For many years Linux was developed with nothing but a filesystem, diff, and patches. Revision control is a valuable tool, but not an absolute requirement for large scale work.
According to some Windows NT Internals book I read many years ago, it started out well-structured, with a nice hardware abstraction layer and all that. But since Windows dropped Alpha processor support, I think the abstraction started to suffer bit-rot and made things much worse than if they had no abstraction at all.
You don't know what you're talking about. Windows still runs on three separate architectures: x86-32bit, x86-64bit, and IA64 (Itanium). The hardware-specific stuff in Windows is generally limited to NTOSKERNEL.EXE, HAL.DLL, and device driver files. That's about it.
So, maybe 0.001% of the code in Windows would need to be touched to port Windows to ARM. Microsoft is not stupid. The real issue is with 3rd-arty applications and drivers - almost none of these are compiled for anything but x86-32bit. Some server applications and OS utilities have x86-64 or IA-64 versions.
Finally, supporting another hardware platform means dealing with a whole new set of device drivers, and buggy hardware, which is where most of the stability "bugs" people encounter in windows actually reside. So the market has to look fairly profitable for MSFT to jump in. The reason they support Itanium is because x86-64 wasn't quite there yet and it enabled the sale of very expensive ($25K/socket x 8+ sockets) SQL Server Enterprise edition licenses.
Yes, Linux is more secure by design than windows, but this attitude makes ppl dumb and lazy.
Linux is most definitely no more secure by design than Windows NT. It is actually far worse in many areas from a design perspective.
Linux is usually more secure as Implemented and deployed than Windows. But this has far more to do with the expertise of the sysadmins than the design of Linux. Microsoft.com seems to stay online despite running on beta versions of the MSFT suite.
An OST file is a database. One big file with lots of internal structure, like records, pages, and indexes.
Oracle, Postgresql, mssql, db2 all put terabytes of data into just a few files. This setup is a lot more efficient than dealing with millions of small files on just about any filesystem.
Why would I choose MS's virtualization engine when Red Hat and Novell have ones that are more mature?
Because Hyper-V, in BETA, was better than any KVM-based virtualization solution. Seriously.
VMware ESX is the competition for Hyper-V. Xen is the enemy of Hyper-V's enemy, and so they share an open-standard disk format and management API with Hyper-V.
Getting KVM going with clustered storage, live migration, virtual VLANs, etc. costs about a man-year of staff time. And the solution will be very, very brittle.
VMware can provide way more features, performance, and stability with a few hours of work. Xen and Microsoft are behind, but still way better than anything based on KVM.
We query inindexed, multi-terabyte tables in Microsoft SQL Server all the time. They DBs don't go down or offline. In MSSQL, use WITH NOLOCK if you can afford dirty reads (you usually can for analytics). I am surprised Oracle has no equivalent hint that would allow a long-running query to not generate large transaction state. Even if you turn on MVCC in MS SQL Server, you can still specify locking behavior for individual statements.
We're a Rackspace customer in their DFW datacenter. This is the third power-related outage they've had in the last two years at that supposedly world-class facility.
The first wasn't really their fault: truck driver with health condition runs into their transformers. Generators kick in, but chillers don't re-start quickly enough. Temps skyrocket in minutes, emergency shutdowns. Maybe the transformes should have had some $50 concrete pylons surrounding them?
The second outage was the result of a botched generator upgrade.
This latest outage was the result of a botched UPS maintenance.
None of the outages was long enough to trigger our failover policy to our DR site, but our customers definitely noticed.
While their messaging has been very open and honest about the problems, and the SLA credits have been immediate, we pay them nearly $20K per month. Nedless to say, we are shopping, and looking into a "multiple cheap colos" architecture instead of "Tier-1 managed hosting". Nothing beats geographic redundancy.
Apple does not have a monopoly on music jukebox software.
Of course they do. About a dozen Zunes and other "non-iPod music players" have been actually sold in all of history.
