Author Topic: EEVblog #691 - Dumpster Dive Xeon Servers  (Read 34146 times)

0 Members and 1 Guest are viewing this topic.

Offline dexters_lab

  • Supporter
  • ****
  • Posts: 1890
  • Country: gb
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #25 on: December 09, 2014, 10:45:08 pm »
nice score dave  :-+

is it me or is something clipping your audio, it dont sound right... or maybe you were too excited!!

Offline darko31

  • Regular Contributor
  • *
  • Posts: 103
  • Country: cs
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #26 on: December 09, 2014, 11:50:55 pm »
Dave, have you met your wacky company neighbours? If they throw away things like that, who knows what goodies they are using.
 

Offline darko31

  • Regular Contributor
  • *
  • Posts: 103
  • Country: cs
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #27 on: December 10, 2014, 12:04:36 am »
Well, now that you mention it. :D

No, I meant more of a sneak peek of the stuff they're using, meeting the IT crowd behind the scenes or something like that.
 

Offline synapsis

  • Regular Contributor
  • *
  • Posts: 140
  • Country: us
    • Blackcow
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #28 on: December 10, 2014, 12:26:56 am »
Question for the IT pros: Why is there an SD card slot directly on the motherboard of the ML330 machine?
 

Offline Monkeh

  • Super Contributor
  • ***
  • Posts: 7992
  • Country: gb
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #29 on: December 10, 2014, 12:31:39 am »
Question for the IT pros: Why is there an SD card slot directly on the motherboard of the ML330 machine?

You may choose to run an OS entirely from the SD card, so that all the drives can be dedicated to storage. It has a USB port, too.
 

Offline Howardlong

  • Super Contributor
  • ***
  • Posts: 5317
  • Country: gb
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #30 on: December 10, 2014, 12:44:21 am »
Question for the IT pros: Why is there an SD card slot directly on the motherboard of the ML330 machine?

You may choose to run an OS entirely from the SD card, so that all the drives can be dedicated to storage. It has a USB port, too.

Would that not be horrendously slow? Sorry, been out of enterprise IT for a couple of years, maybe things have changed on that front.
 

Offline Monkeh

  • Super Contributor
  • ***
  • Posts: 7992
  • Country: gb
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #31 on: December 10, 2014, 12:46:46 am »
Question for the IT pros: Why is there an SD card slot directly on the motherboard of the ML330 machine?

You may choose to run an OS entirely from the SD card, so that all the drives can be dedicated to storage. It has a USB port, too.

Would that not be horrendously slow? Sorry, been out of enterprise IT for a couple of years, maybe things have changed on that front.

Why would you need to access the SD card after boot? This is what RAM is for.
 

n45048

  • Guest
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #32 on: December 10, 2014, 03:37:21 am »
In any case using these servers to archive video that Dave hardly ever accesses is a silly suggestion. They use more power, take up more space and make more noise than plugging a USB drive in on the occassion it is needed. I'd sell them and buy a USB drive or two.

Why? Who said they need to be running all the time? They can be powered on when needed but it he did choose to run them 24/7 they can double as file/print servers for his office or act as off-site storage for his home stuff. With the added RAID capabilities, a dedicated storage server has several advantages over single USB drives.

I know RAID isn't a substitute for a good backup regime but it's a good start.
 

Offline warp_foo

  • Supporter
  • ****
  • Posts: 117
  • Country: us
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #33 on: December 10, 2014, 04:39:16 am »
Question for the IT pros: Why is there an SD card slot directly on the motherboard of the ML330 machine?

You may choose to run an OS entirely from the SD card, so that all the drives can be dedicated to storage. It has a USB port, too.

Would that not be horrendously slow? Sorry, been out of enterprise IT for a couple of years, maybe things have changed on that front.

I'm not sure the SD slot is intended to run Windows. You could, I suppose. In general, we install vmware on the SD slots (dual slots, mirrored), and then deploy guest VMs on the HD or SSD storage. (Or preferably, on NAS or SAN storage.)

m
Where are we going, and why are we in a handbasket?
 

Offline Tothwolf

  • Regular Contributor
  • *
  • Posts: 95
  • Country: 00
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #34 on: December 10, 2014, 07:04:51 am »
I'm not sure the SD slot is intended to run Windows. You could, I suppose. In general, we install vmware on the SD slots (dual slots, mirrored), and then deploy guest VMs on the HD or SSD storage. (Or preferably, on NAS or SAN storage.)

