Slashdot Mirror

The most beautiful data center I ever saw was the USC ISI data center in Marina Del Rey, California. It's been there since at least the 1970s, and was part of the early creation of the Arpanet/Internet. It's on the top floor of a 12 story building, that at least in the olden days had no other tall buildings around it because of earthquake risk. It was easier to cool because it was near the roof air conditioning compressors. It has panoramic windows all around overlooking the beach and the ocean. There were special rails to keep hardware from falling out the windows in case of earthquake.

https://www.isi.edu/about/hist...

What linux virtualization solution would you recommend for hosting Win10 + games with gpu virtualization?

I haven't tried it myself yet, but there are multiple reports of success with Ryzen+KVM+GPU+W10, for example this one.

You say its 'good enough' to run AAA games. What sort of performance hit am I really facing?

My impression is, very little. GPU virtualization gives the guest OS direct access to PCI registers, the overhead can get very close to zero. This report from 2014 shows overhead consistently less than 3%, often a lot less, and remarkably, sometimes actually faster in the VM. I'm not sure how that last one works.

The big overhead for VMs tends not to be CPU, but memory consumption, make sure you have enough to make both host and guest comfortable. You should be fine with 16 GB, but more memory is always better, I'm liking how it feels with 32 GB. You will want a separate SSD for Windows, I think, but that's not going to break the bank.

Do some games "just-not-work" What sort of stability loss am i looking at?

Again, I'm not doing it myself right now (I have too many unplayed games already without a bunch more from Windows) but I see multiple reports of success with GTA 5 and I don't see any horror stories. My feeling is, your system as a whole will be more stable than it is now, and the VM+Windows part of it will be exactly as stable as now.

I've got an i7 and a gtx1080, if that's a factor.

Though I am a newly-minted Ryzen fanboy, I love Intel too except for their business practices. VM stability seems exactly the same for Intel and AMD. That is very cool. Number of VM crashes I had over the years on Intel or AMD: exactly zero, and I really thrash those VMs.

What's the situation with multi-monitor support with something like this?

Dunno. I'm waiting for your report. The question you ought to ask is, what's the situation with sharing the GPU between host and guest? Lots of active discussion on it. It's a thing, and multi-monitor passthrough is a thing.

And peripheral pass through? (usb headsets, usb controllers).

KVM has good USB passthrough, but for mouse and audio where performance is not an issue you probably want the virtual devices. There are a whole pile of online resources on it, e.g. here and the community is active. Mostly people seem to be using libvirt and virt-manager. I don't, I just read the man page and run KVM/QEMU from the command line. Do that only if you enjoy that kind of thing.

There is a great and supportive community here.

A team at USC/ISI has been working on this kind of thing for quite some time (there are subsequent stories showing successful-ish results, but I figured you wanted something early):

http://www.isi.edu/news/story/...

This is cool stuff. Here's some other stuff I found recently for anyone interested in messing with bitstreams, creating an open-source FPGA, or doing hardware more easily. Hardware designers feedback is appreciated.

Open Source Bitstream Generation without R.E. or license violations: http://www.isi.edu/~nsteiner/p...

Archipelago - an open-source FPGA with toolflow support: http://www.eecs.berkeley.edu/P...

Cx, open-source, hardware & synthesis language: http://cx-lang.org/

QFlow Open-source Flow from behavioral synthesis to detail routing: http://opencircuitdesign.com/q...

Have fun people! Especially building on the first two. I'd appreciate experienced people telling me how good the Cx system is for (a) people doing FPGA with high-level synthesis tools and/or (b) beginners using behavioral verilog wanting something better.

I'm wounded. I'm a troll, but not that kind; I'm the kind that lurks in caves or under bridges, but harmless because sunlight is an existential threat to me.

I wasn't really kidding, either. Those were my exact thoughts 30 years ago as I sat at my Z-100 computer logged into my local DEC machine (across the flight line) through a 300 baud acoustic modem, manually updating the system's HOSTS file from the latest "master copy" I FTP'd down from ISI.Good time... good times.

