Posts by Chris Mellor 1

HP PCIe-flashed ProLIant rocket coming?

I'm looking into flash usage by HP gen 8 ProLIant servers and came across this claim in an HP blog (http://h30507.www3.hp.com/t5/Reality-Check-Server-Insights/Gen8-Dynamic-Workload-Acceleration-Solid-State-Optimization-amp/ba-p/109191)

"As a result of this technology [Dynamic Workload Acceleration] HP ProLiant recently achieved the #1 TPC-C benchmark[3] in the industry."

The [3] reference is to: "#1 TCP-C Benchmark: www.tpc.org/results/individual_results/HP/HP_ProLiant_DL385G7_TPCC_111114_01_es.pdf"

Following up the hot link gets this:

HTTP Error 404.0 - Not Found"

Looking in the TPC_C results I found an HP ProIiant DL385 G7 listed at number 10 in the top non-clustered results list with 1,207,9782 tpmC. An Integrity Superdome is listed higher at number 3 with 4,092,799 tpmC. No tpm-C topping result here.

So what is HP on about in this blog?

The ProLiant DL385 G7 result is dated 27 August 2010 (http://c970058.r58.cf2.rackcdn.com/individual_results/HP/HP_ProLiant_DL580G7_2.26GHz_es_100830_Energy_v2.pdf) and it used 38.4TB of 10K SAS disk storage and 30.72TB of 120GB SSD storage.

What I think we 'might' have here is an all-flash ProLiant DL385 G8 (not G7) result, using PCIe flash and not SSDs, which beats the current non-clustered tpmC result record held by an IBM Power 595 Server model 9119-FHA with 6,085,166 tpmC. It hasn't been validated yet hence the hot link reference has been withdrawn or is in error and has been cancelled.

Intriguing stuff.

Object storage and tape

HDS is the only object storage supplier I can find that can move its objects to tape for off-line vaulting. Is it realistic for object storage suppliers to say they don't need tape?

EMC Thunder and Pandora's box

I reckon everyone that has a PCIe flash product and a networked flash array is going to be looking at EMC's Thunder and thinking they could do it too, and maybe partner with back-end array vendors. We're looking in Pandora's box here and EMC has opened the lid.

I think HP may have a strategy here; the P4900 all-flash array, the work with Violin, Donatelli's knowledge of the EMC mindset - I can see some HP Thunder-box coming; ditto a Dell one. Don't know about IBM but the SVC guys are clever and if they get a high-speed server interconnect in their hands they could build something.

NetApp seems content to lumber along a distance behind - curious.

Fusion-io though; DAvid Flynn is clever and Flynn will not want to have EMC VFCache and Thunder steal his market. ACM will hold things for a while but he must get server sharability for his products. With that and a replication capability between them and deduplication - he must surely be getting dedupe! - he could survive against a VFCache-Thunder-VMAX/VNX combination until he adds in cloud storage for secondary data or has back-end array partnerships to accomplish that and so move out of the individual server acceleration niche he is in, and in danger of getting boxed in, at the moment.

Violin also has plans. Don't know what they are.

Xsigo is frothing at its marketing mouth.

This is the hottest section of the market and the main players are EMC, Dell, Fusion-io, HP and Violin, with lots of others crowding the sidelines and wanting to move centre stage.

What will happen and who will emerge as major players?

Chris.

Could you give me an example. For me I prefer to store data/information in the form I het it it. That's fast and convenient and the stuff is easier to find. What's not to like?

Chris.

Big Data and Hadoop

Should you bring the business intelligence (BI) analytics engine to Hadoop or the Hadoop data to the BI engine?<p><p>

EMC has Hadoop data stored on Isilon scale-out NAS and them brings that data to GreenPlum. Terradata, Netezza and Vertica have connectors to extract data from a Hadoop store and bring it to their engines, says RainStor.<p>

RainStor announced today its database stores Hadoop data natively and can run its analytics on that data directly, either Hadoop queries or SQL queries. It says it has brought the engine to the Hadoop data and there's no need for massive Hadoop data copying exercise to bring data to the engine.

Is that view right or wrong?

Chris.