Apple has a comparable percentage of "digital music player/content" sales as Microsoft has of the desktop operating system market. The iPod/iPhone + iTunes is certainly a "monopoly" if you consider Windows a mononoply. I would argue the iPod is even more of a monopoly, with lots of DRM, hardware/software tying, and intentional interoperability issues.
Can you imagine if Microsoft sold you your PC hardware, locked it so it would only run Windows, and the only place you could acquire additional software for it was in the approved Microsoft online store? Welcome to Apple's vision of the future.
HTTPS, WebDAV, and CIFS are all built into on Windows, and enable secured transfers. Or you can run insecure FTP through a PPTP, IPsec, or STTP tunnel. These require only the software already present on Windows boxes. Which is why there is not massive demand for SFTP on Windows (although I use personally it heavily for transfers between Windows and UNIX-ish systems).
There are good SFTP clients for windows, many open source. There are just no good open source SSH server solutions on WIndows (I've tried them all and stuck with inexpensive commerical offerings when I needed one).
My, aren't you lazy! First link on Google when searching for "windows SSH": WinSSHD.
As for "native" SSH for Windows, there you have it. No Cygwin at least.
As for "native open source" SSH on Windows, there obviously isn't great demand. Alternative solutions for the same problems have been available longer on Windows: 1) Telnet with NTLM or Kerberos authentication 2) RDP 3) PSEXEC. RDP has native encryption, any of the others can be wrapped in the natively provided Windows VPN/tunneling solutions (PPTP,IPsec,L2TP,SSTP).
Executable white-listing has been built into Windows Group Policies since 2001. If you have a windows server domain, XP and later clients can be made to run only executables signed with certain certificates, and/or executables with specific hashes. We have used this for kiosk machines at trade shows.
The problem is that maintaining the database of "allowed executables" is horrifyingly complicated. Every patch, DLL, printer driver, language variant, Flash plugin, Adobe Reader, etc. needs to be in the database. There are some 3rd party management tools for that, I think, but we never even got that far before we gave up. It was simply too much to manage even in a 200-workstation network.
Bandwidth isn't the issue for most use cases. It's latency that matters. And this the area is where SSDs crush spinning platters - random access latency is 20-100 times better on SSDs.
This fascination the IT industry has with bandwidth numbers is often meaningless penis-measuring by the under-informed. Latency has a far greater impact than bandwidth for the majority of computer subsystems. Reducing latency inside the processor, cache, main memory, storage, or network usually provides a much bigger benefit than increasing bandwidth.
I can give you a 1 GBps internet connection with 250 ms latency, and it will seem slow to the web-browsing user compared to a 3 Mbps DSL line with 15 ms latency. (And yes, I know about TCP window tuning, but that doesn't help for browsing.)
So iSCSI sharing from two simultaneously active nodes sharing the same mirrored image is certainly possible, you'd just have to switch to the other iSCSI target. How does LeftHand manage consistency? For example, if power gets lost on a node and writes get redirected, how does it ensure there were no in-transit io's going through the dead node? Or does it take the latency hit and not return IO complete until the write is committed on the other nodes?
Well, that is at least part of their special sauce I suppose. I assume they take the latency hit to a degree. Our testing showed spindle-level IOPS throughput overall at longer queue depths, but higher latency than other more traditional systems we demoed running IOmeter. We also yanked cords, power, and jiggled a known-bad network cable during heavy write tests and never observed missing or corrupt data. Some of that may have been covered up by retries on the client side though (haven't read the iSCSI specs to see what is requried of iSCSI clients in that area).
I thought DRBD was only active/active if you ran different services (volumes, NFS shares, whatever) on each of the two active nodes. I don't think you can share the same iSCSI volume from both nodes in a DRBD cluster, can you? That's pretty inefficent and hackish, becasue you have to decide which node is primary for a particular volume; reconfiguration would be manual and likely require downtime.
In any case, the LeftHand gig is N-way plus M redundancy. So you get to use the aggregate network bandwidth and disk IOPS of all of the cluster nodes at the same time for each volume. iSCSI initiators that support miltple connections per session don't even have to deal with sending everything to a "virtual IP"; they can send direct to the node that contains the interesting blocks. Note that VMware doesn't do that bit just yet, but they have iSCSI plug-ins for Windows (and I think Linux) that do. iSCSI HBAs can also do that I believe.