I thought it was fairly conclusive that it wasn't a cypher - the symbols simply lack the entropy to represent language. It's just what you'd expect from someone combining a few symbols in nonsense ways as a hoax, and not statistically what cyphertext looks like at all. A bit disappointing, really.

That is wrong. The word entropy is similar to English, and, while the second order entropy is low, it is similar to Polynesian languages.

This is a nice nice review of Voynich studies.

One can do this kind of research without hacking into other peoples machines as well as reference xkcd, e.g. look at here or here. The second one is also far more comprehensive in terms of alive IPs discovered.

Nevertheless, I enjoyed reading this "research paper".

You should look into using DETER Testbed for this type of class. http://www.isi.edu/deter/ In short, DETER is an Emulab clone for Network Security type of experiments. There is an on-going effort to promote DETER for education purpose and the DETER team has access to several types of virus. https://education.deterlab.net/ Hope this helps. Young

Here you go...

And here...

You may want to consider a scientific workflow system. These systems handle both data storage (including meta-data and provenance -- where the data came from), and design and execution of computational experiments. If you are concerned about the complexity of the meta-data (e.g., pH value..) and would like to make sure to be able sort things according to this, you want to give "Wings" a try. You can try out the sandbox to get an idea: http://wind.isi.edu/sandbox.

1) Yes, you could.

2) When you have a workable method for sending a postcard to every IP address, let me know. Mapping IP address to street address is a neat trick if you can pull it off. Just don't rely on WHOIS, for obvious reasons.

* ns2: http://nsnam.isi.edu/nsnam/index.php/Main_Page (GPL)
* Wide Area Network Emulator http://wanem.sf.net/|WANem (GPL)

I agree with happylight . network simulator is i think what you want . It includes utility where you can visualise the data packets sent from computer nodes . You can also write simple scripts to adjust network settings such as capacity of link etc . It also demonstrates the protocols . we have used it in our computer network course at college . Find more at this link : http://www.isi.edu/nsnam/ns/edu/index.html .

Not sure if it's exactly what you want but check out ns2 / nam. It's a pretty good network simulator. It's open source and runs on linux.

Not sure if it's exactly what you want but check out ns2 / nam. It's a pretty good network simulator. It's open source and runs on linux.

"DNS" was a "HOSTS.TXT" file FTP'd down from ISI.

Now stop doing zone transfers across my lawn, you punks!

This sounds similar to the goals of the NewArch project, http://www.isi.edu/newarch/ back in 2003? The original idea seems to have died with the funding?

Right. Most people are sitting on unaddressable addresses. The ANT census is pretty explicit on this point. Roughly 4% of the IPv4 address space is in use, 30% is not allocated at all, and the remainder (66%) is trapped due to inefficient allocations.

Sounds a bit like this. I've always thought that it looks pretty cute.

The researchers do address your points:
http://www.isi.edu/ant/address/

Internet Census Taking,
We believe we have taken the first census contacting each address of the visible Internet since 1982 (RFC-832). A census is an enumeration of all allocated addresses. We probed all 2.7 billion allocated addresses (compared to 315 in 1982). We are able to probe only the visible Internet: all addresses that use public addresses and choose to respond.
Accuracy,
No census of billions of addresses will be perfect; we underestimate occupancy for three reasons: A few percent of probes and replies are lost due to congestion. Addresses such as those behind firewalls choose not to receive or reply to our requests. Other computers use private addresses. We evaluate loss in our technical report; evaluation of the other cases, the invisible Internet, is future work.

A preprint of the paper is available
at http://www.isi.edu/~johnh/PAPERS/Heidemann08c.html

I encourage folks to review the actual, peer-reviewed paper before commenting on the methodology.

If you want to see the original pop-up free article on the census is here - http://www.isi.edu/ant/address/ and there is a pretty cool interactive map here http://www.isi.edu/ant/address/browse/index.html . Enjoy.

