Projecte

General

Perfil

Accions
Historial

Opció Twisted Storage

El JoseManuelCastillo va descobrir aquesta interessant alternativa a LOCKSS: http://www.twistedstorage.org

From: Jose Manuel Castillo <>
 Subject: Future Development of Twisted Storage
 To: 
 Cc: Ferran Jorba <>
 Date: Fri, 23 Mar 2007 11:37:10 +0100
 User-Agent: Mozilla Thunderbird 1.5.0.10 (X11/20070306)
Dear Mr. Wegrzyn,
My [[[[name]]]] is José M Castillo and I work in the IT Department in
 Autonoumous University of Barcelona (www.uab.cat), in Catalonia, Spain.
First of all, let us make you know how interesting and potentially
 useful is for us Twisted Storage, and how we appreciate your efforts
 into developing this project.
We are currently studying a solution for a CAS system for our
 library. They need to preserve a large (several TBs, and growing)
 quantity of scanned material (TIFF files).
We were very happy and excited when we found Twisted Storage, because
 we thought (and still do) that it would suit our needs perfectly, and
 also, because it was an open source project. Since then, we follow
 with a lot interest the development checking frequently
 twistedstorage.org and also your blog.
In the roadmap you pointed that every two or three weeks it would come
 out a new release of the software, but version 0.1.5 is about five
 months old. Please, we would like to ask you if you plan to keep on
 with the project and releasing new versions of it.
Again, thank you very much for Twisted Storage and for your attention!
Best Regards,
-- 
 José Manuel Castillo
 Servei d'Informàtica
 Universitat Autònoma de Barcelona
From: "Chas. Wegrzyn" <>
 Subject: Re: Future Development of Twisted Storage
 To: Jose Manuel Castillo <>
 Cc: Lance Urbas <>, Ferran Jorba <>
 Date: Fri, 23 Mar 2007 07:02:32 -0400
 User-Agent: Thunderbird 1.5.0.10 (X11/20070302)
Jose,
I have been working on the next release of the Twisted Storage
 platform. I have been working with some people in the data storage
 business to help figure out what is missing from the software. In the
 next 3-4 weeks I am planning on releasing a new version which will
 address all the features listed on the web site as well as aid some
 graphical utilities to make deploying and managing the system easier.  I
 am also working on an interface - using FUSE - to provide an NFS like
 interface to the system. This will provide a simpler way to get data
 into and out of the system (besides using SOA).
Are there features that you think more important than others for your
 project? I am looking to work with some groups to flush out features and
 get the product deployed and would consider your advice important.
All the best,
 Chuck Wegrzyn
From: Ferran Jorba <>
 Subject: Re: Future Development of Twisted Storage
 To: "Chas. Wegrzyn" <>
 Cc: Jose Manuel Castillo <>,
  Lance Urbas <>, Ferran Jorba <>
 Date: Fri, 23 Mar 2007 13:12:39 +0100
 Organization: Universitat Autonoma de Barcelona
 User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
Dear Charles,
I'm José Manuel colleague; he is more a systems guy (and he is the one
 who discovered Twisted Storage), I'm more into applications.  So we
 complement each other.
I talked to José Manuel about LOCKSS (that you may already know it;
 http://www.lockss.org), and when he looked at it, he complanined to be
 too intrusive; the machines are "locked in".  At UAB, we have 120
 library opacs (public access PCs) with Linux since 2001.  They are
 relatively new PCs, with a single partition of 4 Gb (plus swap, that
 is).  All the rest of the disk is unused.  This means about 7 Tb of free
 disk, already there, to be used!
José Manuel thought that we could combine this free disk, with Twisted
 Storage, to provide a sort of LOCKSS-like functionality using our
 infrastructure.
>  I have been working on the next release of the Twisted Storage
 > platform. I have been working with some people in the data storage
 > business to help figure out what is missing from the software. In the
 > next 3-4 weeks I am planning on releasing a new version which will
 > address all the features listed on the web site as well as aid some
 > graphical utilities to make deploying and managing the system easier.  I
 > am also working on an interface - using FUSE - to provide an NFS like
 > interface to the system. This will provide a simpler way to get data
 > into and out of the system (besides using SOA).
All this are excellent news!
>  Are there features that you think more important than others for your
 > project? I am looking to work with some groups to flush out features and
 > get the product deployed and would consider your advice important.
