What Are You Doing Right Now
-
@Dashrender said:
@dafyre said:
@scottalanmiller said:
@dafyre said:
DAY 4
- All vendors see similar margin ratios. This is a well known issue. No matter how many vendors you talk to they will always push SAN-based IPODs. The process of asking vendors for IT advice is what created the IPOD in the first place as it is the cheapest way to sell a SAN which has the highest margins. They all did it and millions of companies that were doing exactly this process fell victim to this design because it is the natural result of combining sales people giving advice, no IT checks and balances and the nature of storage margins. It is also what pushed RAID 5, all based on sales margins.
I don't argue that this is generally the case. We saw this happen when we were trying to purchase our storage setup. The question that we (the IT Team) kept coming back to "What happens if SAN 1 fails?" We asked the storage vendors we were speaking to about this, and two of them were like "you replicate from SAN1 to SAN2"... We liked the HP guys because they were the first one that told us that theirs was an active/passive cluster. No down time if SAN1 fails because SAN2 would automatically take over with no down time... Sadly, we did have to test this scenario several times, and (not so sadly) it worked beautifully.
That part is great. But where the issue comes in is that no one in the chain will ever step back and say "what about not having the SAN at all?" SAN is great in the right use cases, so maybe it was right here. But as everyone in the chain was paid by selling SANs, they answers were going to always be based around the assumption that a SAN was going to be purchased.
In this case not having a SAN did not fit the business goal that we were mandated with. Keep our files up and working, even if the main server building burns down with as little down time as possible. That was our business mandate. (This happened due to a couple of issues beyond our control becoming a major panic point of the administration at the time) Backups are great to have, but at the time, that was not the business goal (we still had them of course!). That is the largest reason why we did not back away from the SAN idea.
How did SAN solve this? the building burns down and your SAN with it.
And more importantly, how did a SAN solve this in a way that a normal server would not have? SAN can do it by having replicated SANs. But you can do it with replicated local storage, DAS or NAS too. Every storage technology can handle that scenario if you replicate between buildings.
I've done this with local storage across counties before.
-
@scottalanmiller said:
@dafyre said:
In this case not having a SAN did not fit the business goal that we were mandated with. Keep our files up and working, even if the main server building burns down with as little down time as possible. That was our business mandate.
But SAN doesn't enable that in any specific way. SAN actually makes that worse and then you have to mitigate the servers risk AND the new SAN risk that was just added. You can have SAN in this setup, but the SAN can't be part of the solution. But it is standard for sales people to act so convincingly like it is that most companies believe it and buy the SAN anyway. SAN has only one purpose and that is cost savings at scale despite the added risks that it brings. SAN can be part of an HA strategy, but SAN itself lowers the availability rather than raising it. You can always do even higher availability without the SAN.
How does it not? The "SAN" I am talking about is an Active/Passive cluster so it is not an IPOD at the storage level. We already had the redundant servers, power, and network links. We didn't have to worry about "syncronizing" files across multiple servers and stuff like that since the SAN handled it for us.
In our case, the SAN fit the business goals perfectly. SAN1 went in the Main Server Room rack, and SAN2 went in the backup server room rack in another building with multiple paths back to the campus network. We already had redundant servers, so adding the SAN not only allowed us to keep files up if the main DC burnt to the ground, it allowed us to keep the campus operating. And these systems being RAID 5, and 8TB of drives, you can imagine how often each unit failed more than 1 drive at a time. However, our campus never knew that there were issues and our administration was happy that we were able to meet their mandate at a cost that didn't give the beancounters nightmares for months. (Only a week or two, ha ha).
-
@scottalanmiller said:
@dafyre said:
@scottalanmiller said:
So we had a VERY successful shopping trip today. We managed to get groceries AND the power adapters that we needed. We are SO excited. What a difference this is going to make.
Food is kinda nice... but so is being able to charge your portable stuff too, ha ha ha!
We were on four days of lentils already!
Just add sausage ?
-
@dafyre said:
How does it not? The "SAN" I am talking about is an Active/Passive cluster so it is not an IPOD at the storage level. We already had the redundant servers, power, and network links. We didn't have to worry about "syncronizing" files across multiple servers and stuff like that since the SAN handled it for us.
It doesn't do it because you have that functionality already without a SAN. What did the SAN ADD. I'm not saying that the SAN took it away, just that it didn't add it. I can do this today without a SAN. So what did the SAN provide is the question.
