I recently investigated this for Solaris, and I'm sure some of the products will be on Linux as well. Also, check the Linux-HA page at http://www.linux-ha.org/.
The basic idea is to do clustering: two machines that share a single (virtual) IP address, in addition to their real IP address. You then have clustering software that detects a failure in the other machine, or in an application, and fails over to the other machine, starting the applications that failed, and doing a gratuitous ARP to bind the virtual IP address to the MAC address for the surviving machine.
Client apps must be able to re-connect to the server, since their TCP sessions are dropped when the primary server crashes. TCP state failover as in some firewalls (e.g. FW-1 from Check Point) would be very handy but I don't know any OSs that do it.
This requires a shared disk subsystem - initially this is usually SCSI, with two controllers and software that can handle this. As systems grow, they tend to migrate to SANs (storage area networks), usually based on Fiber Channel - this is very fast, as you might expect, and can be built using FC switches, so your SAN can be redundant, as well as your servers. You would of course need RAID 1 or 5 in your disk subsystem.
The next step is to do volume replication - this can be nearly instantaneous (you have a choice of synchronous replication, which slows down every update transaction on the main server, or asynchronous replication, which is a little less safe). The trick here is to make sure that the volume replication software can buffer updates during times when the secondary server/disk is not available - otherwise a single failure stops all transactions...
Finally, global cluster management involves failing over between geographically separated systems - this would require the client apps to know how to switch to a different IP address, though you might be able to rig something up with load balancing technology.
This is a horribly complex area, as I discovered, and it's not simple to get it right. There are many techniques I have not covered - good sites to read up on are veritas.com, logitech.com, sun.com (search for NDR and Sun Cluster), technet.oracle.com (Oracle focused but covers many options), and of course linux-ha.org.
Good point, but IMO browsers should be strict in what they send and liberal in what they send (the old IETF credo), i.e. they should be able to display something even if the WML is bad. If they really can't display anything they should show an informative error message so I can harass the WAP site or the WAP gateway as appropriate.
Since I'm using Orange in the UK, Nokia is also supplying the WAP gateway...
Anyway, as a 'plain user' I would probably have given up on WAP a long time ago due to these hassles - it's quite amazing to me that WAP works so badly for such a high percentage of pages. I happen to have a professional interest in WAP, GPRS, 3G, etc, so I keep on trying occasionally.
Finally, I wrote a program that takes arbitrary input line-by-line, and outputs sorted unique lines with a count of how many of each line it found. It took me 7 lines of Perl code, (counting the comment to run the Perl interpreter:) and 50 lines of C.
Ah, you mean like 'sort | uniq -c'?:)
Still, Perl is very nice when your problem doesn't neatly fit the Unix tools approach, e.g. multiple-line records or whatever.
I have a 7110 and occasionally use it for WAP - I would use it more, but about 1/3 of all pages I browse, including some portal sites, just can't be displayed. So calling this a good browser is rather an exaggeration, though it may well be better than the others you tested.
Linux on S/390 is there because they want to port e-business apps to all their existing platforms, and because they understand Linux is going to commoditise operating systems. Unlike many other vendors, IBM really seems to 'get' Linux.
One useful application for S/390 Linux IMO is hosting very large numbers of dedicated-server websites on a single box - every customer gets their own virtual Linux machine, hosted on VM/390, which has 30+ years of tuning behind it.
The customer gets more power and flexibility, and the hosting provider gets more maintainable and reliable hardware, rather than having racks upon racks of Intel-based servers that will inevitably have frequent individual box failures, even though most of them are available at any one time. The only flaw in this argument is the cost of mainframe hardware and system programmers, but if you host enough websites it could be worth it.
Xunker is right, if you have a RISC-based AS/400, i.e. a modern one, it is using PowerPC under the hood.
IBM has recently committed to porting Linux to the AS/400 platform (not to OS/400, the operating system) - this will be done very much the same way they already ported AIX, i.e. you'll have Linux running directly on the PowerPC hardware, with some extra software to arbitrate between OS/400 and Linux where needed, and handle Linux-OS/400 communications.
Should be very interesting when it's all done - every single IBM platform, including the most proprietary AS/400 and S/390 ranges, will be able to run Linux apps, and porting should be relatively easy providing apps are not byte order dependent etc.