A while ago I wrote to a open-source software for libraries list about
 (http://sourceforge.net/mailarchive/forum.php?thread_id=31534460&forum_id=4708)
 solutions to store digitalisation files.  Here I wrote what are our
 priorities: 
But one of our most urgent problems is keeping those original TIFF
  (and their corresponding PDFs) safe beyond just storing them
  somewhere: I mean having more than one copy, doing backups, veryfiying
  checksums, automatically fixing the damaged files, maybe changing
  formats, etc.  This second part is already invented, and it is called
  LOCKSS (http://www.lockss.org).  It is a software with anything I
  could ask for, and more.
What I have been unable to find in the LOCKSS site is a configuration
  model where some libraries, in a local consortium, join together to
  keep jointly this material.  Ok, I understand that LOCKSS is designed
  to keep the material of [external] publishers.  But when I first
  learned about CLOCKSS (Controlled LOCKSS) I immediately thought that
  it would address the scenario we are facing in our consortium.
  However, I cannot find it in their web pages.
So, what is missing in Twisted storage?  Maybe detecting damaged files
 and recovering them from the other partners, using checksums or
 whatever, à la LOCKSS.
Best wishes,
José Manuel
 Ferran
From: "Chas. Wegrzyn" <>
 Subject: Re: Future Development of Twisted Storage
 To: Ferran Jorba <>
 Cc: Jose Manuel Castillo <>,
  Lance Urbas <>
 Date: Fri, 23 Mar 2007 09:02:19 -0400
 User-Agent: Thunderbird 1.5.0.10 (X11/20070302)
Ferran,
Nice to hear back from you. I know about LOCKSS and don't care for their
 approach. I always thought that storage - aggregated should be as easy to deal
 with as peer-to-peer applications.  While I always thought about Twisted
 Storage being a "central set of servers" with attached storage, a few days
 before your email I started thinking about a very distributed storage service
 very similar to what you are building!
Twisted Storage already has the ability to keep multiple copies of files!
 There is policy data kept per user of the system (I call it tenant
 information). This policy information describes where and how multiple copies
 are to be distributed, how long individual copies are to live (afterward they
 can be deleted) and even if you need to have the contents digitally signed and
 encrypted. If one of the copies disappears (before it is supposed to), the
 system will recover it from one of the other sites without your intervention.
 When you try to get back a copy of your file, the system will even find the
 copy that can be recovered the. quickest and return it to you.
As I said one of the complaints I received is that using the system is
 difficult. There was no management console and getting data into or out of the
 system required using webservices.  The newer version, which will be released
 shortly, has a management console that is built on the underlying Twisted
 Storage code: it looks like a regular peer-to-peer application just that it
 doesn't do storage. To me that shows the flexibility of the p2p approach.
Secondly, the NFS interface (and maybe the Windows CIFS interface) will make it
 easy to get data into and out of the system. You will just mount some server as
 an NFS (or CIFS) mount point. You won't have all the features of the system (it
 can do a lot via the webservices interface that isn't available via the NFS).
Can you tell me when you plan to build out your environment? It sounds like you
 have the hardware in place and the running systems and just need the storage
 platform. What version of Linux (distribution) are you using?
Regards,
 Chuck Wegrzyn
From: Ferran Jorba <>
 Subject: Re: Future Development of Twisted Storage
 To: "Chas. Wegrzyn" <>
 Cc: Ferran Jorba <>,
  Jose Manuel Castillo <>,
  Lance Urbas <>
 Date: Sun, 25 Mar 2007 23:59:49 +0200
 Organization: Universitat Autonoma de Barcelona
 User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
Dear Mr. Wegrzyn,
>  Nice to hear back from you. I know about LOCKSS and don't care for
 > their approach. I always thought that storage - aggregated should be
 > as easy to deal with as peer-to-peer applications.  While I always
 > thought about Twisted Storage being a "central set of servers" with
 > attached storage, a few days before your email I started thinking
 > about a very distributed storage service very similar to what you are
 > building!
I'm happy we agree.  When I talk to my colleagues, either librarians or
 computer people, either here from UAB or from the universities around, I
 insist that I want our backup to be in their computer center, 200 km
 away, and so on, so we all have copies of all our documents.  Not
everything from everybody, because that would be a waste, but enough
 so that a lost or damaged file can be reconstructed or found.  I also
 insist that we cannot do backups; this has to be done automatically.