I'm not saying that the SAN is an IPOD, I understand how it is used here. It's HA and does the job. I'm just wondering why the SAN was selected when you could do this for free without it.
-
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
-
@dafyre said:
In our case, the SAN fit the business goals perfectly.
This is where I disagree. Is saving money not a business goal? How can it not be in a business? Even a non-profit or a government can use the money for other things. Doing more good, paying more bonusses, whatever.
Did the SAN cost money? What did it add to make up for the extra money? Did you have huge scale and so the purpose of the SAN was to lower cost at scale and HA was something that you were doing either way? SAN cannot be select for the purpose of HA. It can be selected in situations where HA is needed, but SAN does not provide HA, it just doesn't always take it away.
-
@johnhooks said:
@scottalanmiller said:
OMG... I swear the people over there are freaking incompetent... the number of people on SW that just "fear" anything virtual, hosted or remote is crazy.
https://community.spiceworks.com/topic/1374462-starting-over-do-i-need-avtive-directory
It should be worth noting that "avoiding the MS train" is not a business decision. Yes you can build a comprable suite of tools completely open source and free, but how long will that take, and how much work to maintain in that fairly small environment? Would that choice help or hurt the businessbin the long run? Prepare proposals both ways, then let the management make that decision.
It can be done in less time, less resources, and less management. It's like people think you need to compile every program that you run in Linux and then manually patch upgrades in config files by comparing with diff.
The media has left those that have never touched Linux with this impression -
-
@johnhooks said:
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
Why wow? Assuming he is going to stick fully to Windows (VMWare should be dumped as an option) instead of going with a single instance, I'd use the license to create two VMs and split the load of those options between them.
-
@scottalanmiller said:
@dafyre said:
How does it not? The "SAN" I am talking about is an Active/Passive cluster so it is not an IPOD at the storage level. We already had the redundant servers, power, and network links. We didn't have to worry about "syncronizing" files across multiple servers and stuff like that since the SAN handled it for us.
It doesn't do it because you have that functionality already without a SAN. What did the SAN ADD. I'm not saying that the SAN took it away, just that it didn't add it. I can do this today without a SAN. So what did the SAN provide is the question.
I'm not saying that the SAN is an IPOD, I understand how it is used here. It's HA and does the job. I'm just wondering why the SAN was selected when you could do this for free without it.
We could have done file synchronization, or DFS, sure. But we had some major issues in the limited time we tried DFS, and we weren't going back to that. The File Synchronization would likely have worked as well, but at the time, that was not on our radar (we likely didn't think too long about it).
What we gained, though was the ability to live migrate VMware machines from one server to the others for maintenance that would have normally taken out 2/3s of our infrastructure otherwise. (Extended power outage in the main server room for instance). This was before Shared Nothing live migration was even a term.
-
@Dashrender said:
@johnhooks said:
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
Why wow? Assuming he is going to stick fully to Windows (VMWare should be dumped as an option) instead of going with a single instance, I'd use the license to create two VMs and split the load of those options between them.
Wow because of VMWare and having all of that on the DC. He could split it, but that wasn't mentioned in that post.
-
@dafyre said:
What we gained, though was the ability to live migrate VMware machines from one server to the others for maintenance that would have normally taken out 2/3s of our infrastructure otherwise. (Extended power outage in the main server room for instance). This was before Shared Nothing live migration was even a term.
Again, though, all things that have been available without a SAN. None of these are features that SAN provides. There are literally no features that come from SANs. SANs do not add features to your infrastructure, they just add storage consolidation which adds risks and overhead but can save money at massive scale. So they are always a tradeoff specifically against the things that you are stating as the business requirements.
Yes, that you are using VMware severally limits your options but there has always been VSAN (or VSA) before that from at least three vendors specifically for VMware plus the option of fixing the bigger picture of VMware to solve these problems without spending they money on a SAN.
My guess is that VMware was purchased via the same process of going through the sales people? With Xen, KVM or Hyper-V you can do all of this for even less with more features. I realize that this is not a new project and it is hard to know exactly what the options were at the time compared to what they are now and VMware was a very good option just a few years ago unlike today. But SAN has never been an enabling technology for any of this - it's just what the sales channels push because all of the money is there.