Ironically, IBM tried in the 80s and 90s to provide a way of easily porting apps across all its different OSs and architectures, through a set of development tools and APIs known as System Application Architecture (SAA). They failed. It would be rather cool if Linux solved this problem for them:)
I don't think it's just about glibc - I use Netscape 4.75 on glibc 2.1.2, which is the latest for my distro (Mandrake). Netscape crashes so frequently I have a terminal window dedicated to restarting it - mainly when I close one window, or when I do something requiring POP3 or HTTP authentication.
Oops - you are right, i.e. most of Abilene is on IPv4 not IPv6. However, there is a 'toy backbone' of 2 core routers and 2 campus routers running IPv6, according to the Abilene IPv6 pres at http://www.ipv6forum.com/navbar/globalsummit/slide s/html/michael.lambert/sld021.htm
vBNS, the other Internet2 backbone, also has a similar 4-router configuration, though in this case the routers are all core type routers, serving Chicago, San Francisco, Maryland, etc.
Both backbone teams seem to be in 'experiment with IPv6' mode, no doubt due to the learning curve and scarcity of routers that actually support IPv6.
I forgot to say that Opera on Linux is the only browser that manages to max out my puny 56K modem connection (as reported by the neat Knetload applet, which graphs throughput). [I've also tried Netscape 4.75 and Mozilla's recent nightly builds, which are horribly slow for some reason. I tried to get Galeon going but ran into.so problems.]
I think this line efficiency is because you can set it to use a large number of simultaneous connections, which is a bit mean on the website but very efficient for dialup users, whose TCP time-to-connect is quite long due to modem latency. Hopefully the server OSs will evolve to handle this sort of thing efficiently.... Of course this could be HTTP/1.1 pipelining but I doubt it as I am doing all this via Junkbuster (junkbuster.com) ad filtering, and Junkbuster requires HTTP/1.1 to be turned off (it only analyses new connections.)
Quite apart from maxing out the modem, Opera is very fast at rendering pages; it also has good in-memory caching so it *never* goes to the network or disk when you hit the Back button (unlike Netscape).
More Opera info is at www.opera.com, including a pretty usable tech preview for Linux (i.e. alpha) - crashes and has missing features, but is quite usable, in fact I'm using it now (except that I can't post to Slashdot since the HTTP login stuff is not working...). They also have versions for EPOC (Psion etc), BeOS, OS/2, etc, and hope to support non-Linux Unices as well.
Opera doesn't actually have its own JVM; however, the downloadable versions for Windows come with a JVM from Sun, in the same download and as part of the same install sequence.
As for the cost - Opera is particularly good on older hardware (my mother used it on a 486SX/25 with 8 MB RAM and Windows 3.1 for quite a while, and it was a lot better than Netscape, and I used to use it on a slow P200 laptop). So if you can pay $30 to avoid a hardware upgrade, you are way ahead financially.
Opera is missing a few features such as password management, etc, but its speed and ability to turn off images with one click is enough for me. I'll be registering the Linux version as soon as it gets out of alpha/beta, as it is stunningly fast on my AMD K6-2/350 as well.
IPv6 has been developed largely in public. Although you can use it with static IP addresses, based on the NIC MAC address, you don't have to use this - in fact you can assign whatever IPv6 address you feel like to an interface and use that instead (as long as it's routable).
IPv6 addresses can't be hard coded into NICs, because part of the address derives from the network provider and the site.
The result is that IPv6 can be just as anonymous as IPv4, with some reasonable setup, though by default it's possible your IP address will be quite static. No doubt privacy-enhancing tools will make it easy to randomly choose the lower part of your MAC address to get some privacy back, just as analogous tools block cookies etc.
If you must be so paranoid, why not at least learn about how IPv6 works before you start posting?
Actually not all of Internet2 is IPv6 - Abilene is, I think, but some of the other testing e.g. the Qbone for QoS is still on IPv4, for logistical reasons.
There's been lots of work on migration of IPv4 to IPv6 - or more correctly, coexistence, since it's quite possible IPv4 will never disappeare completely, just like DOS... The details are fairly complex, but there are various tunnelling schemes (some including automatic tunnel setup as required) as well as protocol translators that let an IPv6 domain talk to IPv4 land via (you guessed it) something like a NAT.
