Since most file servers have about 95% unused processor cycles and a limited amount of disk I/O both compression and dedupe can be significant wins provided they don't create an I/O profile that is a smaller percentage more random than their effective compression (ie if they add 10% randomness to the I/O profile but provide 30% compression then it's probably a net win). The fact that they potentially increase cache effectiveness is just gravy since cache is a few orders of magnitude faster than spinning disk and at least an order of magnitude faster than even SSD's.
It's probably heavily dependent on the content, types of disks, and number of disks. With a few spinning disks where multiple large files are typically streamed sequentially, adding 10% randomness would be a pretty severe penalty. An SSD has essentially zero negative effect from randomness though, so they would probably benefit greatly.
It's worth pointing out that many modern SSDs actually perform compression on their storage internally to increase performance.
It's fine as long as you use it properly. I use it for IIS logfiles. I want to keep the logfiles but rarely actually access them, and they are append only, and they are plain text. Very high compression at a very small loss of performance.
Compressing binary data in your working set is, as you point out, probably a bad idea, but as long as you don't do anything stupid you shouldn't have any problems.
Indeed. We use Netapps for our VMs, which have built in dedupe. But log files won't dedupe, so using compression on directories that store logs is an easy way to save space. It's also an easy way to keep from having to expand the disk size, and to keep log directories under control that can grow rapidly.
I went to a week of training recently that was hosted in my city, so I just drove instead of staying in hotels like everyone else. The office the training was located in was downtown near the top of a sky scraper, with a great view. It was a serious pain to deal with traffic and parking every day, not to mention the wireless was over loaded and the facilities were small. The companies actual main offices were located in a really nice green belt area right off a freeway, and would have been a cinch to get to. When asked, the only reason for using the skyscraper was prestige to impress the people taking the class.
This is why I love Tree Style Tabs. You get the tab bar on the left (or wherever else you like it), tabs structured hierarchically, collapsible trees and all that fancy stuff, including vertical screen estate.
I'm also quite pleased with Tree Style Tabs. When I'm researching something, I can have dozens of tabs open and organized by how I got to each one. No other browser can have as many tabs and be even remotely useful for finding them.
Not Austin. We are an evacuation destination for Houston;-). A few years ago, a hurricane caused enough panic in Austin for people to strip store shelves for supplies, which is completely retarded considering how far inland we are. Wherever you go, there are stupid people.
Austin does have its share of infrequent and sudden damaging weather, such as ice storms, hail, heavy rains, and tornadoes. That said, some minor preparation will protect you from all of those, except tornadoes. But tornadoes are so infrequent and have such localized damage that the chances of being damaged by one are small enough to completely ignore.
Austin does have an inactive fault line running through it, so if it suddenly went active again then most of the city would likely be leveled with a good 6 earthquake.
Yes, that's exactly what I'm doing -- PXE-booting the whole OS over iSCSI. I'll edit to make that clearer.
I've looked at some of the pricey solutions that do various versions of this, and what you've done for free is very impressive. I do want to confirm that I'm reading it right though, and that you're basically booting the same image, so the machine names and such will all be identical within Windows, is that correct? Have you seen that cause any issues for any games? Have you tried any LAN games like this (preferably more recent than Quake)?
Clearly having the same machine name and SSID makes it unusable for an Active Directory environment, but it doesn't sound like you're all that far off from the commercial products meant for enterprise environments.
I can think of a couple of ways around it, but each with their own disadvantages, and what you did seems simplest and clean. That said, I'd definitely prefer height adjustable rolling chairs (floor scratches and storage space be damned).
Incidentally, where does the sound come from on the stations? And how loud is it in that room, as I don't see many sound absorbing surfaces.
Don't forget Lotus Notes. IBM has some very large accounts using that.
OTOH, Groupwise is on the way out.
Yeah, IBM has some legacy customers, and a few whose C-level folks got a little too tipsy with an IBM sales person. I don't think I've met a single person in the past decade forced to use it that hasn't loathed it.
