/nnano/journal/vaop/ncurrent/full/nnano.2012.189.html The paper is protected by a paywall.

The IBM press release is available at http://www-03.ibm.com/press/us/en/pressrelease/39250.wss

I recommend reading the comments on the New York Times article. My favorite comment so far is:

MC - NYC
The Singularity edges closer...

Haven't found the actual paper, yet, but I think it's "High-density integration of carbon nanotubes via chemical self-assembly" as mentioned here: http://researcher.watson.ibm.com/researcher/view_pubs.php?person=us-aaronf&t=1

For the SSD's that IBM sells, it gives them an "Endurance" rating for a certain number of "Total Bytes Written." According to http://www.redbooks.ibm.com/abstracts/tips0879.html The implication for this discussion thread is that if you build a RAID 1, 5, 6, or 10 array with new SSD's you should expect them to all reach end of life at about the same time.

The drives described in this paper from March, 2012 are rated for Endurance: "36 TB of total bytes written (TBW) at 90% full disk based on predefined usage pattern for 64 GB SSDs and 72 TB of TBW for higher capacity drives."

"Enterprise Value SSDs and Enterprise SSDs have similar read and write IOPS performance, but the key difference between them is their endurance (or life time) (that is, how long they can perform write operations because SSDs have a finite number of program/erase (P/E) cycles). Enterprise Value SSDs have a better cost/IOPS ratio but lower endurance compared to Enterprise SSDs. SSD write endurance is typically measured by the number of program/erase (P/E) cycles, that the drive incurs over its lifetime, listed as TBW in the device specification.

"The TBW value assigned to a solid-state device is the total bytes of written data (based on the number of P/E cycles) that a drive can be guaranteed to complete (% of remaining P/E cycles = % of remaining TBW). Reaching this limit does not cause the drive to immediately fail. It simply denotes the maximum number of writes that can be guaranteed. A solid-state device will not fail upon reaching the specified TBW. At some point based on manufacturing variance margin, after surpassing the TBW value, the drive will reach the end-of-life point, at which the drive will go into a read-only mode. Because of such behavior by Enterprise Value solid-state drives, careful planning must be done to use them only in read-intensive environments to ensure that the TBW of the drive will not be exceeded prior to the required life expectancy.

"The endurance of Enterprise Value drives is specified based on the following access pattern: 50% random data and 50% sequential data with block size mixes of 5% of the data as 4 KB block size, 5% of the data as 8 KB block size, 10% of the data as 16 KB block size, 35% of the data as 64 KB block size, and 35% of the data as 128 KB block size. The Enterprise Value drives described here are capable of 36 TB (64 GB SSD) or 72 TB (128 GB, 256 GB and 512 GB SSDs) of lifetime writes, with the workload stated above as the worse case. For the device to last in five years inside of the 72 TB of TBW, the drive write workload must be limited to no more than 40 GB of writes per day. For the device to last in three years, the drive write workload must be limited to no more than 65 GB of writes per day."

I don't like his comment on DCBZ

No, I want out-of-order and "high-level" instructions that actually work across different implementations of the same ISA, and across different classes of hardware (iow, span the whole "low-power embedded" to "high-end server" CPU range). So for example, I think having a "memcpy" or "memset" instruction is a great idea, if it allows you to have something that works optimally for different memory subsystems and microarchitectures.
As an example of what not to do, is to expose direct cacheline access with some idiotic "DCBZ" instruction that clears them - because that will then make the software have to care about the size of the cacheline etc. Same goes for things like "nontemporal accesses" that bypass the L1 cache - how do you know when to use those in software when different CPU's have different cache subsystems? Software just shouldn't care. Software wants to clear memory, not aligned cachelines, and software does not want to have to worry about how to do that most efficiently on some particular new machine with a particular cache size and memory subsystem.

DCBZ is described here

http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?topic=/com.ibm.aix.aixassem/doc/alangref/dcbz.htm

If you look at a memset on a traditional CPU you're actually wasting time writing zeros individually to each location on a cache line. That works, but what you really mean at the start of a line is not

"Pull the cache line in from main memory if necessary, perhaps evicting a line containing something else. I'm writing a zero to the first location, second location ... last line". This means you read the line from memory and then fill it full of zeros. It will then sit in the cache dirty until it gets written back. Reading it into the cache means something else got kicked out.