I know that we were doing SAN-less failovers like you describe for big medical customers in Texas by 2008 for free with all local storage. It's not a new thing. SANs use local storage to do this under the hood. So it always has to exist in servers before it does in SANs.
-
@johnhooks said:
@Dashrender said:
@johnhooks said:
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
Why wow? Assuming he is going to stick fully to Windows (VMWare should be dumped as an option) instead of going with a single instance, I'd use the license to create two VMs and split the load of those options between them.
Wow because of VMWare and having all of that on the DC. He could split it, but that wasn't mentioned in that post.
Might not have licenses to split it.
-
@scottalanmiller said:
@johnhooks said:
@Dashrender said:
@johnhooks said:
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
Why wow? Assuming he is going to stick fully to Windows (VMWare should be dumped as an option) instead of going with a single instance, I'd use the license to create two VMs and split the load of those options between them.
Wow because of VMWare and having all of that on the DC. He could split it, but that wasn't mentioned in that post.
Might not have licenses to split it.
I thought it came with licenses for 2 VMs?
-
@dafyre said:
We could have done file synchronization, or DFS, sure. But we had some major issues in the limited time we tried DFS, and we weren't going back to that.
DFS would be one approach but not what I was meaning to imply. I was still assuming using the hypervisor platform for failover and getting all of the features identical to how you have them today.
-
@johnhooks said:
@scottalanmiller said:
@johnhooks said:
@Dashrender said:
@johnhooks said:
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
Why wow? Assuming he is going to stick fully to Windows (VMWare should be dumped as an option) instead of going with a single instance, I'd use the license to create two VMs and split the load of those options between them.
Wow because of VMWare and having all of that on the DC. He could split it, but that wasn't mentioned in that post.
Might not have licenses to split it.
I thought it came with licenses for 2 VMs?
Two, yes, if he has Standard. I didn't read the thread to see.
-
I am doing my daily routine of
- hoping I can find a deal for a 1-day pass to Universal Orlando
- checking the prices for Universal Orlando
- being disappointed there are no deals
What did they say about doing the same thing over and over and expecting different results?????
-
@scottalanmiller said:
@johnhooks said:
@scottalanmiller said:
@johnhooks said:
@Dashrender said:
@johnhooks said:
Buy a tower server, install a hypervisor, VMWare or HyperV, and install a server instance. Given the size of the business I reckon you can get away with a single server instance running; DHCP, DNS, AD, GP, WSUS, PS and FS. That is a lot of acronyms! Anyway, the central management of AD, GP and WSUS in particular will make your life a lot easier, no running around installing updates on every machine.
Wow.
Why wow? Assuming he is going to stick fully to Windows (VMWare should be dumped as an option) instead of going with a single instance, I'd use the license to create two VMs and split the load of those options between them.
Wow because of VMWare and having all of that on the DC. He could split it, but that wasn't mentioned in that post.
Might not have licenses to split it.
I thought it came with licenses for 2 VMs?
Two, yes, if he has Standard. I didn't read the thread to see.
Ok. Ya I don't think it was mentioned.
-
@scottalanmiller said:
@dafyre said:
We could have done file synchronization, or DFS, sure. But we had some major issues in the limited time we tried DFS, and we weren't going back to that.
DFS would be one approach but not what I was meaning to imply. I was still assuming using the hypervisor platform for failover and getting all of the features identical to how you have them today.
lol. Today they have it easy. They use Scale Computing in a new primary DC with a UPS, 25kw generator for extended outages, and multiple paths back to campus network.
Edit: The new primary DC actually was a room that already existed on campus -- including the 25kw generator!
-
@dafyre said:
lol. Today they have it easy. They use Scale Computing in a new primary DC with a UPS, 25kw generator for extended outages, and multiple paths back to campus network.
Much better approach. That Scale does this so well is a great example of how powerful the SAN-less approach is. Scale does this wonderfully.
In other news, we just learned that we are getting three more Scale nodes!! So excited.
-
@scottalanmiller said:
@dafyre said:
lol. Today they have it easy. They use Scale Computing in a new primary DC with a UPS, 25kw generator for extended outages, and multiple paths back to campus network.
Much better approach. That Scale does this so well is a great example of how powerful the SAN-less approach is. Scale does this wonderfully.
In other news, we just learned that we are getting three more Scale nodes!! So excited.
For lab or for production?