I really wish there were a real competitor to Exchange, but MS is just so far ahead of the pack on this that it's ridiculous.
While there are plenty of non-Exchange mail servers, the number of non-Exchange mailbox servers in enterprise environments (>1000 users) is pretty small. The closest I can think of are universities using hosted mail solutions like Google Apps, but I still wouldn't call that "enterprise".
The truth is that Exchange does far more than just email, and it's those USEFUL extra features that make just about every enterprise use Exchange + Outlook. I've even seen organizations split off less important users to a simple email only system, but they still run Exchange for a major chunk of the employees, and that setup is rare.
pfSense is good for office firewalls but it's severe overkill (at least in hardware) for most home uses.
Pretty much all consumer router hardware I've seen locks up at least once a month with a high number of connections and/or data. I've never seen pfSense do that. pfSense also has some pretty advanced QoS features that can do a lot to keep high bandwidth applications from interfering with other people in the home. (My wife likes to upload HD videos of the kids, and it's nice that it doesn't kill the connection for everyone else.
Like an Alix Board and run pfSense on it, with the available packages, there are likely few network related tasks you'll find that pfSense 2.0 on Alix hardware cannot handle. You can also put in whatever wireless card you want, but I prefer to run a dedicated AP. Used Cisco Aironets can be found on Ebay for under $100 and are rock solid.
I second this (or sixth, seeing the comments). You could also use any old PC laying around and it will have orders of magnitude of more power than any consumer router. I personally like old laptops and netbooks because they have built in battery backups + screens.
Nice link but no were in it does it state any of these deaths did or did not get the vaccine, or even if the strain they died of could have been prevented.
So you point to missing data, which the site readily admits is difficult to determine, and then ignore everything else?
The problem with statistics they have and always will be complete bull shit.
Some people (immune compromized) *can't* get vaccinated, so rely on the 'herd immunity'. Infants 'not yet vaccinated' rely on herd immunity. Also some people who do get immunized simply don't 'take', and thus are unknowingly still at risk.
The higher the number of non-immunized people, the higher the risk of collateral damage.
Exactly. Essentially, even if I do everything right and get my son all of his vaccinations at the proper time, there is still the chance that he is susceptible to any number of diseases. He could still be killed or disabled for the rest of his life because some idiots out there think "herd immunity" means they don't need to get vaccinated.
I'm going to go and get myself a nice cold drink because just thinking about it gets my blood boiling.
The motherboard is always the point of failure in my experience. I've never wrecked a CPU. I suppose it's possible with overheating, but that's pretty much it.
I've seen exactly one CPU die in the thousands of systems I've had to deal with over the years. I believe it was a Dell Optiplex 620. Everything else in the system had been replaced, but when we replaced the CPU the system started working. Myself and the technician they sent out were both shocked, having never seen the CPU actually be the cause. So, it can happen, it's just incredibly rare.
Yeah, cause hoisting stuff into space is cheap, right?
No, but I think the idea is to ensure that you don't lift too much mass into space. The transmitter is a relatively small size/mass. The real issue is if they can find a way to make the solar collection/panels small enough. You could create a giant solar sail to reflect light to a much smaller solar panel, but that would likely result in heat issues. If they could develop a paper thin solar panel, that is light enough, then it might be practical to ship it into space.
When you consider the military spends some $400/gallon for gas in places like Afghanistan, because it costs so much to move it in guarded convoys, things start looking up. A collection site for microwaves is basically a bunch of wires strung back and forth, which should be pretty easy to set up and much easier to repair than a large set of generators.
What is your setup and why in Spagetti Monster's name are you running VM over NFS ethernet? Either go RAID or fiber to a SAN. That does not sound like a good Enterprise setup. Are we talking 10GB over a dedicated line (doable) or 1GB over switch (WTF)? If it's the latter, I bet you have a hell of a write to disk latency problem if you run more than 1 production VM. Personally, I'd go fiber SAN if RAID wasn't an option, but then again, I run 12 VMs on 2 64bit servers alongside a small array of dedicated servers. I'd laugh an AoE based VM proposal right out of my office.