In time, hopefully, the IPv6 domains will get larger and larger and gateway directly to each other - the 6bone, which is an international IPv6 network, is currently a mixture of tunnels over IPv4, and some 'real' links that are native IPv6. There are even ISPs that have rolled out native IPv6 service, e.g. NTT is one that has done quite a lot in this area.
IPv6 is particularly useful to Asia and other non-US/European regions, which didn't get much IPv4 address allocation and now really need the address space. It's also important for the massive mobile Internet roll-outs that are happening over the next few years. Just as soon as Microsoft, Cisco and others start shipping IPv6 as standard (quite soon now) it will have a chance of taking over, though it will take anywhere from 5-10 years IMO.
The Grid doesn't really exist yet - it's a bundle of various technologies, including middleware, service brokers, and lots of quite specialised software, along with some standard protocols such as IPv6 et al, and no doubt some specialised ones.
Internet2, by contrast, has little to do with host software per se - it's all about engineering the network for higher and more consistent performance (using QoS), making multicast work better, and getting IPv6 working in production mode (not all at the same time, necessarily).
So, comparing the Grid to Internet2 is a bit like comparing GNOME to the Internet...
Some good URLs on Grid computing are:
http://www.gridcomputing.com/ - long list
http://www.gridforum.org/ - various working groups, has list of other work under Related Initiatives.
An excellent in-depth book is
The Grid : Blueprint for a New Computing Infrastructure, by Ian Foster (Editor), Carl Kesselman (Editor)
ISBN: 1558604758
http://www.amazon.com/exec/obidos/ASIN/155860475 8/
The Grid will be a very impressive technology when fully realised, e.g. Grid-enabled apps could do things like real-time medical modelling and diagnosis using a bunch of remote supercomputing resources acquired dynamically on a computational market. Not to mention the gaming possibilities:)
DVDs in New Zealand will be region coded, just like any other region of the world (although there is one 'whole world' region code that is very rarely used.)
Some Chinese manufacturers are now selling DVD players with 'hidden' region hacks in their firmware - of course the codes to unlock these hacks are magically made known via the Internet. So there are probably ways of getting a region-free DVD player in NZ, quite apart from the whole PC and software hack approach.
I thought the article was quite good, but it's a shame he included sly attacks on Gnome along with mainly factual comments. I was unimpressed with the quotes from anonymous 'Gnome experts who also know KDE', and the rumours about tensions within the Gnome camp, but the points about the languages and API are worth considering.
It's debatable what makes Linux Unix-like - certainly the kernel API was designed from the start to be Unix-like, and there are Unix-like tools from BSD and the X project in particular, so it's far from accurate to say that the GNU tools are what makes it Unix-like.
I expect this will start another GNU/Linux flamefest:) It's clear that GNU has contributed massively to the typical Linux distro, I just wish they would not try to take *all* the credit...
Presumably people have looked at PICK (allows multivalued fields) and object oriented databases for this sort of thing? OODBMSs are very good at handling data with this sort of complex structure - see www.exceloncorp.com (Excelon, formerly Object Design), www.objectivity.com, and www.versant.com for some commercial examples. There are also some open source near-equivalents but I don't know the URLs - one is called Texas Persistent Store, I think.
As with RDBMSs, there are some arguments for going with commercial ODBMSs if you have very demanding requirements. The commercial ones also have lots of extra tools, many of them dealing with XML, which involves complex nested data structures and is also suited to ODBMSs.
A properly drafted NDA does not impose any restrictions on you as long as you don't disclose the confidential information. If you come by that information some other legal way, e.g. another company gives it to you, there are no restrictions. At least that's how my company's NDAs work - they are also time limited.
Lots of the most interesting jobs mean that you have to sign an NDA to have a clue what they are doing - or would you rather sign up to the job with no clue other than the skills they want from you?
Slashdot Mirror
Forgot to point out that volume replication means between two servers across some sort of network, it's not just mirroring/RAID-0.
Must...use...preview...button...
I recently investigated this for Solaris, and I'm sure some of the products will be on Linux as well. Also, check the Linux-HA page at http://www.linux-ha.org/.