 Also, those files have to be online (although not not at high-speed), so
 they can be checked periodically.  As a matter of fact, the LOCKSS
 project has the same goals, except that, according to José Manuel
 experience, LOCKSS imposes too much burden to be used as we were
 expecting.
>  Twisted Storage already has the ability to keep multiple copies of
 > files!  There is policy data kept per user of the system (I call it
 > tenant information). This policy information describes where and how
 > multiple copies are to be distributed, how long individual copies are
 > to live (afterward they can be deleted) and even if you need to have
 > the contents digitally signed and encrypted. If one of the copies
 > disappears (before it is supposed to), the system will recover it from
 > one of the other sites without your intervention.  When you try to get
 > back a copy of your file, the system will even find the copy that can
 > be recovered the. quickest and return it to you.
I see.  According to a fascinating paper I read recently (Failure Trends
 in a Large Disk Drive Population,
 http://labs.google.com/papers/disk_failures.pdf), Google keeps at least
 three copies of their files.  Of course, Google goes for immediate speed
 (http://www.google.es/intl/en/corporate/tenthings.html), but we don't
 need it.  Maybe two copies and their MD5 checksum in a third site could
 be enough, so there is three signatures to compare.  What do you think?
Although I haven't played with it yet (José Manuel has played with it a
 little bit), my interest is that those second copies automatically
 should be placed as far away as possible.
The Petabox Internet Archive Forum
 (http://www.archive.org/web/petabox.php) has some messages about how the
 Internet Archive has their datacenters in different continents.  This
 post is illuminating:
Our data protection system is to have at least 2 copies and preferably
   in distant locations (we have found that human error accounts for real
   loss as well, so having different administrative bodies helps). We
   keep copies in San Francisco and at the Library of Alexandria in
   Egypt. [...] I am in Europe for the next 2 months setting up a
   European Internet Archive that will host those machines in
   Amsterdam. I would be very interested in talking with anyone about
   what we are doing if this is of interest.
   (http://www.archive.org/iathreads/post-view.php?id=15555)
Although I find it unconvincing that they use rsync to syncronize them
 (http://www.archive.org/iathreads/post-view.php?id=19587).
A couple of interesting posts:
 http://www.archive.org/iathreads/post-view.php?id=41661
 http://www.archive.org/iathreads/post-view.php?id=62512
>  As I said one of the complaints I received is that using the system
 > is difficult. There was no management console and getting data into or
 > out of the system required using webservices.  The newer version,
 > which will be released shortly, has a management console that is built
 > on the underlying Twisted Storage code: it looks like a regular
 > peer-to-peer application just that it doesn't do storage. To me that
 > shows the flexibility of the p2p approach.
Looks good.
> Secondly, the NFS interface (and maybe the Windows CIFS interface)
 > will make it easy to get data into and out of the system. You will
 > just mount some server as an NFS (or CIFS) mount point. You won't have
 > all the features of the system (it can do a lot via the webservices
 > interface that isn't available via the NFS).
I'm lost here.  Or not.  Maybe I need to be a little practical
 experience with Twisted Storage.
> Can you tell me when you plan to build out your environment? It sounds
 > like you have the hardware in place and the running systems and just
 > need the storage platform. What version of Linux (distribution) are
 > you using?
We have already received the first 500 Gb USB external disk with
 thousands of 75 Mb TIFF scanned files of old periodicals, plus their
 OCRd PDF equivalents for Web publishing.  Fortunatelly, this first disk
 was only half full, but there are more to come!  In our disk array we
 have less than 1 Tb left for the library, so we are actively looking for
 solutions in the short, medium and long term.
We are using Debian stable; as you may know, at this moment is 3.1
 (Sarge), but next one (code[[[[name]]]]d Etch) is about to be released in the
 next few months; we could consider upgrade those 120 library Opac PCs to
 Debian Etch even before the official release, if needed.  We'll have to
 do it anyway.
To be continued...
Ferran
From: "Chas. Wegrzyn" <>
 Subject: Re: Future Development of Twisted Storage
 To: Ferran Jorba <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Mon, 26 Mar 2007 06:24:47 -0400
 User-Agent: Thunderbird 1.5.0.10 (X11/20070302)
Ferran,
It turns out in the Twisted Storage platform, if you have three copies
 on machines the system is capable of automatically recovering the
 copies. It was one of the features I felt was very important. The system
 is designed to require minimal human interface (once the policy is set
 the system will work automatically).