One high-end SAN feature LeftHand doesn't have is IO proritization per volume... they expect you to handle that from the initiator side, which doesn't work when the initiators are different machines. And support has gone downhill since the HP takeover - I need like 3 separate HP website acocunts just to download patches.
DRBD + LVM or another roll-it-yourself solution wouldn't compare from a High-Availability standpoint; that would be an active/passive solution. LeftHand gear (now owned by HP) is a true N+1 or N+2 cluster. All nodes are active, and data is striped amongst all nodes with one or two copies. Lose a node, and iSCSI redirects keep everything working without any interruption at all. We actually use a cluster of five of their hardware appliances (basically HP DL320s with 12 disks each running the software on the metal) for our VMware installation, and are testing the Virtual SAN Appliance for our DR site.
Other "real SAN" features include very space-efficent snapshots without reservations (LVM is pretty useless for snapshots because of the reservations), volume clones with shared blocks, asynchronous replication with bandwidth controls, thin provisioning, grow/shrink, etc. If they could figure out a way to do truly clustered NFS/CIFS with deduplication on the same platform, they could kill NetApp in a year or two. HP bought LeftHand as a hedge against Dell/EqualLogic, just in case the "clustered storage" thing took off. It might still.
It would be nice to see some sort of virtual SAN integrated into the VMs
Something like this you mean? Turns the local storage on any VMware host into part of a full-featured, clustered, iSCSI SAN. Not cheap though (about $2500 per TB)
Take an economics class, please. Even the most leftist college econ professor would agree that wealth is created every day. The economoy is not a zero-sum game.
Example: we have a stack of wood and a pile of nails, and you with some free time. You decide to build a house out of the wood and nails. Now, everyone knows a finished house has 10x the real value of a stack of wood and nails. Your labor and ingenuity created that wealth.
Now, you might ask, how do we measure "value". A rational person would say value is "what can I sell it for?" or "what is it worth to me?". But even fuzzy non-sensical things like "value to society" are still increased in this example, as our little house can just be a homeless shelter for displaced auto workers.
Firefox has the same behavior after every update... it always opens a mozilla.org page without your consent. This is an informative page "you've just updated, here are some notes/preferences/whatever" in both the IE and Firefox cases. No reason to get your britches in a twist.
Or you could just use a GUID-generating algorithm that produces identifiers which are monotonically increasing (yes, it is possible) - MSSQL even has a special function for that, NEWSEQUENTIALID(), and of course the same thing can be done on client side as well.
Using a NEWSEQUENTIALID() sort of function, even on the client, creates a bottleneck: multiple threads must coordinate with some central location that stores the last updated value. Global shared state doesn't scale, and it makes code ugly. Using a per-server or per-client prefix value like a hashed MAC address as the part of the key also creates (to a lesser degree) the same problem, in that inserts do not always go to the end of a table.
Of course, for MSSQL you can do away with generated keys completely, and use GUIDs, which can be generated on the client in the first place. But that makes indexes that much bigger, and queries that much slower...
We were worried about that once upon a time, but it turns out to have been a case of preemtime optimisation. We benchmarked one of our prototype apps with both INTEGER and UNIQUEIDENTIFIER primary keys. There was no real difference in performance (likely because it was operations on the other columns that actually took up all the time, not index seeks).
The real thing to watch out for with GUIDs as primary keys is index fragmentation... inserts are spread randomly throughout the table if you cluster on the GUID. This, in turn, causes your DB to use about twice as much space as it would with sequential keys unless you rebuild clustered indexes frequently. Our solution was a compromise: use GUIDs as primary keys (and get all the of the attendant advantages, especially multi-master replication). We also create a clustered index on some other column in the table that makes more sense for the application, such as a date/time value, or an appropriate string value that gets searched a lot.
In my experience an MS-SQL server could NOT keep up with heavy hits for a corporate database. We had to replicate the MS-SQL tables to MySQL in order not to have to restart a server every 30 minutes.