I run many servers and embedded systems with Linux on SD and CompactFlash cards. Some of them also originally came with Windows XP Embedded or Windows XP Professional on them, so yes, you can run Windows from a memory card. They can also be helpful as a recovery device, so you could store a backup copy of the OS, or a system recovery partition, etc separately from the RAID array.
 

Offline Tothwolf

  • Regular Contributor
  • *
  • Posts: 95
  • Country: 00
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #35 on: December 10, 2014, 07:37:10 am »
Dave, in addition to potential use as a fileserver, the ML330 would also be a good machine for post processing and OCR work if you ever decided to start scanning your large collection of vintage electronics magazines and books. For that fact, with two machines, assuming your internet connection in the lab would allow for it, you could use one machine for post processing and internal file storage and use the other as a publicly accessible archive. Another potential use for a smaller machine would be gps locked ntp or you could run your own internal network services such as dns, dhcp, etc. I've found an embedded machine with a Pentium M can work very well for network services on a small network though, so even the smaller of the two machines would really be overkill for that task.

HP will generally also supply restore DVDs for these machines if you contact them. The original OS might not be something you'd want to run, but the utilities and configuration software would still be worth having. Of course if you decide to sell the machines, the restore discs also add a good bit of value too.

The thing that puzzles me about the ML330 is that it has 3ea 8GB and 3ea 4GB memory modules. Usually these should be installed in matched sets so the system can use a faster interleaved memory scheme. If the modules lack an extra HP sticker on them, I suspect they might have been customer upgraded/replaced and they didn't know that they really should be installed in matched sets. Of course another explanation is that someone else beat Dave to the machines and nicked the hard drives and a couple of memory modules...
 

n45048

  • Guest
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #36 on: December 10, 2014, 07:47:40 am »
Another potential use for a smaller machine would be gps locked ntp

If you're that serious about time sync, you're better off with a NTP appliance. They come in 1RU form factors and are very low power compared to a PC. Not cheap though. It might not be possible in his office to run a GPS antenna where it has proper LOS to the satellites.

For everyday applications though, you can just connect through to a public time server provided by NIST or similar. DHCP has a user-definable value which allows you to specify the time server address to clients.
« Last Edit: December 10, 2014, 07:51:12 am by Halon »
 

Offline Tothwolf

  • Regular Contributor
  • *
  • Posts: 95
  • Country: 00
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #37 on: December 10, 2014, 09:10:58 am »
Another potential use for a smaller machine would be gps locked ntp

If you're that serious about time sync, you're better off with a NTP appliance. They come in 1RU form factors and are very low power compared to a PC. Not cheap though. It might not be possible in his office to run a GPS antenna where it has proper LOS to the satellites.

For everyday applications though, you can just connect through to a public time server provided by NIST or similar. DHCP has a user-definable value which allows you to specify the time server address to clients.

If you plan to make a time service available outside of a local network, you generally need something with more CPU than a standalone ntp appliance can offer. In addition to being quite expensive, such ntp appliances just don't have much in the way of CPU and can't handle lots of requests. Often you'd pair up one of those appliances with a larger server to create a stratum-1 or stratum-2 ntp server. I've built ntp appliances from scratch before and have a newer pair of stratum-0 gps/ntp devices in the late parts gathering stage.

There are indeed lots of public servers available, but sometimes their reliability can be questionable and they are never going to be as accurate as a local gps locked stratum-1 ntp server. The rule of thumb with ntp is if you have more than a couple of local clients, you really should also consider running your own internal stratum-3 ntp server to decrease the load on the public servers too.
 

n45048

  • Guest
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #38 on: December 10, 2014, 10:02:02 am »
Really? A 'couple' of local clients? NTP isn't that intensive of a protocol. We run GPS NTP appliances at work that serve entire offices and radio repeaters. They aren't all that powerful. I'd say one of our 1RU appliances serves a network of about 150 clients? That said the accuracy of 1 second or under is near enough.
« Last Edit: December 10, 2014, 10:03:50 am by Halon »
 

Offline VK3DRB

  • Super Contributor
  • ***
  • Posts: 2252
  • Country: au
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #39 on: December 10, 2014, 10:14:00 am »
Some experiences with Xeons...

I did a course on the Intel Xeon processor and Sitka Server design way back in the mid 1990's It was a course run by Intel at a rented hotel opposite Manly Beach in Sydney in the summer. The students were so distracted by the young bikini clad women walking up and down the sand at the beach, Intel had get in some special screens to black out the view.

