I bought a MD1200 and I hated it so much. So heavy, loud and obnoxiously bright. So I ended up building my own DAS, which is way quieter because it uses bigger fans and it has a faster, more efficient controller while being able to handle more drives. Much better
Hi Brad, I have often found in my office/lab that heat has always been a big problem. The principle reason is this area is downstairs in my home, therefore has double brick walls and also faces east. The brick walls act like a heat sponge during the morning and make it near impossible to remove heat from the air inside the room during the height of summer. At one point I had 3 42RU racks in a much larger room with most of the gear running 24/7 which included 7 or 8 HDD shelves and DL380+580 servers associated network gear. At one point this “large’ room had a high temperature of around 36 degrees in the middle of summer. Not good at all. I ended up dividing the same space, about 8 x 3.6M, in half. Although I have not finished yet, I’ve calculated I should be able to run the same gear in a smaller room, but with hot and cold aisle type set up, the room can be kept at about 28 degrees C. The hot aisle faces the hot front brick wall of my home. Still not ideal, but this is only applicable when running my whole lab flat out……which doesn’t happen often. But still allows me to use the same heat pump gear, much more efficiently. Which is just one of the things we learn as we go. Also having worked in the HVAC &Building Automation industry at one point has helped. Take care Brad.
OK so first, the brick wall is bring in heat from the outside. If you notice in my videos I have a silver covered walls. There is a reason for that,. That is to prevent the very thing you are dealing with. You need to insulate the inner wall away from the the outer head of the outside. place up ferning strips 1x3 inch or 2x4 inch wood boards up on the brick walls and then put up the insulation that has a high R factor to retard heat retention. That will help a lot with your issue,. Hope this helps some.
Thanks for making this video. Just to understand, does this mean when you use the previous SAS/sata connector cable you are halving the bandwidth to 6bbps as opposed to using this cable which is SAS to SAS providing 12Gbps?
Is it as easy as changing out the boot drive SSD with another one even if the replacement drive is not empty? Im referring to the cards that go in that J4/J3 slot that work as master/slave on ada1 and ada0
If the drive is ratted as hot swap able then yes. that easy but make sure you have your hot swap disks setup in your ZFS or Raid configs as well so the storage system knows it has a hot swap disk to use
How does your proposal differ from a mainframe? I have a feeling that once you start building a low latency topology of hyper-converged machines, that's kind of where you end up. A video comparing those concepts would be interesting from you, as you've put a lot of thought into this!
Think of it more like a main resources vs. simple cluster compute power for you have access to all the I/O inputs such as USB and so on. Build a I/O cluster in your house and all the PC systems DP are part of it, Printers, USB devices an so on are available for use with in the cluster, but this could go farther to include monitoring systems, security and camera's CNC and other lab gear... and so on. The devices come just like PNP and not DMA addresses (old school) One scanner or printewr can server the whole network and pops up in your print queue, or a comman NAS for use fully secure by being in the cluster. think of this as user cluster and not a HPC or HCP cluster for home IT lab use or more as a central home or building cluster.
Back in the days when TX-10/100 was the norm and having tx-1000 connection was the fastest but only a few ports. but today I have 100 GB switches to 10 GB switches to 1 GB nodes... So that may not be the case anymore.
There were a few clustering operating systems back in the day that allowed distributed processing without having to rewrite code. The projects died off when multi-core CPU's became ubiquitous. I have always liked the idea of spreading the load across multiple machines - I have lots of computers. Your idea of extending the processing further across GPU's NPUs etc is awesome!
@@leadiususa7394 "So can't we have a Cluster service that blows away the ProxMox clusters out of the water. Some much gets wasted on the PC or server that runs but not to its fullest." To this end though, it REALLY depends on what you do/what you're doing. I have a 3 node Proxmox HA cluster, and one of the nodes has a load average of 3.1 (out of a 4-core processor), which means that node is used, to at least 75% of its capacity (all of my nodes are the OASLOA Mini PC which has an Intel N95 Processor in it). The other nodes are sitting at around 9% and 3% CPU utilisation, but the memory utilisation is 75% and 83% respectively. So, it REALLY depends on what you're doing. My node that is using 75% of the CPU, I have an Ubuntu VM running so that I can run SPICE so that I can drive three monitors with it (because it's a LOT harder to do the same with an Ubuntu LXC container). The other two nodes run three Ubuntu LXC containers EACH, hence the low CPU utilisation, but has a high enough memory utilisation. So it REALLY depends on what you're doing.
Was watching a completely unrelated video and youtube recommend. So your in the algorithm somewhere. I will say your lab setup is intriguing. Just started to consider using proxmox in my homelab recently.
The obsolescence cycle coincides with the megapixel / lens multiplier cycle too. The image sensor resolutions didn't outpace the capacity for displaying them unscaled so many times over to meet provisioned user needs or even on-device features. They don't enable remote service delivery to the user, they supply remote data for any purpose parties using tools and applications with administrative access enable them.