Browser gores into loop

In "Welcome To The Latest Forum Features" I clicked withdraw on a topic - "Line Endings" - which I had created in that forum, was presented with a withdraw screen and, when I clicked "Back to the forum" the browser flashed and redrew the screen again and again. When I closed it down and restarted it and re-entered this form the topic was still there but couldn't be replied to.

Browser issues

Hi. I can get in and operate here with Safari but not, it appears with Opera, with which I cannot log-in. Are there browser limitations? Chris.

Navigation

I'd love a direct way to get back from replying to a thread, such as Latest Forum Features, to the main User Forums' page (http://forums.theregister.co.uk/section/forums/user_forums/) by clicking a button rather than using the browser's back button. Is that feasible?

LIne endings

Does this<br>work?

The nails smacked formly on their heads

It's the total disk write time and file/object seek time that seem to me too to be crucial attributes of holographic storage. Some kind of file system for the latter and a fast-as-possible drive for the latter. But developing the drive - that's what tormented InPhase to distraction and then imploding destruction.

Geoff Barrall says ....

An extract of a mail sent to me by Geoff Barrall:-

"I wanted to make sure that you heard from me that I enjoyed working for Eric at Overland and learned a lot during my time there. However both Eric and I had discussed on the way in and out that a turnaround was not necessarily a long term home for me. My first stock grant at the company was structured (unusually) as a one year vest based on those discussions. Eric and I discussed every aspect of my departure and the folks who would be coming to Connected Data and everything around the exit was handled professionally. I hope we part as friends.

The article makes it seem like I'm constantly on the move. This seems a little unfair. Other than Overland I've only had two full time jobs in the last 13 years. I was at BlueArc for six years between 2008 and 2004 (the article lists 2001 as my exit) and the Data Robotics (now Drobo) for five years between that start of 2005 and the end of 2009. I've remained connected to BlueArc for the whole 14 years and still proudly serve on their advisory board. "

------------------------------

Chris,

A couple of questions

Hi,

1. What happens if I track a forum? Do I get emails alerting me to posts in that forum?

2. How do I create threads in this (moderators') forum? There doesn't seem to be a way to do that.

Cheers ... C

Storage Forum

This is the Register Storage Forum for the discussion of any and all computing storage issues.

Blaise Mouttet comments

Blaise Mouttet has sent in the following email, which is reproduced here:

Chris,

HAPPY NEW YEAR!

Regarding your recent article on the "memristor brouhaha":

Stan Williams is dead wrong that phase change memory (and other unipolar memory types) can be accurately considered a memristor (even given the recently revised 2011 definition which is arguably fraudulent in and of itself). Neither Stan Williams nor Leon Chua have any expertise in phase change memory (or various of the other memory types Williams claims to be "memristors"). No one as far as I know has provided a correct proof or reasonable model showing that phase change memory can be accurately modeled as a memristor or as a 1st order memristive system. It is irresponsible for Stan Williams to be spreading this type of misinformation for the consumption of the popular press unless such a proof is first produced and well vetted in the scientific literature. I believe that Williams is doing this for business reasons that have nothing to do with correct science. This is very dangerous if it is allowed to continue unchecked.

I am cc'ing David Wright who has a great deal of expertise in modeling phase change memory materials and who has published at least one paper I am aware of peripherally related to the memristor concept.

I am also cc'ing Dalibor Biolek who was one of the first researchers to create SPICE models for memristive systems and who recently published an important paper distinguishing between the pinched hysteresis loops of bipolar (TYPE 1, Fig.1a) and unipolar (TYPE 2, Fig. 1b) memory.

In the 1976 paper by Chua and Kang an example of an impossible Lissajous curve for memristive one-ports was given based on property 4 of memristive systems (see Fig. 5b). The characteristic of this curve is similar to that found due to the voltage snap back in phase change memory which David Wright or any other researcher in phase change memory can confirm. Thus either Chua and Williams are ignoring this property or (more likely) have too little knowledge of phase change memory to realize that a claim that phase change memory is a memristor is easily provable to be scientifically false.

The more recent article by Dalibor gets more directly to the point by noting in the conclusion that memristance-charge or memductance-flux relationships cannot generate pinched hysteresis loops of type-II (which are typically associated with unipolar switching found in phase change memory).