>>  Twisted Storage already has the ability to keep multiple copies of
 >> files!  There is policy data kept per user of the system (I call it
 >> tenant information). This policy information describes where and how
 >> multiple copies are to be distributed, how long individual copies are
 >> to live (afterward they can be deleted) and even if you need to have
 >> the contents digitally signed and encrypted. If one of the copies
 >> disappears (before it is supposed to), the system will recover it from
 >> one of the other sites without your intervention.  When you try to get
 >> back a copy of your file, the system will even find the copy that can
 >> be recovered the. quickest and return it to you.
 >
 > I see.  According to a fascinating paper I read recently (Failure Trends
 > in a Large Disk Drive Population,
 > http://labs.google.com/papers/disk_failures.pdf), Google keeps at least
 > three copies of their files.  Of course, Google goes for immediate speed
 > (http://www.google.es/intl/en/corporate/tenthings.html), but we don't
 > need it.  Maybe two copies and their MD5 checksum in a third site could
 > be enough, so there is three signatures to compare.  What do you think?
 >
 I read the same paper. Before I started writing the current version of
 Twisted Storage (I wrote an earlier system called LTS before this; it is
 used at a commercial environment today) I thought about having thousands
 of machines and tens of thousands of disk drives. I assumed there would
 be failures daily. It is why I designed and implemented Twisted Storage
 to be self healing; it just made a lot of sense.
As the system stands today, one of the policy options you can enable is
 to have the digital signature created of the document. That signature,
 and an option encryption key (and salt) can be sent somewhere else for
 storage. But, as I said it is an optionally enabled feature.
Can you tell me what you would use the signature for?
> Although I haven't played with it yet (José Manuel has played with it a
 > little bit), my interest is that those second copies automatically
 > should be placed as far away as possible.
 >
 > The Petabox Internet Archive Forum
 > (http://www.archive.org/web/petabox.php) has some messages about how the
 > Internet Archive has their datacenters in different continents.  This
 > post is illuminating:
 >
 Thank you for the reference on the petabox archive. I haven't seen it
 yet, but I am planning on doing so.
>  Our data protection system is to have at least 2 copies and preferably
 >  in distant locations (we have found that human error accounts for real
 >  loss as well, so having different administrative bodies helps). We
 >  keep copies in San Francisco and at the Library of Alexandria in
 >  Egypt. [...] I am in Europe for the next 2 months setting up a
 >  European Internet Archive that will host those machines in
 >  Amsterdam. I would be very interested in talking with anyone about
 >  what we are doing if this is of interest.
 >  (http://www.archive.org/iathreads/post-view.php?id=15555)
 >
 So long as the machines are connected via a TCP path, I don't care where
 the devices are located.  Will your network of machines support
 multicast? If not, will all the machines be in a broadcast subnet?
 Twisted Storage is optimized to use those two mechanisms (if available)
 to enable broad communication between peers.
> Although I find it unconvincing that they use rsync to syncronize them
 > (http://www.archive.org/iathreads/post-view.php?id=19587).
 >
 Rsync - fine for data processing. Poor for just about everything else.
 And I am speaking from practical experience (in the VoIP world).
> A couple of interesting posts:
 > http://www.archive.org/iathreads/post-view.php?id=41661
 > http://www.archive.org/iathreads/post-view.php?id=62512
 >
 Thanks for the information. Very interesting.
>>  As I said one of the complaints I received is that using the system
 >> is difficult. There was no management console and getting data into or
 >> out of the system required using webservices.  The newer version,
 >> which will be released shortly, has a management console that is built
 >> on the underlying Twisted Storage code: it looks like a regular
 >> peer-to-peer application just that it doesn't do storage. To me that
 >> shows the flexibility of the p2p approach.
 >
 > Looks good.
 >
 >> Secondly, the NFS interface (and maybe the Windows CIFS interface)
 >> will make it easy to get data into and out of the system. You will
 >> just mount some server as an NFS (or CIFS) mount point. You won't have
 >> all the features of the system (it can do a lot via the webservices
 >> interface that isn't available via the NFS).
 >
 > I'm lost here.  Or not.  Maybe I need to be a little practical
 > experience with Twisted Storage.
 >
 As one of the policies options you can have Twisted Storage talk to a
 specially designed application (that I am currently writing) that will
 make the data repository look like an NFS mount. Once you store data in
 the "repository" it is automatically added to the Twisted Storage
 platform. Can you tell me how you are planning to get data into and out
 of the storage system?