I'd say you really don't know what you're doing. SQL 7.0 had some odd locking behavior, but that is over 10 years old. We have one SQL 2005 application that supports hundreds of thousands of users and 1M+ transactions per day. On a single 8-core box with 16 GB of RAM (with warm standby in case of hardware failure). SQL gets restarted about once every six months for a security patch. And it requires almost no tuning, at least compared to Oracle.
SQL Server is clearly Microsoft's best product, and probably the best general SQL database available in terms of feature/performance/usability balance.
The whining tone of the "fact sheet" makes me giggle. Microsoft "must" do this, they "must" do that. Simply because someone published something and called it a "standard". I especially like the "Microsoft must support the revised ODF standard in the first Office release or Office service pack after a revised standard is published"... totally unrealistic. Rapid implementation of standards has never been accomplished in any part of the IT industry. Hell, support for SQL-92 is still spotty, much less SQL:2006. HTML 4.01 is still not fully implemented in any browser. And don't get me started on SOAP.
Hey, Microsoft published OOXML as a standard, also through an established standards organization. Are OpenOffice.org, Google Docs, etc going to be "required" to support OOXML in their next releases? If no, why not? Because Microsoft is a big corporation with gobs of money? So are IBM, Google, and Oracle.
but they are still using most of the Kernel of Windows NT - The main change is the Network system?
They are not using the NT kernel. Almost all of Singularity is written in managed code (a C# derviative), only a very small core is unmanaged code (all of the kernel of any other mainstream OS, including NT and Linux, is typically unmanaged C/C++ code or assembler). That's the whole point really: Singularity is a totally different sort of OS which uses software-based contracts to handle process isolation and security. The goal is to build an OS that can make guarantees about the behavior of the software which it runs. Correctness and reliabiity are the primary design priciples, but despite this performance doesn't seem too bad based on their benchmark results.
Slashdot Mirror
For many years Linux was developed with nothing but a filesystem, diff, and patches. Revision control is a valuable tool, but not an absolute requirement for large scale work.
Put the bong down, right now. You're not even making sense to the /. crowd.
According to some Windows NT Internals book I read many years ago, it started out well-structured, with a nice hardware abstraction layer and all that. But since Windows dropped Alpha processor support, I think the abstraction started to suffer bit-rot and made things much worse than if they had no abstraction at all.
You don't know what you're talking about. Windows still runs on three separate architectures: x86-32bit, x86-64bit, and IA64 (Itanium). The hardware-specific stuff in Windows is generally limited to NTOSKERNEL.EXE, HAL.DLL, and device driver files. That's about it.
So, maybe 0.001% of the code in Windows would need to be touched to port Windows to ARM. Microsoft is not stupid. The real issue is with 3rd-arty applications and drivers - almost none of these are compiled for anything but x86-32bit. Some server applications and OS utilities have x86-64 or IA-64 versions.
Finally, supporting another hardware platform means dealing with a whole new set of device drivers, and buggy hardware, which is where most of the stability "bugs" people encounter in windows actually reside. So the market has to look fairly profitable for MSFT to jump in. The reason they support Itanium is because x86-64 wasn't quite there yet and it enabled the sale of very expensive ($25K/socket x 8+ sockets) SQL Server Enterprise edition licenses.
Yes, Linux is more secure by design than windows, but this attitude makes ppl dumb and lazy.
Linux is most definitely no more secure by design than Windows NT. It is actually far worse in many areas from a design perspective.
Linux is usually more secure as Implemented and deployed than Windows. But this has far more to do with the expertise of the sysadmins than the design of Linux. Microsoft.com seems to stay online despite running on beta versions of the MSFT suite.
An OST file is a database. One big file with lots of internal structure, like records, pages, and indexes.
Oracle, Postgresql, mssql, db2 all put terabytes of data into just a few files. This setup is a lot more efficient than dealing with millions of small files on just about any filesystem.
Why would I choose MS's virtualization engine when Red Hat and Novell have ones that are more mature?