Willliams is making false claims about the memristor either out of his own ignorance or deliberate attempt at fraud to advance HP's business agenda (at the cost of legitimate companies developing ReRAM such as Sharp, Samsung, and Unity Semiconductor). This should not be allowed and serves only to create confusion in the public.

Given the above information Williams should at minimum withdraw his allegation that phase change memory is a memristor and apologize.

I am currently preparing an article which I will be posting on wordpress in a few days going into more detail about why various of the arguments made by Williams in his rebuttal are completely bogus.

Prices

The 64GB Synapse Cache costs $149 and the 128GB one $249; lower than we thought.

Chris.

Prices all wrong

Chad Sakacc sent me this comment and I'm posting it here to get it up asap:-

>>>>>>>>>>>>>>

Chris – your ensuing article is totally off base - http://www.theregister.co.uk/2011/09/20/nexenta_vmworld_2011/- and (perhaps I’m sensitive here) paints the wrong picture.

First of all – where did you get prices for the configs? Reality is that Nexenta, NetApp, and EMC solution prices are, in the end, about the same.

Also, was there any support for the “Nexenta supported 50% of the load” quote (where did THAT come from)? Or was it just made up?

And let me be abundantly clear – EMC came out with flying colors, had no issues, and while I won’t throw the other guys under the bus (because there WERE problems), your article is off base.

>>>>>>>>>>>>>>

Great comment. The quoted prices came from Nexenta sources. The idea that "Nexenta, NetApp, and EMC solution prices are, in the end, about the same," is surprising and I'm asking Chad for numbers to verify that.

The 50 per cent load number came from Nexenta; it's a CEO quote in the story. Where he got it from is his affair. Either he is telling the truth or he is not.

Chris.

It's the 2600-HD Engenio base, not the E5400

Well, well. I've been told this (anonymously):-

Actually, Chris, the DCS3700 is NOT the E5400. Rather its the 2600-HD annouced last November. see-- http://www.wwpi.com/index.php?option=com_content&amp;view=article&amp;id=9377:lsi-debuts-engenio-2600-hd-hpc-storage-system&amp;catid=221:other-storage-media-ssd-hdd-blades&amp;Itemid=2701223

The E5400 is a much higher performance controller than the 2600-HD, comparable to HDS's AMS 2500 (6Gbit/s SAS backplane, higher max IOPS but lower max throughput than a DS5300/E7900).

While Oracle and SGI agreed to take the E5400 some time back, IBM never did. This "roadmap divergence" may be part of why Phil Bullinger, the head of Engenio under LSI, is no longer at Engenio or LSI (but now has been hired to head up Oracle's storage division).

In IBM's defence, however; the E5400 does not fit all that well in their portfolio. As a replacement for the DS5020, its performance numbers are just a little to close (and sometimes better) to the DS5300.

--------------

Isn't that interesting?

Chris.

Ahead of his time with the 'net

Sent to me:-

I wanted to comment a little on your piece re: Tom West. It seems you missed an important part of the timeline in coming to the conclusion that he missed the network desktop.

As described in this metafilter comment I made:

http://metatalk.metafilter.com/13606/Jessamyn-in-Valley-News#384503

Tom saw the promise of the internet quite a bit more than most, and built a team ('thiinline') to try to capitalize on that. We built the highest-density 1U purpose-built web server, the first intended-for-the-home 'home server' with 802.11b wireless, and various odds and ends, including the first web tablet and an internet-connected teddy bear.

In the end it turned out that (a) we were actually ahead of our time, and (b) despite what you said in the article, Tom's power over funding was not particularly magical, and when faced with technophobes like Ron Skates, Tom's magic wasn't enough sometimes; arguably DG was already in decline, but we didn't get the staff or the funding to execute on the idea and had to constantly fight the bureaucracy.

Hope that helps.

-------------

Chris.

Removing radioactive element

The article originally said the radio-active element of the depleted uranium molecule had been removed, and this was based on the U of Nottingham release about Liddle's SMM, which said this: "Although it may have somewhat negative PR it seems depleted Uranium — a by-product from uranium enrichment and of no use in nuclear applications because the radioactive component has been removed ..."