If you want to see the original pop-up free article on the census is here - http://www.isi.edu/ant/address/ and there is a pretty cool interactive map here http://www.isi.edu/ant/address/browse/index.html . Enjoy.

There is an interactive map on their site that allows you to zoom into the IP space pretty nicely. Our uni has a B range of addresses and we use only two Cs of that right now. When we split off from the main building and got onto city fiber, they decided that, rather than give us a private IP range like the other campuses, we would be allocated one of the C ranges.

Of course, no one knew what they were doing so getting the ASA and default routes set-up properly was a nightmare, but hey, we're using more of our IP space now! (sarcasm intended)

JDB is a package of commands for manipulating flat-ASCII databases from shell scripts. JDB is useful to process medium amounts of data (with very little data you'd do it by hand, with megabytes you might want a real database). ... By storing data in simple text files and processing it with pipelines it is easy to experiment (in the shell) and look at the output. The original implementation of this idea was /rdb, a commercial product described in the book ``UNIX relational database management: application development in the UNIX environment''.

I think you need to read more about the IPv4 allocation. "NAT" needs to be deployed by corporations for their internal workstations. This is the true cause of the shortage. Companies like Apple (gasp) have ancient class-A allocations--Yes you read that right, Apple controls a block of 16 Million IPs. Of these, only a few thousand are routable or whose ports are publically reachable (due to firewalls).

So stop the FUD. The question isn't whether home users will get NAT'd by their ISP. The question is whether 30 year old allocations need to be rebalanced to place more addresses in the hands of network service providers.

At my company all of our workstations are behind a NAT. We have fewer than 16 routable IPs but control the equivalent of an old "C" block allocation.

You should take a look at this: census vs. allocation. Only a small fraction of IPs are really in use.

There is no new IP. There is only unallocated IP.

Assuming that you expand IP as "intellectual property", U.S. law[1] does not recognize a single "property" (exclusive) right in "intellectual" creations. It recognizes exclusive rights in works of authorship, which are separate from the exclusive rights in inventions, both of which are separate from the exclusive rights in distinctive marks. Which do you mean?

I don't know what typeface they used the image on for http://www.isi.edu/ant/address/ but it looks very odd. It's hard to concentrate on the whole image. I mean, look at the 'c's!

Don't you just hate it when the internet wraps onto the ceiling. All those packets are horrible on the acoustic tiles.

And once it gets up there you know its going to be hard to get it back down.

http://www.isi.edu/news/print.php?story=163 has some more top-level details. The chip has six identical RISC processors with associated embedded DRAM memories and a stream-processing programmable hardware core, along with a lot of on-chip and off-chip I/O interconnect bandwidth.

Links to pictures and videos from this page: http://www.isi.edu/robots/superbot.htm

http://www.isi.edu/craft/CC/Welcome_files/resource s/media/CCmachine.wmv

Looks like the machine is building a large penis

You aroused my curiosity, and it turns out that the video at http://www.isi.edu/craft/CC/Welcome_files/resource s/animation.html (thanks to mindriot for pointing this out) shows a simple solution. For those on limited bandwidth connections, the basic gist of it is that the floor & walls are "printed" and then a separate robot arm picks up some flat (almost I-beam looking things) that it lays across the roof. The I-beams are then "printed" over to hold them in place & seal them.

http://www.isi.edu/CRAFT/

Much more details.

A few links could of course have helped this article... I think contourcrafting.org seems to be more or less the right page for the California project. The videos and animations are quite worth seeing.

For the Loughborough one, the closest I could come up with was Dr Soar's website...

I'm going to zig to every other replier's zag.

Try this objection on for size: I'm supposed to be able to read a patent and implement the device. I recently saw Mythbusters do just that for a couple of old patents. But software patents by my estimate get you far less than 1% of the way to an implementation.

A patent on the web would describe "a client machine that requests content from a server, and the server receives it." A series of further elaborations might be made mentioning that it could be cached, could travel over the internet, could travel wirelessly, etc.