This paragraph is full of so much fail that I'm not even sure where to start.
1. VMWare recommends running VMs from an NFS mount over Ethernet. It's a simple low latency network file based protocol that allows one to directly access the VM images without mucking about with proprietary clustered file systems. We have Oracle databases, mounted over NFS to VMs, whose size put your entire storage size to shame.
2. You use RAID/fiber/SAN, but I don't think you know what they mean. RAID is a way of arranging data across multiple disks, and every SAN uses RAID. They aren't mutually exclusive. Local storage could also use RAID, but depending on your topology there may not be a need, and using local storage in most VM setups would be an unnecessary risk.
3. You said "fiber" but you may have meant Fibre Channel (FC) (note the arrangement of the R and E). If you meant FC then that's a preference and existing architecture thing. There are good arguments either way. If you meant fiber optics, then that's goofy. There is no functional difference between Ethernet over fiber and Ethernet over copper in the confines of a datacenter, they will be the same speed and protocols.
4. Modern switching adds negligible latency. It's so small that the storage itself is orders of magnitude more latency. Not using switches would be insanely inflexible.
5. Unless you are constantly maxing out your connection in bandwidth, you're unlikely to see any significant latency issues from sharing VMs on even a 1Gbps link. Assume that a VM from one server is saturating a 1Gbps link, and another server sends data. A modern switch will basically interleave the packets from the two sources. So the first packet from the second source will be delayed by the length of whatever packet is currently being sent from the first source (a miniscule amount of time). The total time for the packets after that from the second source will be about twice what it would have been with no other traffic. As the vast majority of read/writes from servers tend to be tiny, the total is negligible. If you had lots of VMs trying to transfer large amounts of data simultaneously, then it would cause serious latency issues, but in practice that essentially never happens.
6. 12 VMs on two servers should easily fit on a 1Gbps network, but most SANs will come with at least 2x 1Gbps NICs for bonding to a switch. That would ensure that all of the VMs on a host wouldn't be able to saturate the link to cause issues with the other VMs on other hosts. If you paid the insane amounts for FC on a setup this small, you paid way too much.
7. iSCSI is used in almost every organization, and AoE is a more efficient and a far simpler protocol. The only real technical advantage of iSCSI is it is IP routable, but I've never met anyone who routes between subnets. Sure you could route iSCSI over the internet, but why? In practice, because iSCSI has been around and used forever it's highly supported, whereas AoE has barely scratched the surface. That said, AoE appears to be well supported on certain platforms such as XenServer and the OP's setup.
Good luck with those web forums when your millions-of-dollars-an-hour-in-lost-revenue business is down beyond it's maintenance window while you wait for those web forum responses for your obscure edge case you ran into during $someMaintenanceTask. Even better luck with that once the response comes back 'RTFM noob!'.
Real IT checks their pride at the door and pays the man for proper mission critical support when you're dealing with enterprise infrastructure.
Heh, man you must lead a charmed life. I've seen Oracle essentially ignore the SLA times on premium support contracts for both hardware and software bugs, and not provide useful answers when they do get around to calling you back. This is at both one of the top US telecoms, and a major hardware manufacturer. I've seen them be just as stellar for smaller organizations too.
And Oracle isn't a lone example either. I've seen Microsoft and many other major players be just as unhelpful, despite paying for support contracts that could be redirected to pay for a couple of dedicated personnel on staff to support the same systems. Not that they aren't without their uses, but your post really gives them too much credit.
I'm really curious what your setup looks like. All the theory I've read behind AoE sounds great (who really thought that iSCSI needed to be routable over IP?), but it's essentially unknown in the marketplace, and I don't think I've ever met someone in person who uses it. What manufacturer, how many NICs, switches, etc? Are you using a redundant system?