My personal opinion as IT-Professional: Personal Devices are a godsend for corporate IT. You can force your users to use ONLY specified Services and connections with their devices. And we all know, your general User is very much too uneducated and simple minded to handle a whole operating system. The Money on the other hand is quickly fixed , if you consider manufacturing large quantities for your corporate directly from factories. As this is of course only possible for bigger companies, smaller ones need to rely on vendors. I remember Apple offering a specific version of their SE2 to us, because we had a limit of 450,- for a singular device. Anything above that would've needed special approval. So we got the most barebones packaged SE2's without charger and any kind of booklet.
I used a sata disk to install centos. This drive does not have raid. I then attached a nettapp x371a sas ssd 12gbs drive. The system has recognized my drive. But when I install sg3 into centos and then use the sudo sg_scan command, only the drive where centos is installed appears, the sas drive is not recognized. Therefore I cannot use the command sg_format --format --size=512 /dev/drivename. Can you help me see where I'm wrong? Or do I have to use untubu?
Setup the drive as a HBA or raid 0 as a single drive and then you should be able do it. Remember you need a true SAS controller (PCI or PCI-E) to see the drive for a motherboard will not see it.
I depends on two things, the OS, (if you are going to use a server OS software vs. desktop OS.) and the way the 2-U rack mounted SMP system is configure. This is normally loaded with a base server OS and does a special task like encoding or some task MMX functions or off site ADS-DC server / controller but rarely is it used as a personal PC but don't use Window 11 on a system like this for your Desktop OS, Use Ubuntu desktop 22.04 instead for it will let you use all the features this system offers.
I can see the CPU and memory together, they already brought the Lcache from the motherboard to the CPU chip long ago but try to stuff my 4090 with a cpu, we are not there yet but looking forward to it.
Thanks Brad. I have done similar service work in the controls and automation industry. 90% of the time dealing with PCs or servers running control systems, often coming across customers gear, varying between 5-20 years old. Requires a similar bunch of tools and equipment……spent 4 years building my toolkit, basically be prepared for anything.
These sorts of things have been happening for a long time, where things are being put closer and closer together to improve latency This will be a big step which will take lots of time to hit the market and it will be interesting to see what happens to computers to make this happen.
Yes and WiFi PCI-E card can be added, You can use any slot but know some are full height and other are half height slots. Do you research on the OS you are using to pick the best card you want.
Hi Leadius, You make great content! Do you have an email that I can contact you with? I got a 3PAR 8200 and would like some advice. I don't know if I should continue with a fiber channel setup, or sell the controller nodes and install use it as a jbod instead. Thanks.
@@leadiususa7394 I should elaborate a little. I have the 8200 2 node controllers, which provide both the fiber channel and iscsi capabilities, but I don't know what the 2 node controllers are for if the enclosures can be connected to a SAS HBA card directly. It just feels like a waste if I don't utilize the expensive node controllers. 😅 And what do you mean by JBOD on IP? Do you mean use the iscsi connector?
@@iisimaginary Well, it you get a HP Controller set you can setup a dual path dual part 4 way fail over but I choose just to do a single dual path fail over for my needs for I am not doing a enterprise level 4 way fail over that require multi vendor ISP services for BGP type usage. But go for it if you want
@@leadiususa7394 I see, so the 4 FC / 4 ISCSI controllers are for a potential 4 way fail over, what an amazing infrastructure! Then it seems like selling the 2 node controllers for a 2 port HBA card, and use the enclosure as a JBOD is the best route. What about the JBOD on IP? I thought the enclosures only had 2 DP (Mini SAS) connections, and not IP based connections?
@@iisimaginary That should be done but only at the dual fail over loop. Not 4 way but you don't need with with Min-SAS or SAS, Juast two paths is needed. .
finally found something about this, i am watching this on compaq p2 series pc which i got on my 6th birthday 18 now and surprisingly still working perfectly just put a extra 2gb ram and the hdd went bad so replaced that
I have a HPE DL380 gen9 with 8 x 2.5'' slots, there is room to add another 8, but I HATE that I have to buy a drive cage for it instead of just the backplane and cables. They could easily had the slots/grooves included in the chassis.
I bought a SAS HDD 3.5 and a sata to sas adapter the HDD wont spin up. Can someone help me what should I do to make it work on my desktop? I was thinking to get a SAS crontroller nvme from aliexpress as I have a ITX desktop..
i bought a 3.5 inch, 10TB drive for my pc second hand, and did not read the description carefully enough.. It is a SAS drive not SATA. I just want the drive to work on my PC. Can anyone reccommend a controlller card and which cables i will need, to install it in my desktop PC. I am on a very limited budget, so I don't need anything powerful or something that will run a lot of drives, I just want it to work
Sure, get a ISL SAS controller with a dual port SAS connectors on the card for internal use then get the molex power and cable set for it and you are good to go. let me know if you need links t o recommended hardware on ebay.