I didn't check this - FAIL!

It's been pointed out that this is hogwash. A commentard said: "U-238 still decays with a half-life of 4 x 10⁹ y, which is around 10 times longer than U-235. Still, not a lot."

But still, definitely still radio-active. So the article has been corrected and I'm grateful for the education.

Chris.

OCZ spokesperson says....

Sent to me by an OCZ spokesperson:-

(1) "OCZ joins PCIe flash fray" and "OCZ has joined in the PCIe flash market"

The language used in the article insinuates that we have just joined the PCIe market when in fact we've been in it for awhile now; our first PCIe SSD was in 2009:

http://www.ocztechnology.com/aboutocz/press/2009/337

We actually have a successful line of both enterprise and client PCIe SSDs which no other company can say:

http://www.ocztechnology.com/products/solid_state_drives/pci-e_solid_state_drives

(2) "Coupling software RAID with incompressible data reveals some even sillier numbers. The random write numbers are presented in MB/sec terms and not IOPS, which is the standard measure for virtually every other supplier and OCZ's previous SSDs. That makes you suspicious, and then the numbers themselves are seriously weird: 64MB/sec for the 300GB card, 65MB/sec for the 600GB card and a near-derisory 19MB/sec for the 1.2TB card."

This is because we have recently added benchmarking scores that focus on incompressible data and we and the rest of the industry feel AS-SSD does this well; Iometer does not offer the ability to have that much control over the type of data. The benchmark reports in MB/s which is why we've added MB/s to the compressible data specs as well (for easy comparison).

As with all enterprise-targeted products, we recommend that clients work with us to determine which SSD solution is right for their application/environment. In most cases, nobody comes close to our price/performance ratio.

- - - - - - -

Thank you OCZ,

Chris.

Not so brilliant - 2 hiccups

Sent to me:-

With a minor hiccup, two actually: (1) DIMM slot does not properly signal power-loss (no wonder - DIMMs don't care) and (2) memory hub has not been designed with microsecond-level time-out in mind. Otherwise it's just fine, provided one can get away with these two (which I doubt :)).

Thanks,

Chris.

Iron Mountain DID tell customers

Sent to me by Iron Mountain spokesperson:-

Iron Mountain did recently notify customers of our Virtual File Store and Archive Service Platform that we are retiring these two commodity cloud-storage solutions. This decision only affects those using Virtual File Store, a low-cost cloud storage option for inactive files, and technology partners who use the Archive Service Platform as a general purpose cloud for storing their customers' data. As the Gartner report notes, public cloud service offerings like these have seen modest levels of adoption.

- - - - - -

Thank you IM,

Chris.

Dave Mosley has NOT disappeared

Sent by Seagate and an extract from an SEC filing:-

William David Mosley, 44, was appointed Executive Vice President, Operations. He will be responsible for the Company’s global heads, drive, and media manufacturing operations. Before this appointment, Mr. Mosley served as Executive Vice President, Sales and Marketing since September 2009. He previously held positions as Executive Vice President, Sales, Marketing and Product Line Management from February 2009 to September 2009, Senior Vice President, Global Disk Storage Operations from 2007 to 2009 and Vice President, Research and Development, Engineering from 2002 to 2007. Prior to 2002, Mr. Mosley held other progressively senior management positions in the Research and Development and Operations and Manufacturing organizations of the Company.

--------------------------------------

Thank you Seagate,

Chris.

Texas Memory Systems responds

Sent to me by TMS:-

There are a few things that are not presented clearly in the write performance comparison from the slide deck that you uncovered and I thought it might be helpful to shed some light on this. Flash chips need to have entire blocks (256 KB) written and erased at the same time. Since applications write in smaller chunks than this there is some over-provisioned capacity and between the host and the flash chips is something called a Flash Translation Layer. I’ll try to summarise the details of dealing with the writes here. For more information, SNIA’s Solid State Storage Initiative’s performance test specification goes into quite a bit of depth: http://www.snia.org/forums/sssi/knowledge/education/SSS_PTS_Whitepaper_Nov2010.pdf

The Flash Translation Layer keeps an index of where a block is physically written and the logical address that is presented to the host. After a continuous small block random write workload, eventually there are not any pre-erased blocks available to write to. The flash controller has to perform “moves” of data from a few blocks that have some valid data and some stale data. These moves are performed to get a few completely full blocks and a new empty block to write to. A move operation ties up the chip and is just a little less work than a write. The amount of over-provisioned capacity determines how many background moves you will have to perform for each new write that comes in under the worst possible scenario. On the RamSan-20 datasheet http://www.ramsan.com/files/download/589 we list this number for our random writes: 50,000 IOPS. Outside of this worse case, we can achieve 160,000 write IOPS.