>> Can you tell me when you plan to build out your environment? It sounds
 >> like you have the hardware in place and the running systems and just
 >> need the storage platform. What version of Linux (distribution) are
 >> you using?
 >
 > We have already received the first 500 Gb USB external disk with
 > thousands of 75 Mb TIFF scanned files of old periodicals, plus their
 > OCRd PDF equivalents for Web publishing.  Fortunatelly, this first disk
 > was only half full, but there are more to come!  In our disk array we
 > have less than 1 Tb left for the library, so we are actively looking for
 > solutions in the short, medium and long term.
 >
 Interesting. How would you see these being added to the Twisted Storage
 platform?
> We are using Debian stable; as you may know, at this moment is 3.1
 > (Sarge), but next one (code[[[[name]]]]d Etch) is about to be released in the
 > next few months; we could consider upgrade those 120 library Opac PCs to
 > Debian Etch even before the official release, if needed.  We'll have to
 > do it anyway.
 >
 Thanks for the information. Do you expect RPMS to do software installs?
> To be continued...
 >
 > Ferran
From: "Chas. Wegrzyn" <>
 Subject: Re: Future Development of Twisted Storage
 To: Ferran Jorba <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Thu, 29 Mar 2007 07:46:02 -0400
 User-Agent: Thunderbird 1.5.0.10 (X11/20070302)
Ferran,
I am aware of the [[MogileFS]] project. It is a fine piece of software for
 what it does. If you want to build a small file system application it
 will work just fine. But I had some problems with it.
My first implementation of Twisted Storage - LTS (which I sold to
 InfoCrossing) - was a system that required a centralized database, like
 that in [[MogileFS]]. The problem is that when you have a system that needs
 to scale to thousands, tens of thousands or hundreds of thousands of
 files you run into a real problem. The first is that your system grows
 in two different ways. The first way is in the data storage - you need
 to add more boxes to support the storing of files. The second is that
 you have to scale up your database server (it becomes the bottleneck to
 the entire system). Twisted Storage solves the second problem by
 supporting the idea of a distributed database - it grows the same way as
 all the other parts of the system.
Secondly, there is no policy and protection in [[MogileFS]] as there is in
 Twisted Storage. By giving users direct access to the files you allow
 the possibility of hacking and changing the contents. Twisted Storage is
 meant to prevent that from occurring.  Once a file is deposited in
 Twisted Storage it is protected from direct access (at least without
 hacking the code and changing it). In addition the contents can be
 protected through encryption and digital signatures.
While it is enough to store multiple copies of a file the real problem
 is how does the system recover from failures? Twisted is designed to
 automatically recover from system and hard disk failures. It does this
 through its use of distributed databases (there is a database in each
 and every Twisted peer).
Finally, Twisted is designed to do more than merely [[[[name]]]] a file. A
 future version (some time this year) will do full content indexing,
 allowing documents to be found by searching for words (it will use an
 open source index utility).
Now for your environment perhaps Twisted does too much and all you need
 is [[MogileFS]]. If that is the case, it will be the right choice.
All the best,
 Chuck Wegrzyn
From: Ferran Jorba <>
 Subject: Re: Future Development of Twisted Storage
 To: "Chas. Wegrzyn" <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Thu, 29 Mar 2007 16:29:00 +0200
 Organization: Universitat Autonoma de Barcelona
 User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
Dear Mr Wegrzyn,
thanks for answering.  Again, we don't want to be misunderstood: Twisted
 Storage seems to us the most ambitious and complete solution for us.  We
 really like it!  But the release José Manuel tested last week didn't
 work on Debian Stable, at least without much effort.  We are eager to
 test your new release with our limited human resources.
But until then, we may give a chance to [[MogileFS]], even knowing the
 limitations you have told us, because we must implement a solution
 quickly, as the TIFF files are arriving at alarming rate and our disk
 array is almost full.
But of course, in the future we'd like to use Twisted Storage.
By the way, we, as a public university (a non profit organization),
 which license should we use?  According to your LICENSE file in the
 distribution, 
Twisted Storage is distributed un the GNU General Public License
   version 2 for private and academic use only. Use in a product or as a
   service or part of a service requiresa commercial license. For
   commercial licenses please contact Retriever Technologies, Inc at
   .