If you'd been a decade earlier, you could have done it with a SCSI drive and two host controllers, all assigned to different IDs. Then you could have had both drives able to access the drive at the same time. I have no idea how you would have avoiding trashing the file system or poisoned file cache, but I'm sure there's a way.
Netmon actually does a pretty good job of dividing traffic up into different processes, which really speeds up narrowing down your capture to just the important bits. Wireshark will divine more information from your capture natively, and I frankly just like its interface more. That said, they are pretty similar with one important difference, Wireshark has to install some third party driver to support captures while Netmon installs a small Microsoft supported driver that is likely to be updated via Microsoft Updates. That means that we have no problem installing Netmon to be available on our Windows Server images. So in the end, we just use Netmon.
The specific system I used pulls ~35W idle, and measurements have put immediate electrical costs between $36-39 per year. That's a little over twice the draw of what you say a U-Verse box draws, and mine is substantially more capable and stable (I say this as a U-Verse user). Really though, this was just a random system, and you could pick up an Atom that idles 5W, or a Core2/i system that idles 30W, but that would cost you enough that it would take several years to recoup the costs via electricity. And you might want to replace it with something else in a few years anyway.
So yeah, you do have to decide if the extra $5-30/year in electrical costs, plus initial investment, is going to be worth the extra capabilities. The original poster says he will have a 100Mbps connection, and if he wants to use features like QoS, then no consumer router is going to have the necessary horsepower to push 100Mbps. He'll either have to get some business grade networking gear, or use some sort of PC.
I would re-check the numbers on that. Many routers being recommended in this discussion have clock speeds of over 600 MHz. I don't know how that compares to P3 MHz in terms of performance on router tasks, but I doubt your old P3 is orders of magnitude more powerful.
As far as little power goes, most consumer routers I've seen, including the ones I've owned, use just a few Watts for the entire system, including power supply inefficiencies. I doubt your P3 system gets that low.
So, while I don't disagree with rolling your own router so that you get all the flexibility you want, I am not convinced a P3 would be a big win in terms of processing power, and I am sure it would be a big loss in terms of electricity usage.
Granted, the comparison is something like (to use a car analogy) trying to compare the towing capacity of a large truck to a moped. The processors used in those routers are highly specialized and will actually perform a number of network related tasks at a very high rate, but unfortunately are dog slow at anything more generalized. For a simple NAT, it would probably be much faster, and would certainly be faster for basic routing. If you're using any more than the barest of features (QoS/VLANs/transparent proxy/complex logging/etc), a Pentium III with 2GB of RAM could be orders of magnitude* faster and more stable than whatever you find at Best Buy (assuming heavy workloads). It'll also be a fraction the performance of decent networking gear, but at a fraction the price. I'm assuming the poster was interested in extra features because he was asking about loading a custom firmware.
Really, I was just using the P3 as an example of something I scrounged together for zero cost that works okay. It isn't even a particularly good example. If the poster could spend a few hundred, he could get something many, many times more powerful than what I have, and consumed less power. My box pulls around 35W, which is quite a bit more than your typical consumer router, but I'm okay with eating the $40/year in electrical costs for the added ability.
Honestly, if he really wants to push a full 100Mbps AND do anything fancy, he's going to have to get a full system (or business grade networking gear). There really isn't any question about it.
* I really just like using the phrase "orders of magnitude". I could try to explain how I was using it correctly, but that would be pedantic, and I'm pretty sure it doesn't matter and no one cares.
Slashdot Mirror
Since most file servers have about 95% unused processor cycles and a limited amount of disk I/O both compression and dedupe can be significant wins provided they don't create an I/O profile that is a smaller percentage more random than their effective compression (ie if they add 10% randomness to the I/O profile but provide 30% compression then it's probably a net win). The fact that they potentially increase cache effectiveness is just gravy since cache is a few orders of magnitude faster than spinning disk and at least an order of magnitude faster than even SSD's.