In many ways the sustained number is not terribly important to real-world applications as anytime that you are not doing 100% random writes a good flash controller will perform background move operations to defragment the valid and stale data. To get to the worst case, you have to randomly write across all of the capacity and more without ever stopping to read. While some logging applications have this type of constant write workload they are never random.

So that brings us to the interesting question: can a flash controller be designed that doesn’t have this drop off in performance? Absolutely – you merely need to have more chips on the backend than the frontend can drive. That means that for most workloads you have idle chips. Notice that in the comparison from the presentation, a 10 TB flash chassis (8 TB after over provisioning) is being compared to the performance of single PCIe cards with 450 GBs or less. If you multiply the performance of any one of the cards presented below by ratio of their capacity to 8 TBs that they are being compared to, every single one beats the 1.4 GB/s that is provided by Violin.

Texas Memory Systems had to ensure we had a fast enough front-end on our high-capacity RamSan-630 system to drive all of the chips that are available in 10 TBs of flash (after over provisioning). That is why we support ten QDR InfiniBand (40 Gbps each) or ten 8 gbps FC ports and can supply over 10 GB/s of bandwidth from a single 3u system. Flash is still quite a bit more expensive than disks, so adding capacity without adding the front end performance doesn’t serve customers well.

------------------------

Thanks TMS!

ST-412 history notes

Sent to me by Steven L. Kaczeus:-

Originally this product was ST-512 as a double density product of the ST-506. When I changed the read/write heads to thin film then we changed the product number to ST-412. My following project was the ST-225 product which was a half high and again double density device. The ST-225 was the highest volume selling product for Seagate Technology. This product also made Seagate a very successful company.

-----------------

Thanks Steven,

Chris.

Not ignoring storage flash

A Marvell spokesperson ent this to me:-

[We] aren’t ignoring the benefits of having flash cache in servers, such as NetApp's arrays, and of having flash drives replacing hard drives in storage arrays. DragonFly in fact complements it. DragonFly offers an L1/L2 host I/O cache but back-end SAN/NAS can still benefit from SSDs as an L3 cache. SSDs in storage arrays can be used to speed I/O relative to disks within the array. DragonFly offers the host-level cache to reduce traffic going to/from the array – so they complement each other.

----------

Thanks Marvell,

Chris.

Sent to me by....

... an industry insider:-

A pitch-perfect article on ILM. "Incomplete Literature Musings" might fit the acronym better. It was always a big logical and physical jump for storage vendors to leap from physical disks to information management by way of storing data.

No one vendor, however richly endowed with marketing dollars and R+D budgets can expect to corner this market. A long line have tried, with very mixed results. Who will be the firstbrave vendor to offer cloud storage with ILM as a sweetener?

Despite all the hype, the big storage issues today are much as they were 5, 10, 20 years ago... performance, availability, connectivity and security.

------------------

Chris.

SpectraLogic corrects a few points

Sent to me by SpectraLogic's Molly Rector, VP of product management:-

Our new Data Integrity Verification feature is a first-to-market offering of its kind for several reasons:

The recent announcement that was issued by Spectra Logic on Monday, March 21 included two “new”/”market first” items:

1. Spectra Logic is the first tape library vendor to offer this level of data integrity validation to its customers across its entire product line (SME-large enterprise).

2. Spectra Logic has the only tape library data integrity verification offering that is free of charge (no additional license fee) to all customers—both existing installation base and our future customers.

A few other points for clarification follow. Spectra Logic has been developing media integrity and data integrity features for many years in our libraries. The recent announcement continues to build on this important feature set by now offering it to all customers at no charge regardless of the size of tape library that they have. This feature is no longer just available to Enterprise users but to all of our tape storage users. We think this is critical as all customer data, regardless of how large or small their data set, is critical to their organization.