Because Hyper-V, in BETA, was better than any KVM-based virtualization solution. Seriously.
VMware ESX is the competition for Hyper-V. Xen is the enemy of Hyper-V's enemy, and so they share an open-standard disk format and management API with Hyper-V.
Getting KVM going with clustered storage, live migration, virtual VLANs, etc. costs about a man-year of staff time. And the solution will be very, very brittle.
VMware can provide way more features, performance, and stability with a few hours of work. Xen and Microsoft are behind, but still way better than anything based on KVM.
We query inindexed, multi-terabyte tables in Microsoft SQL Server all the time. They DBs don't go down or offline. In MSSQL, use WITH NOLOCK if you can afford dirty reads (you usually can for analytics). I am surprised Oracle has no equivalent hint that would allow a long-running query to not generate large transaction state. Even if you turn on MVCC in MS SQL Server, you can still specify locking behavior for individual statements.
We're a Rackspace customer in their DFW datacenter. This is the third power-related outage they've had in the last two years at that supposedly world-class facility.
The first wasn't really their fault: truck driver with health condition runs into their transformers. Generators kick in, but chillers don't re-start quickly enough. Temps skyrocket in minutes, emergency shutdowns. Maybe the transformes should have had some $50 concrete pylons surrounding them?
The second outage was the result of a botched generator upgrade.
This latest outage was the result of a botched UPS maintenance.
None of the outages was long enough to trigger our failover policy to our DR site, but our customers definitely noticed.
While their messaging has been very open and honest about the problems, and the SLA credits have been immediate, we pay them nearly $20K per month. Nedless to say, we are shopping, and looking into a "multiple cheap colos" architecture instead of "Tier-1 managed hosting". Nothing beats geographic redundancy.
Apple does not have a monopoly on music jukebox software.
Of course they do. About a dozen Zunes and other "non-iPod music players" have been actually sold in all of history.
Apple has a comparable percentage of "digital music player/content" sales as Microsoft has of the desktop operating system market. The iPod/iPhone + iTunes is certainly a "monopoly" if you consider Windows a mononoply. I would argue the iPod is even more of a monopoly, with lots of DRM, hardware/software tying, and intentional interoperability issues.
Can you imagine if Microsoft sold you your PC hardware, locked it so it would only run Windows, and the only place you could acquire additional software for it was in the approved Microsoft online store? Welcome to Apple's vision of the future.
HTTPS, WebDAV, and CIFS are all built into on Windows, and enable secured transfers. Or you can run insecure FTP through a PPTP, IPsec, or STTP tunnel. These require only the software already present on Windows boxes. Which is why there is not massive demand for SFTP on Windows (although I use personally it heavily for transfers between Windows and UNIX-ish systems).
There are good SFTP clients for windows, many open source. There are just no good open source SSH server solutions on WIndows (I've tried them all and stuck with inexpensive commerical offerings when I needed one).
My, aren't you lazy! First link on Google when searching for "windows SSH": WinSSHD.
As for "native" SSH for Windows, there you have it. No Cygwin at least.
As for "native open source" SSH on Windows, there obviously isn't great demand. Alternative solutions for the same problems have been available longer on Windows: 1) Telnet with NTLM or Kerberos authentication 2) RDP 3) PSEXEC. RDP has native encryption, any of the others can be wrapped in the natively provided Windows VPN/tunneling solutions (PPTP,IPsec,L2TP,SSTP).
Executable white-listing has been built into Windows Group Policies since 2001. If you have a windows server domain, XP and later clients can be made to run only executables signed with certain certificates, and/or executables with specific hashes. We have used this for kiosk machines at trade shows.
The problem is that maintaining the database of "allowed executables" is horrifyingly complicated. Every patch, DLL, printer driver, language variant, Flash plugin, Adobe Reader, etc. needs to be in the database. There are some 3rd party management tools for that, I think, but we never even got that far before we gave up. It was simply too much to manage even in a 200-workstation network.
Bandwidth isn't the issue for most use cases. It's latency that matters. And this the area is where SSDs crush spinning platters - random access latency is 20-100 times better on SSDs.