It's probably heavily dependent on the content, types of disks, and number of disks. With a few spinning disks where multiple large files are typically streamed sequentially, adding 10% randomness would be a pretty severe penalty. An SSD has essentially zero negative effect from randomness though, so they would probably benefit greatly.
It's worth pointing out that many modern SSDs actually perform compression on their storage internally to increase performance.
NTFSs file compression actually rather sucks.
It's fine as long as you use it properly. I use it for IIS logfiles. I want to keep the logfiles but rarely actually access them, and they are append only, and they are plain text. Very high compression at a very small loss of performance.
Compressing binary data in your working set is, as you point out, probably a bad idea, but as long as you don't do anything stupid you shouldn't have any problems.
Indeed. We use Netapps for our VMs, which have built in dedupe. But log files won't dedupe, so using compression on directories that store logs is an easy way to save space. It's also an easy way to keep from having to expand the disk size, and to keep log directories under control that can grow rapidly.
I went to a week of training recently that was hosted in my city, so I just drove instead of staying in hotels like everyone else. The office the training was located in was downtown near the top of a sky scraper, with a great view. It was a serious pain to deal with traffic and parking every day, not to mention the wireless was over loaded and the facilities were small. The companies actual main offices were located in a really nice green belt area right off a freeway, and would have been a cinch to get to. When asked, the only reason for using the skyscraper was prestige to impress the people taking the class.
I wasn't impressed.
This is why I love Tree Style Tabs. You get the tab bar on the left (or wherever else you like it), tabs structured hierarchically, collapsible trees and all that fancy stuff, including vertical screen estate.
I'm also quite pleased with Tree Style Tabs. When I'm researching something, I can have dozens of tabs open and organized by how I got to each one. No other browser can have as many tabs and be even remotely useful for finding them.
Not Austin. We are an evacuation destination for Houston ;-). A few years ago, a hurricane caused enough panic in Austin for people to strip store shelves for supplies, which is completely retarded considering how far inland we are. Wherever you go, there are stupid people.
Austin does have its share of infrequent and sudden damaging weather, such as ice storms, hail, heavy rains, and tornadoes. That said, some minor preparation will protect you from all of those, except tornadoes. But tornadoes are so infrequent and have such localized damage that the chances of being damaged by one are small enough to completely ignore.
Austin does have an inactive fault line running through it, so if it suddenly went active again then most of the city would likely be leveled with a good 6 earthquake.
Yes, that's exactly what I'm doing -- PXE-booting the whole OS over iSCSI. I'll edit to make that clearer.
I've looked at some of the pricey solutions that do various versions of this, and what you've done for free is very impressive. I do want to confirm that I'm reading it right though, and that you're basically booting the same image, so the machine names and such will all be identical within Windows, is that correct? Have you seen that cause any issues for any games? Have you tried any LAN games like this (preferably more recent than Quake)?
Clearly having the same machine name and SSID makes it unusable for an Active Directory environment, but it doesn't sound like you're all that far off from the commercial products meant for enterprise environments.
I can think of a couple of ways around it, but each with their own disadvantages, and what you did seems simplest and clean. That said, I'd definitely prefer height adjustable rolling chairs (floor scratches and storage space be damned).
Incidentally, where does the sound come from on the stations? And how loud is it in that room, as I don't see many sound absorbing surfaces.
Don't forget Lotus Notes. IBM has some very large accounts using that.
OTOH, Groupwise is on the way out.
Yeah, IBM has some legacy customers, and a few whose C-level folks got a little too tipsy with an IBM sales person. I don't think I've met a single person in the past decade forced to use it that hasn't loathed it.
I really wish there were a real competitor to Exchange, but MS is just so far ahead of the pack on this that it's ridiculous.