As a short summary of our announcement dates and history on this topic, in April 2008, Spectra Logic announced Media Lifecycle Management, which was our first piece of the data integrity verification story. MLM checks more than 30 health points on individual cartridges and proactively notifies the administrator when it is time to retire tape media and replace them with new tapes, mitigating data corruption and loss.

Also, as a clarification to some of the points made in the recent write up that were not accurate regarding our offering:

· Data Integrity Verification (PreScan, QuickScan and PostScan) is fully functional with any media from any LTO supplier. It is not limited to just Spectra Logic’s Certified Media. Additionally, Spectra allows its customers to use any resource they deem appropriate for data verification. They can be dedicated to just this purpose or shared with production resource. We do not require loading tapes into a production library and work the scanning around normal operations. As you know, Spectra’s entire product line is based on a modular, customizable design that can be easily adjusted based on individual customer needs and requirements.

· With Spectra’s Data Integrity Verification, dedicated resources can absolutely be used. In fact, Spectra recommends that production resources are used to achieve optimum speed of QuickScan. (With QuickScan, a very fast scan is completed immediately after a job completes while the tape is still mounted in the drive. It would be poor resource utilization to unmount it, move it, and put it in a different drive. We let the customer use any resource they deem appropriate.)

· Also, two tapes/copies of data are not required to conduct the Data Integrity Verification process. This in many cases is considered a best practice in data protection and archive but is not a requirement of the Spectra Logic BlueScale Data Integrity Verification feature specifically.

----------------------

Thanks,

Chris.

Savvio areal density numbers changed

Due to a mix-up the Savvio 10K.5 and 15K.3 areal density numbers were wrongly specified. The correct numbers are 506Gbit/sq in for the Savvio 10K.5 and and 448Gbit/sq in for the Savvio 15K.3.

Chris.

A Xiotech view

Sent to me from Jim MacDonald, chief strategy officer of Xiotech:-

“The success of Fusion IO shows that there is a massive demand for application performance that cannot be fulfilled by traditional storage systems. It is interesting to note though that Fusion IO is not 2x or 5x but often 10x that of hard drive-based arrays, and there is a massive market in the “fast enough and large enough” space which remains largely untapped.”

“Fusion IO is a great technology for traditional data sets, where fast access times are critical. However given the growth of big data there is also a requirement for something that sits between the traditional arrays and the cache-like Fusion IO. This is also a huge growth space, and one in which Xiotech is proud to be leading the charge.”

“It’s important to note that while their name connotes a “Fusion,” Fusion-IO is pure cache. Its singular focus on performance is absolutely one that people need right now, but as Fusion IO attempts to reach new markets it will find that traditional storage features such as replication, clustering and high availability become more essential and these are not something that Fusion IO can add easily with their existing architecture. They also will face an increasing challenge in their core market by ever-decreasing DRAM prices.”

People don't want SAN overhead

Sent to me:-

Your comments are right on. People want the scale, performance and availability of a SAN but they don’t want the overhead of managing it.

As we deliver this to the market we have discovered the "SAN friction" issue is immense and by relieving customers of it we are enabling SAN's to be deployed more broadly.

Thanks for the thoughtful article.

Bob Fernander CEO Pivot3 Inc.

Fidelma Russo's title

Here is EMC's statement:

"EMC always looks to recruit top executive talent. Fidelma joined EMC earlier this year as Senior Vice President reporting to Brian Gallagher, President of the Symmetrix and Virtualization Product Group. This organization is focused on EMC Symmetrix and VPLEX product families."

Chris.

Understated revenue

I have been told that the 2008 numbers I guessed at were about 1/2 of what Data Robotics achieved in that calendar year. That's because Jillian Mansolf did an amazing job in the first full year of shipments. Also Data Robotics shipped for two and a bit quarters in 2007, from June onwards, so there was about half a year's revenue in that year too,

Adjusting my numbers (assuming a $500 ASP) along these lines and guesstimating we get:-

2007 - $5m

2008 - $10m

2009 - $25m

2010 - $47.5m

TOTAL - $87.5m

Halve the numbers for a $250 ASP.

That's probably a more realistic picture.

Chris.

ZFS prescribed

Sent to me :-

------------------

Hi Chris: In reference to this article. "EMC races to catch up with NetApp"

All EMC has to do is integrate ZFS into its kernal to make a run at NetApp's customers. BSD has ZFS integrated now and if a company like EMC put its boffins to work an creating a system based on COTS equipment with integrated CIFS and NFS they would have a winner.

--------------------

Seems a neat idea.

Chris.

UNIX users know this to an extent already

Sent to me by mail:-

----------------------------

You're quite right about deduplication (of course), but this is an issue that Unix folk have faced for years - at least those who make use of hard links. Okay, it's not quite the same as (for most filesystems) the space is automatically reclaimed once the final hard link is deleted (specifically, when the reference count drops to zero) rather than after a scan*, but you take the point.

There's no such thing as a free lunch - you "gain" space in deduplication only in as much as you are reusing bit-patterns; you are not owed that space, and you only get the benefit so long as the duplication remains true. Deletion of one copy (as opposed to two) breaks that.

* Okay, refcounts can get confused, in which case nothing short of an fsck or equivalent is going to return you the space - but that's a bug not a feature.

Mark

P.S. The pedant in me must point out that if you *do* delete enough, of course it will free up space. ;-)

-----------------------

Thanks Mark,

Chris.

Bit level history of early PDPs

Sent to me by Geoffrey G. Rochat:-

--------------The original DEC computer, the PDP-1, was an 18-bit machine, not a 12-bit machine. This error has shown up in a lot of Ken Olsen obituaries over the last couple of days because nobody checks the facts. The PDP-1 begat the PDP-4, then the PDP-7, PDP-9 and finally the PDP-15, and they were all 18-bitters. A few years after the PDP-1 came out DEC launched the PDP-5, and it was a 12-bit machine. The PDP-5 begat the PDP-8, and the PDP-8 series, which was wildly popular in its day, launched what we know of as the minicomputer revolution, before being supplanted by the 16-bit PDP-11. But the PDP-5 and PDP-8 were preceeded by several 18-bit DEC machines, and they trace their heritage back to the TX-0 computer that Ken Olsen worked on at Lincoln Labs as a graduate student. Anybody who wants to verify the chronology is invited to look at Al Kossow's Bitsavers site, www.bitsavers.org, where may be found scans of a vast array of retrocomputing documents.

-------------------

Thanks Geoffrey,

Chris.

EMC first

This was sent to me by EMC's Rick Lacroix, Director of Public Relations ∙ Storage Products and Software:-

------------------------

A point of clarification on today's VAAI integration story. HDS isn't first.

VAAI is GA on VMAX - it was announced in January and began shipping last December, it was a major part of our announcement in London.

We also announced it on our midrange systems at EMC World last May. GA in Q3 of '10

http://www.emc.com/about/news/press/2010/20100511-02.htm

There's lots of VAAI integration work happening in the industry, but the subhead on today's story is inaccurate as it relates to EMC.

------------------

Chris.

New CEO is going to stay

This was sent to me:-

Cirtas VP of Marketing and Product Management Josh Goldstein: "Gary is at Cirtas to stay – he’s not a temporary CEO and he’s not going back to Bessemer. ... Gary joined Cirtas after 3.5 years as an EIR (Entrepreneur In Residence) and out of several other CEO job offers."

Chris.

ZL Technologies lawsuit dismissed

Sent to me by Nancy Erskine:

"I’m the ombudsman at Gartner. It’s a position we established at Gartner more than 7 years ago to ensure the objectivity, independence and accuracy of Gartner Research. I thought you and your readers might be interested in the blog post http://tinyurl.com/26tgtw9 I wrote about the judge’s decision in the ZL case in 2009. I have also posted about the fact that Gartner research opinion is not for sale http://tinyurl.com/yhh2u2b . (In case either of these links don’t work my blog is blogs@gartner.com/ombudsman/."

I understand from a Gartner spokesperson that ZL appealed against the judge’s original decision to dismiss the case in 2009, but its appeal was also dismissed soon afterwards.

Chris.