The basic idea is to do clustering: two machines that share a single (virtual) IP address, in addition to their real IP address. You then have clustering software that detects a failure in the other machine, or in an application, and fails over to the other machine, starting the applications that failed, and doing a gratuitous ARP to bind the virtual IP address to the MAC address for the surviving machine.
Client apps must be able to re-connect to the server, since their TCP sessions are dropped when the primary server crashes. TCP state failover as in some firewalls (e.g. FW-1 from Check Point) would be very handy but I don't know any OSs that do it.
This requires a shared disk subsystem - initially this is usually SCSI, with two controllers and software that can handle this. As systems grow, they tend to migrate to SANs (storage area networks), usually based on Fiber Channel - this is very fast, as you might expect, and can be built using FC switches, so your SAN can be redundant, as well as your servers. You would of course need RAID 1 or 5 in your disk subsystem.
The next step is to do volume replication - this can be nearly instantaneous (you have a choice of synchronous replication, which slows down every update transaction on the main server, or asynchronous replication, which is a little less safe). The trick here is to make sure that the volume replication software can buffer updates during times when the secondary server/disk is not available - otherwise a single failure stops all transactions...
Finally, global cluster management involves failing over between geographically separated systems - this would require the client apps to know how to switch to a different IP address, though you might be able to rig something up with load balancing technology.
This is a horribly complex area, as I discovered, and it's not simple to get it right. There are many techniques I have not covered - good sites to read up on are veritas.com, logitech.com, sun.com (search for NDR and Sun Cluster), technet.oracle.com (Oracle focused but covers many options), and of course linux-ha.org.
Good point, but IMO browsers should be strict in what they send and liberal in what they send (the old IETF credo), i.e. they should be able to display something even if the WML is bad. If they really can't display anything they should show an informative error message so I can harass the WAP site or the WAP gateway as appropriate.
Since I'm using Orange in the UK, Nokia is also supplying the WAP gateway...
Anyway, as a 'plain user' I would probably have given up on WAP a long time ago due to these hassles - it's quite amazing to me that WAP works so badly for such a high percentage of pages. I happen to have a professional interest in WAP, GPRS, 3G, etc, so I keep on trying occasionally.
Ah, you mean like 'sort | uniq -c'? :)
Still, Perl is very nice when your problem doesn't neatly fit the Unix tools approach, e.g. multiple-line records or whatever.
I have a 7110 and occasionally use it for WAP - I would use it more, but about 1/3 of all pages I browse, including some portal sites, just can't be displayed. So calling this a good browser is rather an exaggeration, though it may well be better than the others you tested.
According to a (rather old) PowerPC FAQ at http://www.motorola.com/SPS/PowerPC/library/ppc_fa q/ppc_faq.html, the skinny is:
"IBM uses a custom 64-bit PowerPC processor -- the PowerPC AS -- in their AS/400 business computing systems."
Linux on S/390 is there because they want to port e-business apps to all their existing platforms, and because they understand Linux is going to commoditise operating systems. Unlike many other vendors, IBM really seems to 'get' Linux.
One useful application for S/390 Linux IMO is hosting very large numbers of dedicated-server websites on a single box - every customer gets their own virtual Linux machine, hosted on VM/390, which has 30+ years of tuning behind it.
The customer gets more power and flexibility, and the hosting provider gets more maintainable and reliable hardware, rather than having racks upon racks of Intel-based servers that will inevitably have frequent individual box failures, even though most of them are available at any one time. The only flaw in this argument is the cost of mainframe hardware and system programmers, but if you host enough websites it could be worth it.
Xunker is right, if you have a RISC-based AS/400, i.e. a modern one, it is using PowerPC under the hood.
:)
IBM has recently committed to porting Linux to the AS/400 platform (not to OS/400, the operating system) - this will be done very much the same way they already ported AIX, i.e. you'll have Linux running directly on the PowerPC hardware, with some extra software to arbitrate between OS/400 and Linux where needed, and handle Linux-OS/400 communications.
Should be very interesting when it's all done - every single IBM platform, including the most proprietary AS/400 and S/390 ranges, will be able to run Linux apps, and porting should be relatively easy providing apps are not byte order dependent etc.