There are plenty of non-exchange enterprise users
While there are plenty of non-Exchange mail servers, the number of non-Exchange mailbox servers in enterprise environments (>1000 users) is pretty small. The closest I can think of are universities using hosted mail solutions like Google Apps, but I still wouldn't call that "enterprise".
The truth is that Exchange does far more than just email, and it's those USEFUL extra features that make just about every enterprise use Exchange + Outlook. I've even seen organizations split off less important users to a simple email only system, but they still run Exchange for a major chunk of the employees, and that setup is rare.
pfSense is good for office firewalls but it's severe overkill (at least in hardware) for most home uses.
Pretty much all consumer router hardware I've seen locks up at least once a month with a high number of connections and/or data. I've never seen pfSense do that. pfSense also has some pretty advanced QoS features that can do a lot to keep high bandwidth applications from interfering with other people in the home. (My wife likes to upload HD videos of the kids, and it's nice that it doesn't kill the connection for everyone else.
Like an Alix Board and run pfSense on it, with the available packages, there are likely few network related tasks you'll find that pfSense 2.0 on Alix hardware cannot handle. You can also put in whatever wireless card you want, but I prefer to run a dedicated AP. Used Cisco Aironets can be found on Ebay for under $100 and are rock solid.
I second this (or sixth, seeing the comments). You could also use any old PC laying around and it will have orders of magnitude of more power than any consumer router. I personally like old laptops and netbooks because they have built in battery backups + screens.
Nice link but no were in it does it state any of these deaths did or did not get the vaccine, or even if the strain they died of could have been prevented.
So you point to missing data, which the site readily admits is difficult to determine, and then ignore everything else?
The problem with statistics they have and always will be complete bull shit.
What are you, afraid of numbers?
There are some people who are genuinely contraindicated for certain vaccinations. I only hope this program accounts for that.
I would be very surprised if it didn't allow one to get a doctor's note to not be vaccinated.
Some people (immune compromized) *can't* get vaccinated, so rely on the 'herd immunity'. Infants 'not yet vaccinated' rely on herd immunity. Also some people who do get immunized simply don't 'take', and thus are unknowingly still at risk.
The higher the number of non-immunized people, the higher the risk of collateral damage.
Exactly. Essentially, even if I do everything right and get my son all of his vaccinations at the proper time, there is still the chance that he is susceptible to any number of diseases. He could still be killed or disabled for the rest of his life because some idiots out there think "herd immunity" means they don't need to get vaccinated.
I'm going to go and get myself a nice cold drink because just thinking about it gets my blood boiling.
The motherboard is always the point of failure in my experience. I've never wrecked a CPU. I suppose it's possible with overheating, but that's pretty much it.
I've seen exactly one CPU die in the thousands of systems I've had to deal with over the years. I believe it was a Dell Optiplex 620. Everything else in the system had been replaced, but when we replaced the CPU the system started working. Myself and the technician they sent out were both shocked, having never seen the CPU actually be the cause. So, it can happen, it's just incredibly rare.
You are much less likely to experience work place violence from a Vet.
I'm curious if you have a citation for this, or if it's just your perspective.
Yeah, cause hoisting stuff into space is cheap, right?
No, but I think the idea is to ensure that you don't lift too much mass into space. The transmitter is a relatively small size/mass. The real issue is if they can find a way to make the solar collection/panels small enough. You could create a giant solar sail to reflect light to a much smaller solar panel, but that would likely result in heat issues. If they could develop a paper thin solar panel, that is light enough, then it might be practical to ship it into space.
When you consider the military spends some $400/gallon for gas in places like Afghanistan, because it costs so much to move it in guarded convoys, things start looking up. A collection site for microwaves is basically a bunch of wires strung back and forth, which should be pretty easy to set up and much easier to repair than a large set of generators.