However, I guarantee you the patent won't contain anything like this, which is at least enough to implement it, even if it still only constitutes a specification with no code whatsoever. (That's the HTTP 1.1 RFC, so you don't have to follow the link.)

There are a bare handful of patents this doesn't apply to, but all but one software patent I've ever read looks like this.

The reason most patents seem so obvious is that the mere statement of a problem is hard to make interesting.

The other basic problem is that to the extent that the difficult part is seeing a problem, many patents "patent" the only obvious solution. Realizing that people want a one-click checkout might be a good idea that nobody has. (Note the "might"; actually I think it's the obvious conclusion of a long refinement process of making checkout simpler, but maybe you disagree.) But once you say to me "Hey, I want a one-click checkout", the technical implementation is obvious. In fact, since there's only one basic way to do it it's actually even more obvious than a conventional checkout.

The evaluated http://gate.ac.uk/ which is GPL software but ended up using http://search.cpan.org/~acoburn/Lingua-EN-Tagger/. There are several other tools in this space that can be glued together to create this type of software:

http://www-nlp.stanford.edu/
http://tcc.itc.it/research/textec/tools-resources/ jinfil.html
http://wordnet.princeton.edu/
http://www.alias-i.com/lingpipe/web/faq.html
http://www.isi.edu/licensed-sw/halogen/index.html

Not trivial, but if you wanted to DIY, you don't need to start from scratch. Though, having a bunch of hardware to chug through 1000s of documents would still be needed :).

I did RTFA. If you had RTF design principles of the Internet (see IEN 48), you'd know that it is a "network of networks". So, if large numbers of "local networks" are "brought down by lots of users trying to stream World Cup footage", then it effectively is an "Internet meltdown". As for how large a number of local networks predicted to be affected, one article states: "experts in the networking industry warned that staff using the service could bring UK corporate networks to their knees. If all the corporate networks in the UK were to "meltdown" or be "brought to their knees", then I'd say that qualifies as an "Internet meltdown". At least a "regional" meltdown.

If all the corporations in the UK suffered blackouts due to an unexpected draw of electricity by employees, I think it would be fair to call the result a "power grid meltdown".

Note, I'm sceptical this will happen. That's what prompted my submission. I'm looking for evidence that a global event such as World Cup ever caused major traffic problems across the Internet, either at the leaf networks or the backbone networks.

USC-ISI is actually working on something that could be much more incredible than this. Their project (still in research phase) is known as SuperBots. They want to be able to build large self-configurable robots that can adapt to any terrain. They have videos available. These guys are currently being looked into by the LEAG (lunar exploration analysis group) of NASA's JPL (here is some commentary by an actual JPL engineer who went to the last LEAG conference). It is basically a nerd wet dream to imagine these things on a 10x scale powered by small nuclear engines stomping all over Mars and the Moon doing experiments, exploration, and generally being awesome.

USC-ISI is actually working on something that could be much more incredible than this. Their project (still in research phase) is known as SuperBots. They want to be able to build large self-configurable robots that can adapt to any terrain. They have videos available. These guys are currently being looked into by the LEAG (lunar exploration analysis group) of NASA's JPL (here is some commentary by an actual JPL engineer who went to the last LEAG conference). It is basically a nerd wet dream to imagine these things on a 10x scale powered by small nuclear engines stomping all over Mars and the Moon doing experiments, exploration, and generally being awesome.

For a second there, I thought the article summary was refering to one of the projects here at USC, the Superbot, part of our Polymorphic robotics Laboratory. Superbot is really about chains of tiny robots that can connect together to adapt to the particular task at hand. I know know low-G environments is one of our targets (NASA funded) and we have 2D air hockey table prototypes, and even toyed with underwater robots. But the videos show we have a long ways to go.

Anm

In other words, TCP is a protocol, not an algorithm.

So ... if Vista has some fabulous new algorithms for implementing TCP, then why can't other OSes be patched to benefit from those algorithms also? OR, if Vista is implementing something other than TCP, then how can it be (fully) backwards compatible?