Ironically, IBM tried in the 80s and 90s to provide a way of easily porting apps across all its different OSs and architectures, through a set of development tools and APIs known as System Application Architecture (SAA). They failed. It would be rather cool if Linux solved this problem for them
I don't think it's just about glibc - I use Netscape 4.75 on glibc 2.1.2, which is the latest for my distro (Mandrake). Netscape crashes so frequently I have a terminal window dedicated to restarting it - mainly when I close one window, or when I do something requiring POP3 or HTTP authentication.
Oops - you are right, i.e. most of Abilene is on IPv4 not IPv6. However, there is a 'toy backbone' of 2 core routers and 2 campus routers running IPv6, according to the Abilene IPv6 pres at http://www.ipv6forum.com/navbar/globalsummit/slide s/html/michael.lambert/sld021.htm
vBNS, the other Internet2 backbone, also has a similar 4-router configuration, though in this case the routers are all core type routers, serving Chicago, San Francisco, Maryland, etc.
Both backbone teams seem to be in 'experiment with IPv6' mode, no doubt due to the learning curve and scarcity of routers that actually support IPv6.
Hey, this was a review from an American website, why should they care about the 'international' market :)
I forgot to say that Opera on Linux is the only browser that manages to max out my puny 56K modem connection (as reported by the neat Knetload applet, which graphs throughput). [I've also tried Netscape 4.75 and Mozilla's recent nightly builds, which are horribly slow for some reason. I tried to get Galeon going but ran into .so problems.]
I think this line efficiency is because you can set it to use a large number of simultaneous connections, which is a bit mean on the website but very efficient for dialup users, whose TCP time-to-connect is quite long due to modem latency. Hopefully the server OSs will evolve to handle this sort of thing efficiently.... Of course this could be HTTP/1.1 pipelining but I doubt it as I am doing all this via Junkbuster (junkbuster.com) ad filtering, and Junkbuster requires HTTP/1.1 to be turned off (it only analyses new connections.)
Quite apart from maxing out the modem, Opera is very fast at rendering pages; it also has good in-memory caching so it *never* goes to the network or disk when you hit the Back button (unlike Netscape).
More Opera info is at www.opera.com, including a pretty usable tech preview for Linux (i.e. alpha) - crashes and has missing features, but is quite usable, in fact I'm using it now (except that I can't post to Slashdot since the HTTP login stuff is not working...). They also have versions for EPOC (Psion etc), BeOS, OS/2, etc, and hope to support non-Linux Unices as well.
Opera doesn't actually have its own JVM; however, the downloadable versions for Windows come with a JVM from Sun, in the same download and as part of the same install sequence.
As for the cost - Opera is particularly good on older hardware (my mother used it on a 486SX/25 with 8 MB RAM and Windows 3.1 for quite a while, and it was a lot better than Netscape, and I used to use it on a slow P200 laptop). So if you can pay $30 to avoid a hardware upgrade, you are way ahead financially.
Opera is missing a few features such as password management, etc, but its speed and ability to turn off images with one click is enough for me. I'll be registering the Linux version as soon as it gets out of alpha/beta, as it is stunningly fast on my AMD K6-2/350 as well.
IPv6 has been developed largely in public. Although you can use it with static IP addresses, based on the NIC MAC address, you don't have to use this - in fact you can assign whatever IPv6 address you feel like to an interface and use that instead (as long as it's routable).
IPv6 addresses can't be hard coded into NICs, because part of the address derives from the network provider and the site.
The result is that IPv6 can be just as anonymous as IPv4, with some reasonable setup, though by default it's possible your IP address will be quite static. No doubt privacy-enhancing tools will make it easy to randomly choose the lower part of your MAC address to get some privacy back, just as analogous tools block cookies etc.
If you must be so paranoid, why not at least learn about how IPv6 works before you start posting?
Actually not all of Internet2 is IPv6 - Abilene is, I think, but some of the other testing e.g. the Qbone for QoS is still on IPv4, for logistical reasons.
There's been lots of work on migration of IPv4 to IPv6 - or more correctly, coexistence, since it's quite possible IPv4 will never disappeare completely, just like DOS... The details are fairly complex, but there are various tunnelling schemes (some including automatic tunnel setup as required) as well as protocol translators that let an IPv6 domain talk to IPv4 land via (you guessed it) something like a NAT.