What is your setup and why in Spagetti Monster's name are you running VM over NFS ethernet? Either go RAID or fiber to a SAN. That does not sound like a good Enterprise setup. Are we talking 10GB over a dedicated line (doable) or 1GB over switch (WTF)? If it's the latter, I bet you have a hell of a write to disk latency problem if you run more than 1 production VM. Personally, I'd go fiber SAN if RAID wasn't an option, but then again, I run 12 VMs on 2 64bit servers alongside a small array of dedicated servers. I'd laugh an AoE based VM proposal right out of my office.
This paragraph is full of so much fail that I'm not even sure where to start.
1. VMWare recommends running VMs from an NFS mount over Ethernet. It's a simple low latency network file based protocol that allows one to directly access the VM images without mucking about with proprietary clustered file systems. We have Oracle databases, mounted over NFS to VMs, whose size put your entire storage size to shame.
2. You use RAID/fiber/SAN, but I don't think you know what they mean. RAID is a way of arranging data across multiple disks, and every SAN uses RAID. They aren't mutually exclusive. Local storage could also use RAID, but depending on your topology there may not be a need, and using local storage in most VM setups would be an unnecessary risk.
3. You said "fiber" but you may have meant Fibre Channel (FC) (note the arrangement of the R and E). If you meant FC then that's a preference and existing architecture thing. There are good arguments either way. If you meant fiber optics, then that's goofy. There is no functional difference between Ethernet over fiber and Ethernet over copper in the confines of a datacenter, they will be the same speed and protocols.
4. Modern switching adds negligible latency. It's so small that the storage itself is orders of magnitude more latency. Not using switches would be insanely inflexible.
5. Unless you are constantly maxing out your connection in bandwidth, you're unlikely to see any significant latency issues from sharing VMs on even a 1Gbps link. Assume that a VM from one server is saturating a 1Gbps link, and another server sends data. A modern switch will basically interleave the packets from the two sources. So the first packet from the second source will be delayed by the length of whatever packet is currently being sent from the first source (a miniscule amount of time). The total time for the packets after that from the second source will be about twice what it would have been with no other traffic. As the vast majority of read/writes from servers tend to be tiny, the total is negligible. If you had lots of VMs trying to transfer large amounts of data simultaneously, then it would cause serious latency issues, but in practice that essentially never happens.
6. 12 VMs on two servers should easily fit on a 1Gbps network, but most SANs will come with at least 2x 1Gbps NICs for bonding to a switch. That would ensure that all of the VMs on a host wouldn't be able to saturate the link to cause issues with the other VMs on other hosts. If you paid the insane amounts for FC on a setup this small, you paid way too much.
7. iSCSI is used in almost every organization, and AoE is a more efficient and a far simpler protocol. The only real technical advantage of iSCSI is it is IP routable, but I've never met anyone who routes between subnets. Sure you could route iSCSI over the internet, but why? In practice, because iSCSI has been around and used forever it's highly supported, whereas AoE has barely scratched the surface. That said, AoE appears to be well supported on certain platforms such as XenServer and the OP's setup.
Good luck with those web forums when your millions-of-dollars-an-hour-in-lost-revenue business is down beyond it's maintenance window while you wait for those web forum responses for your obscure edge case you ran into during $someMaintenanceTask. Even better luck with that once the response comes back 'RTFM noob!'.
Real IT checks their pride at the door and pays the man for proper mission critical support when you're dealing with enterprise infrastructure.
Heh, man you must lead a charmed life. I've seen Oracle essentially ignore the SLA times on premium support contracts for both hardware and software bugs, and not provide useful answers when they do get around to calling you back. This is at both one of the top US telecoms, and a major hardware manufacturer. I've seen them be just as stellar for smaller organizations too.
And Oracle isn't a lone example either. I've seen Microsoft and many other major players be just as unhelpful, despite paying for support contracts that could be redirected to pay for a couple of dedicated personnel on staff to support the same systems. Not that they aren't without their uses, but your post really gives them too much credit.
I'm really curious what your setup looks like. All the theory I've read behind AoE sounds great (who really thought that iSCSI needed to be routable over IP?), but it's essentially unknown in the marketplace, and I don't think I've ever met someone in person who uses it. What manufacturer, how many NICs, switches, etc? Are you using a redundant system?