This fascination the IT industry has with bandwidth numbers is often meaningless penis-measuring by the under-informed. Latency has a far greater impact than bandwidth for the majority of computer subsystems. Reducing latency inside the processor, cache, main memory, storage, or network usually provides a much bigger benefit than increasing bandwidth.
I can give you a 1 GBps internet connection with 250 ms latency, and it will seem slow to the web-browsing user compared to a 3 Mbps DSL line with 15 ms latency. (And yes, I know about TCP window tuning, but that doesn't help for browsing.)
So iSCSI sharing from two simultaneously active nodes sharing the same mirrored image is certainly possible, you'd just have to switch to the other iSCSI target. How does LeftHand manage consistency? For example, if power gets lost on a node and writes get redirected, how does it ensure there were no in-transit io's going through the dead node? Or does it take the latency hit and not return IO complete until the write is committed on the other nodes?
Well, that is at least part of their special sauce I suppose. I assume they take the latency hit to a degree. Our testing showed spindle-level IOPS throughput overall at longer queue depths, but higher latency than other more traditional systems we demoed running IOmeter. We also yanked cords, power, and jiggled a known-bad network cable during heavy write tests and never observed missing or corrupt data. Some of that may have been covered up by retries on the client side though (haven't read the iSCSI specs to see what is requried of iSCSI clients in that area).
I thought DRBD was only active/active if you ran different services (volumes, NFS shares, whatever) on each of the two active nodes. I don't think you can share the same iSCSI volume from both nodes in a DRBD cluster, can you? That's pretty inefficent and hackish, becasue you have to decide which node is primary for a particular volume; reconfiguration would be manual and likely require downtime.
In any case, the LeftHand gig is N-way plus M redundancy. So you get to use the aggregate network bandwidth and disk IOPS of all of the cluster nodes at the same time for each volume. iSCSI initiators that support miltple connections per session don't even have to deal with sending everything to a "virtual IP"; they can send direct to the node that contains the interesting blocks. Note that VMware doesn't do that bit just yet, but they have iSCSI plug-ins for Windows (and I think Linux) that do. iSCSI HBAs can also do that I believe.
One high-end SAN feature LeftHand doesn't have is IO proritization per volume... they expect you to handle that from the initiator side, which doesn't work when the initiators are different machines. And support has gone downhill since the HP takeover - I need like 3 separate HP website acocunts just to download patches.
DRBD + LVM or another roll-it-yourself solution wouldn't compare from a High-Availability standpoint; that would be an active/passive solution. LeftHand gear (now owned by HP) is a true N+1 or N+2 cluster. All nodes are active, and data is striped amongst all nodes with one or two copies. Lose a node, and iSCSI redirects keep everything working without any interruption at all. We actually use a cluster of five of their hardware appliances (basically HP DL320s with 12 disks each running the software on the metal) for our VMware installation, and are testing the Virtual SAN Appliance for our DR site.
Other "real SAN" features include very space-efficent snapshots without reservations (LVM is pretty useless for snapshots because of the reservations), volume clones with shared blocks, asynchronous replication with bandwidth controls, thin provisioning, grow/shrink, etc. If they could figure out a way to do truly clustered NFS/CIFS with deduplication on the same platform, they could kill NetApp in a year or two. HP bought LeftHand as a hedge against Dell/EqualLogic, just in case the "clustered storage" thing took off. It might still.
It would be nice to see some sort of virtual SAN integrated into the VMs
Something like this you mean? Turns the local storage on any VMware host into part of a full-featured, clustered, iSCSI SAN. Not cheap though (about $2500 per TB)
Take an economics class, please. Even the most leftist college econ professor would agree that wealth is created every day. The economoy is not a zero-sum game.
Example: we have a stack of wood and a pile of nails, and you with some free time. You decide to build a house out of the wood and nails. Now, everyone knows a finished house has 10x the real value of a stack of wood and nails. Your labor and ingenuity created that wealth.