Seems like the word "compatibility" might need to be scrutinized here.

About time for someone to implementRFC 3093, "Firewall Enhancement Protocol", also known as IP-over-HTTP.

You just don't get what he's saying, and you're not making any sense.

"The only time FTP has less overhead than TCP is when you're retrieving several files."

I'm going to make a guess here and assume you mean HTTP, not TCP.

First, take a look at the FTP RFC.
http://www.freesoft.org/CIE/RFC/959/index.htm

Then, take a look at the HTTP 1.1 RFC:
ftp://ftp.isi.edu/in-notes/rfc2616.txt

You tell me which has more overhead? A notable part of the difference is the encoding; FTP can transfer data straight binary - no MIME types or special encoding to send the data over the channel.

"the overhead of FTP can be significantly higher than HTTP (logon banners)."

Are you kidding?

" For HTTP, you send the request and sit back and wait for the data. "

If browsers were as simple as an FTP client, this might be true. But don't forget about all the banners and lots of extra data that gets communicated between your average browser and HTTP server these days. Not to mention cookies.

"With FTP, you have to login (USER, PASS), which both require you to wait for confirmation before you can PORT and RETR."

All of this is is likely done in less then 100 bytes of data transferred.

"Not to mention the overhead of establishing another TCP socket to pass the data over."

Here's a quick run down of how a TCP connection is established:

1. Packet sent from initiating machine. Very small packet (bytes) with the SYN flag set.
2. Recieving machine gets packet with SYN. Sends packet back (bytes) with the SYN and ACK flags set.
3. Initiating machine sends back another small packet (bytes) with the ACK flag set.

The amount of data necessary to open a raw TCP connection is so miniscule that it's almost not worth mentioning.

"If you need to retrieve a tree structure of files, download several files from a single server, or need to upload files, FTP is the way to do it. If you need to download only one file, or several files in parallel (typical webbrowsing), then HTTP is your friend."

You're looking at this from a user perspective, not a technical one.

FTP is very low overhead (read: almost zero,) it's a very intelligent design, and it works great over slow and unreliable connections to boot.

Nobody is saying we should replace HTTP with the FTP protocol.

you can read the papers here. I recommend the first few pages of "NewArch Final Technical Report" to find out answers to your questions

NewArch Project: Future-Generation Internet Architecture

Although it doesn't not support HTML or images,
notes-mode for emacs (or xemacs) was designed for research log applications. Its real strength is automatic indexing, so when you have 10 years of notes, you can find what you care about.
See http://www.isi.edu/~johnh/SOFTWARE/NOTES_MODE/

Has no one read RFC 3675?

Here's what's going on. Google has a flaw in their algorithm in the way it deals with 302 redirects. Instead of following the internet standard of following the 302 redirect to the new location and indexing that, Google sometimes applies the information, PageRank, and Google ranking position of the 302 destination page back to the page doing the redirecting.

This doesn't sound like a bug at all. This sounds more like Google actually bothered to implement the HTTP specification correctly, and the web site with the redirect did not. Go read it and pay careful attention to the subtle but important differences between 301 and 302. The methods by which user agents (including search engines) should handle these responses are outlined there.

Sensor Network research has been going on for quite some time. CENS isn't the only place doing the research. Some of the original work actually came from SCADDS from USC/ISI, with some of the same people now at CENS. All the UCs are very involved in Sensor Networks, Berkeley for example was the orginator of the current most popular hardware, the motes, now manufactured by Crossbow. Intel also makes their own version of the motes, though they are not sold comertially yet.

Most of the hardware runs in a specific OS called TinyOS, which is open source. Other hardware, like the Stargates (also from Crossbow) run an arm version of debian.

You can find lots of neat info about Sensor Networks from the specialized conferences like Sensys, IPSN, etc.

Most of the projects done with Sensor Networks have been geared towards the biological monitoring fields. However, the funding comes in from NSF as well as DARPA so sometimes it is security/military based.

It's an interesting world out there...

cl