>What is your setup and why in Spagetti Monster's name are you running VM over NFS ethernet?
Maybe because VMware recommend it?
Wish I had mod points for you. Half of his post didn't make any sense.
If you'd been a decade earlier, you could have done it with a SCSI drive and two host controllers, all assigned to different IDs. Then you could have had both drives able to access the drive at the same time. I have no idea how you would have avoiding trashing the file system or poisoned file cache, but I'm sure there's a way.
Netmon actually does a pretty good job of dividing traffic up into different processes, which really speeds up narrowing down your capture to just the important bits. Wireshark will divine more information from your capture natively, and I frankly just like its interface more. That said, they are pretty similar with one important difference, Wireshark has to install some third party driver to support captures while Netmon installs a small Microsoft supported driver that is likely to be updated via Microsoft Updates. That means that we have no problem installing Netmon to be available on our Windows Server images. So in the end, we just use Netmon.
The specific system I used pulls ~35W idle, and measurements have put immediate electrical costs between $36-39 per year. That's a little over twice the draw of what you say a U-Verse box draws, and mine is substantially more capable and stable (I say this as a U-Verse user). Really though, this was just a random system, and you could pick up an Atom that idles 5W, or a Core2/i system that idles 30W, but that would cost you enough that it would take several years to recoup the costs via electricity. And you might want to replace it with something else in a few years anyway.
So yeah, you do have to decide if the extra $5-30/year in electrical costs, plus initial investment, is going to be worth the extra capabilities. The original poster says he will have a 100Mbps connection, and if he wants to use features like QoS, then no consumer router is going to have the necessary horsepower to push 100Mbps. He'll either have to get some business grade networking gear, or use some sort of PC.
I would re-check the numbers on that. Many routers being recommended in this discussion have clock speeds of over 600 MHz. I don't know how that compares to P3 MHz in terms of performance on router tasks, but I doubt your old P3 is orders of magnitude more powerful.
As far as little power goes, most consumer routers I've seen, including the ones I've owned, use just a few Watts for the entire system, including power supply inefficiencies. I doubt your P3 system gets that low.
So, while I don't disagree with rolling your own router so that you get all the flexibility you want, I am not convinced a P3 would be a big win in terms of processing power, and I am sure it would be a big loss in terms of electricity usage.
Sorry, I worded that sentence poorly, or possibly wrong, but you'll want to look at the Megahertz Myth.
http://en.wikipedia.org/wiki/Megahertz_myth
Granted, the comparison is something like (to use a car analogy) trying to compare the towing capacity of a large truck to a moped. The processors used in those routers are highly specialized and will actually perform a number of network related tasks at a very high rate, but unfortunately are dog slow at anything more generalized. For a simple NAT, it would probably be much faster, and would certainly be faster for basic routing. If you're using any more than the barest of features (QoS/VLANs/transparent proxy/complex logging/etc), a Pentium III with 2GB of RAM could be orders of magnitude* faster and more stable than whatever you find at Best Buy (assuming heavy workloads). It'll also be a fraction the performance of decent networking gear, but at a fraction the price. I'm assuming the poster was interested in extra features because he was asking about loading a custom firmware.
Really, I was just using the P3 as an example of something I scrounged together for zero cost that works okay. It isn't even a particularly good example. If the poster could spend a few hundred, he could get something many, many times more powerful than what I have, and consumed less power. My box pulls around 35W, which is quite a bit more than your typical consumer router, but I'm okay with eating the $40/year in electrical costs for the added ability.
Honestly, if he really wants to push a full 100Mbps AND do anything fancy, he's going to have to get a full system (or business grade networking gear). There really isn't any question about it.
* I really just like using the phrase "orders of magnitude". I could try to explain how I was using it correctly, but that would be pedantic, and I'm pretty sure it doesn't matter and no one cares.