Now, you might ask, how do we measure "value". A rational person would say value is "what can I sell it for?" or "what is it worth to me?". But even fuzzy non-sensical things like "value to society" are still increased in this example, as our little house can just be a homeless shelter for displaced auto workers.
Firefox has the same behavior after every update... it always opens a mozilla.org page without your consent. This is an informative page "you've just updated, here are some notes/preferences/whatever" in both the IE and Firefox cases. No reason to get your britches in a twist.
Or you could just use a GUID-generating algorithm that produces identifiers which are monotonically increasing (yes, it is possible) - MSSQL even has a special function for that, NEWSEQUENTIALID(), and of course the same thing can be done on client side as well.
Using a NEWSEQUENTIALID() sort of function, even on the client, creates a bottleneck: multiple threads must coordinate with some central location that stores the last updated value. Global shared state doesn't scale, and it makes code ugly. Using a per-server or per-client prefix value like a hashed MAC address as the part of the key also creates (to a lesser degree) the same problem, in that inserts do not always go to the end of a table.
Of course, for MSSQL you can do away with generated keys completely, and use GUIDs, which can be generated on the client in the first place. But that makes indexes that much bigger, and queries that much slower...
We were worried about that once upon a time, but it turns out to have been a case of preemtime optimisation. We benchmarked one of our prototype apps with both INTEGER and UNIQUEIDENTIFIER primary keys. There was no real difference in performance (likely because it was operations on the other columns that actually took up all the time, not index seeks).
The real thing to watch out for with GUIDs as primary keys is index fragmentation... inserts are spread randomly throughout the table if you cluster on the GUID. This, in turn, causes your DB to use about twice as much space as it would with sequential keys unless you rebuild clustered indexes frequently. Our solution was a compromise: use GUIDs as primary keys (and get all the of the attendant advantages, especially multi-master replication). We also create a clustered index on some other column in the table that makes more sense for the application, such as a date/time value, or an appropriate string value that gets searched a lot.
In my experience an MS-SQL server could NOT keep up with heavy hits for a corporate database. We had to replicate the MS-SQL tables to MySQL in order not to have to restart a server every 30 minutes.
I'd say you really don't know what you're doing. SQL 7.0 had some odd locking behavior, but that is over 10 years old. We have one SQL 2005 application that supports hundreds of thousands of users and 1M+ transactions per day. On a single 8-core box with 16 GB of RAM (with warm standby in case of hardware failure). SQL gets restarted about once every six months for a security patch. And it requires almost no tuning, at least compared to Oracle.
SQL Server is clearly Microsoft's best product, and probably the best general SQL database available in terms of feature/performance/usability balance.
The whining tone of the "fact sheet" makes me giggle. Microsoft "must" do this, they "must" do that. Simply because someone published something and called it a "standard". I especially like the "Microsoft must support the revised ODF standard in the first Office release or Office service pack after a revised standard is published"... totally unrealistic. Rapid implementation of standards has never been accomplished in any part of the IT industry. Hell, support for SQL-92 is still spotty, much less SQL:2006. HTML 4.01 is still not fully implemented in any browser. And don't get me started on SOAP.
Hey, Microsoft published OOXML as a standard, also through an established standards organization. Are OpenOffice.org, Google Docs, etc going to be "required" to support OOXML in their next releases? If no, why not? Because Microsoft is a big corporation with gobs of money? So are IBM, Google, and Oracle.
whenever there is an update to IE MS seems fit to switch my preference *back* to live.com
Ah, another complete and utter falsehood about Windows, brought to you by Slashdot. My day is complete.
but they are still using most of the Kernel of Windows NT - The main change is the Network system?
They are not using the NT kernel. Almost all of Singularity is written in managed code (a C# derviative), only a very small core is unmanaged code (all of the kernel of any other mainstream OS, including NT and Linux, is typically unmanaged C/C++ code or assembler). That's the whole point really: Singularity is a totally different sort of OS which uses software-based contracts to handle process isolation and security. The goal is to build an OS that can make guarantees about the behavior of the software which it runs. Correctness and reliabiity are the primary design priciples, but despite this performance doesn't seem too bad based on their benchmark results.