After the course I developed a machine at IBM that had 4 x 220MHz Xeon processors and SCSI drives totalling a massive 9GB. I cannot remember how much RAM, but the machines were priced at $48,000 each. Value today: 5 cents as a boat anchor if you are lucky. The Sitka motherboard was quite advanced and sported an independent 8051 based processor for health monitoring that could initiate a phone call (early SMS) if there was any problem.

In California there was a machine that made an after hours call to the techo to say one Xeon has failed with an over temperature fault (Xeons in the Sitka was designed for non-stop operation with redundant processors). Short time later another call with another processor going over temperature, then another. The techo then decided to go into work to find out what was going on with the machine. He arrived find the building was in fact on fire. 




 

Offline Monkeh

  • Super Contributor
  • ***
  • Posts: 7992
  • Country: gb
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #40 on: December 10, 2014, 10:22:23 am »
Really? A 'couple' of local clients? NTP isn't that intensive of a protocol. We run GPS NTP appliances at work that serve entire offices and radio repeaters. They aren't all that powerful. I'd say one of our 1RU appliances serves a network of about 150 clients? That said the accuracy of 1 second or under is near enough.

Expose it to the web and advertise it and 'a couple' turns into 'a couple thousand' pretty quickly.

In California there was a machine that made an after hours call to the techo to say one Xeon has failed with an over temperature fault (Xeons in the Sitka was designed for non-stop operation with redundant processors). Short time later another call with another processor going over temperature, then another. The techo then decided to go into work to find out what was going on with the machine. He arrived find the building was in fact on fire.

So what you're saying is your Xeon machine set a building on fire. :P
 

Offline Tothwolf

  • Regular Contributor
  • *
  • Posts: 95
  • Country: 00
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #41 on: December 10, 2014, 12:14:12 pm »
Really? A 'couple' of local clients? NTP isn't that intensive of a protocol. We run GPS NTP appliances at work that serve entire offices and radio repeaters. They aren't all that powerful. I'd say one of our 1RU appliances serves a network of about 150 clients? That said the accuracy of 1 second or under is near enough.

 :-//

USNO NTP Network Time Servers --
"We request that you allow no more than three of your clients to poll our servers directly.  Organizations with many clients should obtain time from local stratum 2 servers."

 

Offline BrianDagobah

  • Contributor
  • Posts: 15
  • Country: us
  • I.T. Pro by day, Electronics hobbyist by night
    • Dagobah System
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #42 on: December 10, 2014, 12:16:58 pm »
Question for the IT pros: Why is there an SD card slot directly on the motherboard of the ML330 machine?

To add to what other folks have said about this. A very common config for ESXi hosts when part of a cluster, is to run the ESXi OS entirely from an SD card without local storage at all. All storage is done via iSCSI, FC, FCoE or whatever to a SAN(which multiple hosts share). This keeps the cost/power of the hosts down and when you're operating in a clustered ESXi environment, loosing a host due to a faulty SD card is really not a big deal. The FOM brings up the VMs on other hosts within a few mS. In most cases you don't even drop a single ping. On the failed host, you just reinstall ESXi (about 10 minutes worth of work), add it back to the cluster and you're back in business. Also, booting without local storage - directly to an SD card is REALLY REALLY fast.

Even in smaller environments where you've got maybe 40 ESXi hosts. Saving USD$1200*/machine on HDD's X 40 is a nice chunk of change.

*Vendor branded (HP)hot swap 600GB SAS drives are $600 each currently. CRAZY!



Brian
Tech Blog: http://www.dagobah-system.com
G+ https://plus.google.com/102681862919770239777
I.T. Pro by day, Electronics hobbyist (not a very good one) by night.
 

Offline max_torque

  • Super Contributor
  • ***
  • Posts: 1275
  • Country: gb
    • bitdynamics
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #43 on: December 10, 2014, 05:16:52 pm »
Whilst it always amazes me what companies "throw away" there are good reasons for this.  As Dave mentioned, hardware assets will have be written down (depreciated) against tax over there useful (to the company, rather than absolute) life.  As such, at the end of that period, there value is zero, even if they have a small commercial monetary value.  Unfortunately, such is the way of modern accounting, if the company were to sell these assets, the resulting income would have to be declared, and would then be taxed (as it would not match the depreciated asset value).  And of course, an accountant somewhere would have to monitor, record and finaly declare that income on the appropriate years tax submission.  And the, admittedly slightly crazy, situation is that for a large company, that process costs more than the asset is worth.  In effect, they can chuck the asset out with no recompense, or sell it, make some income, but incur costs in processing that income that equal or exceed the net worth of that asset.  End result in both cases is the same, but one action also comes with a lot of in-direct costs, such as employees time!