Thanks again,
Ferran
From: "Chas. Wegrzyn" <>
 Subject: Re: Future Development of Twisted Storage
 To: Ferran Jorba <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Thu, 29 Mar 2007 11:28:16 -0400
 User-Agent: Thunderbird 1.5.0.10 (X11/20070302)
Ferran,
I didn't understand. I know that you have a need to find the most
 appropriate solution to your particular problem. Why it didn't work on
 Debian, I don't know. So long as it has a working 2.4 Python system and
 followed the instructions on the website it should have run. By that
 aside it is a pretty new project and requires some hand holding to get
 it working; that is why I am building a management console.
You are covered under the GPL v2 license, no doubt about it. My
 license was to make certain that private companies can't install and use
 the system and resell the services.
Please stay in touch. I would like to know how your project is going
 and what you are learning.
All the best,
 Chuck Wegrzyn
From: Ferran Jorba <>
 Subject: Re: Future Development of Twisted Storage
 To: "Chas. Wegrzyn" <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Fri, 30 Mar 2007 13:17:35 +0200
 Organization: Universitat Autonoma de Barcelona
 User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
Dear Mr. Wegrzyn,
>   You are covered under the GPL v2 license, no doubt about it. My
 > license was to make certain that private companies can't install and use
 > the system and resell the services.
Oh, great!  Thanks!
>  Please stay in touch. I would like to know how your project is going
 > and what you are learning.
Sure we will.  I'm starting my Easter vacations today; José Manuel is
 staying a few more days.  We have starting also to look at cheap storage
 nodes, autonomous PCs, flexible enough to host any of the software
 alternatives we are evaluating, because we've seen we won't have enough
 raw capacity with the public PCs in the libraries.
Abusing of your broader experience, do you have any top-of-the-head
 recommendation?  (without doing any research, I mean).
Thanks again,
Ferran
From: "Chas. Wegrzyn" <>
 Subject: Re: Future Development of Twisted Storage
 To: Ferran Jorba <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Fri, 30 Mar 2007 07:29:26 -0400
 User-Agent: Thunderbird 1.5.0.10 (X11/20070302)
Ferran,
I will be more than happy to help you both out and your project. The more
 practical experience I have in live deployments (even not using Twisted
 Storage) means that I can figure out what needs to go into my project. So
 please feel free to call on me whenever you need.
What are you looking for?
Regards,
 Chuck Wegrzyn
From: Ferran Jorba <>
 Subject: Re: Future Development of Twisted Storage
 To: "Chas. Wegrzyn" <>
 Cc: Lance Urbas <>,
  Jose Manuel Castillo <>
 Date: Fri, 30 Mar 2007 13:47:48 +0200
 Organization: Universitat Autonoma de Barcelona
 User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
Dear Mr. Wegrzyn,
>   I will be more than happy to help you both out and your project. The more
 > practical experience I have in live deployments (even not using Twisted
 > Storage) means that I can figure out what needs to go into my project. So
 > please feel free to call on me whenever you need.
Your generosity overwhelms me.
>  What are you looking for?
Basically, we have searched and seen, besides the big Capricorn Tech
 (http://www.capricorn-tech.com/), Copan Systems
 (http://www.copansys.com/), and the like, specialized vendors that sell
 rackable U nodes that could fit into our computer room lie LaCie's
 Ethernet disks (http://www.lacie.com/es/products/product.htm?pid=10836),
 or Infrant's ReadyNAS (http://infrant.com/products/compare.php).
The other option could be the cheapest (lowest CPU, 512 Mb RAM but as
 much disk as possible) HP rackable PC (we have good experience with
 HP).
My question, in general terms, is that if you have a quick
 recommendation about which kind of hardware should we buy (then, we'll
 have to check if they sell in Spain, that is).
José Manuel will follow this thread, as I'll be out in a few minutes,
 and he know hardware much better than me, the applications guy.
Best regards,
Ferran
PS By the way, you can see the public side of all this storage needs at
    http://ddd.uab.es/?ln=en .  It is a young project, and many things
    are still missing.  We are using a GPL, Apache+Python+MySQL based
    software written at CERN that used to be called CDSware and now
    re[[[[name]]]]d CDS Invenio (http://cdsware.cern.ch/invenio/).  The CERN
    production server has almost a milion documents
    (http://cdsweb.cern.ch/).

Actualitzat per fa més de 17 anys · 0 revisions