What you really mean is

"I'd like to clobber the whole cache line with zeros". I.e. whether or not you have it, just fill it with zeros and mark it dirty. Or, better, write zeros to it bypassing the cache completely - that way you don't pollute the cache."
dcbz is a way to say that.

They explain it here

http://www.ibm.com/developerworks/power/library/pa-memory/index.html

It just means you need to write some PPC specific code for memset. Which IBM give you. If you do it it will run faster. If you do it the old way it will still work.
I.e. it's an architecture specific optimisation. I think he's saying that it would better to make the old way as fast as dcbz. I'm not sure that's possible - the CPU can't turn a bunch of stores into the equivalent of dcbzs. It's one of those situations where the old way is slow because you're not giving the CPU enough information about what you really mean.

If you've got an embedded system that's slow it's worth profiling it. If there architectural things like this that aren't being used in the hot spot functions its worth implementing them.

I don't like his comment on DCBZ

No, I want out-of-order and "high-level" instructions that actually work across different implementations of the same ISA, and across different classes of hardware (iow, span the whole "low-power embedded" to "high-end server" CPU range). So for example, I think having a "memcpy" or "memset" instruction is a great idea, if it allows you to have something that works optimally for different memory subsystems and microarchitectures.
As an example of what not to do, is to expose direct cacheline access with some idiotic "DCBZ" instruction that clears them - because that will then make the software have to care about the size of the cacheline etc. Same goes for things like "nontemporal accesses" that bypass the L1 cache - how do you know when to use those in software when different CPU's have different cache subsystems? Software just shouldn't care. Software wants to clear memory, not aligned cachelines, and software does not want to have to worry about how to do that most efficiently on some particular new machine with a particular cache size and memory subsystem.

DCBZ is described here

http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?topic=/com.ibm.aix.aixassem/doc/alangref/dcbz.htm

If you look at a memset on a traditional CPU you're actually wasting time writing zeros individually to each location on a cache line. That works, but what you really mean at the start of a line is not

"Pull the cache line in from main memory if necessary, perhaps evicting a line containing something else. I'm writing a zero to the first location, second location ... last line". This means you read the line from memory and then fill it full of zeros. It will then sit in the cache dirty until it gets written back. Reading it into the cache means something else got kicked out.

What you really mean is

"I'd like to clobber the whole cache line with zeros". I.e. whether or not you have it, just fill it with zeros and mark it dirty. Or, better, write zeros to it bypassing the cache completely - that way you don't pollute the cache."
dcbz is a way to say that.

They explain it here

http://www.ibm.com/developerworks/power/library/pa-memory/index.html

It just means you need to write some PPC specific code for memset. Which IBM give you. If you do it it will run faster. If you do it the old way it will still work.
I.e. it's an architecture specific optimisation. I think he's saying that it would better to make the old way as fast as dcbz. I'm not sure that's possible - the CPU can't turn a bunch of stores into the equivalent of dcbzs. It's one of those situations where the old way is slow because you're not giving the CPU enough information about what you really mean.

If you've got an embedded system that's slow it's worth profiling it. If there architectural things like this that aren't being used in the hot spot functions its worth implementing them.

We had an IBM 1130 computer in my senior year. Previous years, I'm told, we had teletypes and access to the Xerox mainframe at West Chester State, but I didn't get interested until 12th grade, so I only knew the 1130. I believe it was a model 3B with the internal disc cartridge drive, 1403 model 7 printer, 1442 card reader/punch.

The Computer Science course taught us basic programming in FORTRAN IV. There was a "computer math" course by a different teacher; I don't know anything about the course, but the teacher was the guy running the whole show, including the computer club, so I knew him anyway. Most of what we did in the computer club amounted to fooling around a lot in APL and learning more detail about how the computer worked.

Bonus: The computer science teacher was young, cute, female, and single. :)

And please. I know I set myself up for a pun. Note that I spelled "basic" in all lower case. I obviously am not speaking of the programming language. Thank you.

http://www-03.ibm.com/ibm/history/exhibits/1130/1130_intro.html
http://en.wikipedia.org/wiki/IBM_1130

Nuclear has the lowest deaths per terawatt-hours produced than any other significant energy production method, by a couple orders of magnitude: http://www-958.ibm.com/software/data/cognos/manyeyes/visualizations/deaths-per-twh-by-energy-sources (and yes, the figure includes all nuclear power accidents). Note that even hydro is far, far worse. Here's another, more striking, visualization: http://transitionvoice.com/wp-content/uploads/2011/03/death-rate-per-watts-Seth-Godin.jpg