In time, hopefully, the IPv6 domains will get larger and larger and gateway directly to each other - the 6bone, which is an international IPv6 network, is currently a mixture of tunnels over IPv4, and some 'real' links that are native IPv6. There are even ISPs that have rolled out native IPv6 service, e.g. NTT is one that has done quite a lot in this area.
IPv6 is particularly useful to Asia and other non-US/European regions, which didn't get much IPv4 address allocation and now really need the address space. It's also important for the massive mobile Internet roll-outs that are happening over the next few years. Just as soon as Microsoft, Cisco and others start shipping IPv6 as standard (quite soon now) it will have a chance of taking over, though it will take anywhere from 5-10 years IMO.
The Grid doesn't really exist yet - it's a bundle of various technologies, including middleware, service brokers, and lots of quite specialised software, along with some standard protocols such as IPv6 et al, and no doubt some specialised ones.
5 8/
:)
Internet2, by contrast, has little to do with host software per se - it's all about engineering the network for higher and more consistent performance (using QoS), making multicast work better, and getting IPv6 working in production mode (not all at the same time, necessarily).
So, comparing the Grid to Internet2 is a bit like comparing GNOME to the Internet...
Some good URLs on Grid computing are:
http://www.gridcomputing.com/ - long list
http://www.gridforum.org/ - various working groups, has list of other work under Related Initiatives.
An excellent in-depth book is
The Grid : Blueprint for a New Computing Infrastructure, by Ian Foster (Editor), Carl Kesselman (Editor)
ISBN: 1558604758
http://www.amazon.com/exec/obidos/ASIN/15586047
The Grid will be a very impressive technology when fully realised, e.g. Grid-enabled apps could do things like real-time medical modelling and diagnosis using a bunch of remote supercomputing resources acquired dynamically on a computational market. Not to mention the gaming possibilities
I think the reason why this story made the news and not yours is that SGI's PR company is better than VA Linux's :)
DVDs in New Zealand will be region coded, just like any other region of the world (although there is one 'whole world' region code that is very rarely used.)
Some Chinese manufacturers are now selling DVD players with 'hidden' region hacks in their firmware - of course the codes to unlock these hacks are magically made known via the Internet. So there are probably ways of getting a region-free DVD player in NZ, quite apart from the whole PC and software hack approach.
Errr, Bluetooth will get you about 10 meters - hardly a replacement for the use of CSD as a carrier for WAP. Perhaps you mean GPRS?
I thought the article was quite good, but it's a shame he included sly attacks on Gnome along with mainly factual comments. I was unimpressed with the quotes from anonymous 'Gnome experts who also know KDE', and the rumours about tensions within the Gnome camp, but the points about the languages and API are worth considering.
If only Qt was LGPLed...
It's debatable what makes Linux Unix-like - certainly the kernel API was designed from the start to be Unix-like, and there are Unix-like tools from BSD and the X project in particular, so it's far from accurate to say that the GNU tools are what makes it Unix-like.
:) It's clear that GNU has contributed massively to the typical Linux distro, I just wish they would not try to take *all* the credit...
I expect this will start another GNU/Linux flamefest
Presumably people have looked at PICK (allows multivalued fields) and object oriented databases for this sort of thing? OODBMSs are very good at handling data with this sort of complex structure - see www.exceloncorp.com (Excelon, formerly Object Design), www.objectivity.com, and www.versant.com for some commercial examples. There are also some open source near-equivalents but I don't know the URLs - one is called Texas Persistent Store, I think.
As with RDBMSs, there are some arguments for going with commercial ODBMSs if you have very demanding requirements. The commercial ones also have lots of extra tools, many of them dealing with XML, which involves complex nested data structures and is also suited to ODBMSs.
> The only sites that require these features are porn sites.
Yes, cookies/http-password/ssl are never used by e-commerce sites or well-known discussion sites...
Or do you consider amazon.com and slashdot.org to be porn sites?
A properly drafted NDA does not impose any restrictions on you as long as you don't disclose the confidential information. If you come by that information some other legal way, e.g. another company gives it to you, there are no restrictions. At least that's how my company's NDAs work - they are also time limited.
Lots of the most interesting jobs mean that you have to sign an NDA to have a clue what they are doing - or would you rather sign up to the job with no clue other than the skills they want from you?
I thought this was OSF/1 based? At one time DEC called it DEC OSF/1 I think. The earlier DEC Ultrix was BSD based, though.