Mad world we live in really........   :-DD
 

Offline madires

  • Super Contributor
  • ***
  • Posts: 7754
  • Country: de
  • A qualified hobbyist ;)
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #44 on: December 10, 2014, 05:31:16 pm »
The servers aren't thown away because they are written down. They are thown away because they are old and too slow. Desktops are good for 3-4 years and servers might run 5 years. And that's reflected by the official recovery periods for PCs and servers. If none of the employees takes the old stuff home, it's dumped.
 

Offline BrianDagobah

  • Contributor
  • Posts: 15
  • Country: us
  • I.T. Pro by day, Electronics hobbyist by night
    • Dagobah System
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #45 on: December 10, 2014, 06:23:10 pm »
The servers aren't thown away because they are written down. They are thown away because they are old and too slow. Desktops are good for 3-4 years and servers might run 5 years. And that's reflected by the official recovery periods for PCs and servers. If none of the employees takes the old stuff home, it's dumped.

RIGHT! And probably end-of-life. When a product is end-of-life, the manufacturer won't support it any longer and many times parts are no longer available. Companies that rely on that equipment to be running 100% of the time, can't risk having equipment go down without being able to get parts or service in a short amount of time. Every company I've worked for chucks everything at the end of the warranty or end-of-life. It's kind of sad to see that good equipment go. Which is why I typically re-purpose it for a DEV environment or slow/cheap storage. A few years later it ends up in the recycle bin, but at least I squeezed a little life out of it.
Brian
Tech Blog: http://www.dagobah-system.com
G+ https://plus.google.com/102681862919770239777
I.T. Pro by day, Electronics hobbyist (not a very good one) by night.
 

Offline casper.bang

  • Frequent Contributor
  • **
  • Posts: 311
  • Country: dk
  • Pro SE, amateur EE.
    • BangBits
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #46 on: December 10, 2014, 08:57:20 pm »
Here in Denmark Dumpster Diving is actually illegal :)

Yeah, because there's very strict control with the sorting - but that's a small price to pay for such a high recycling percentage of 60%-70%. That aside, I wonder if Dave has a hoarding problem. For an engineer he spends an obscene amount of time and energy on old crap - I'd rather he spend his time on fundamental Friday, tutorials, reviews etc.   :-\
 

Offline Hade

  • Contributor
  • Posts: 12
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #47 on: December 11, 2014, 10:22:55 am »
Yeah, because there's very strict control with the sorting - but that's a small price to pay for such a high recycling percentage of 60%-70%.

It's illegal in the UK too, but it has nothing to do with recycling. It's just illegal to take somebodies stuff, end of.
You can take stuff from a beach however, as there is an exception for beach combing.

I'd rather he spend his time on fundamental Friday, tutorials, reviews etc.   :-\

Agreed. The video subjects have been crap recently. I'm getting bored. I love the repairs, teardowns and tutorials. I want to LEARN.
 

Offline casper.bang

  • Frequent Contributor
  • **
  • Posts: 311
  • Country: dk
  • Pro SE, amateur EE.
    • BangBits
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #48 on: December 11, 2014, 10:53:55 am »
Agreed. The video subjects have been crap recently. I'm getting bored. I love the repairs, teardowns and tutorials. I want to LEARN.

Dave is of course entirely entitled to spend his time on whatever he wants. From following him a few years now, I get the impression that he is very little impressionable (I.e. there are many many requesting an update to the power supply stuff) so the way I try to influence is donate when I like his content - been a while though. A review of the Rigol DS1000Z series feels like slam dunk at this point and has been for a while.  :-//
 

Offline VK3DRB

  • Super Contributor
  • ***
  • Posts: 2252
  • Country: au
Re: EEVblog #691 - Dumpster Dive Xeon Servers
« Reply #49 on: December 11, 2014, 11:39:03 am »
In California there was a machine that made an after hours call to the techo to say one Xeon has failed with an over temperature fault (Xeons in the Sitka was designed for non-stop operation with redundant processors). Short time later another call with another processor going over temperature, then another. The techo then decided to go into work to find out what was going on with the machine. He arrived find the building was in fact on fire.
Quote
So what you're saying is your Xeon machine set a building on fire. :P

Not stated and not implied. The Xeons and their host server effectively acted as a fire alarm. By the time the technician arrived, the entire building was on fire and outside the building was teaming with firemen and fire trucks. The machine was incinerated, but it was faithful to the end.

Interesting that "firemen" is one word and "fire trucks" are two words. Looks like another bug in the English language.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf