Server advice

[JordanT91]

I'm after a bit of advice regarding servers.

 

At present I have 3 Dell tower servers, all Hyper-V hosts:

 

• PowerEdge T420
  
  • 2 x Intel Xeon E5-2430
    
  • 64GB RAM
    
  • 6 x 600GB SAS drives (RAID 10)
    
  • Purchased March 2013
    
  • Warranty expired years ago
    

   

  [*]PowerEdge T320

  • 1 x Intel Xeon E5-2430
    
  • 64GB RAM
    
  • 4 x 600GB SAS drives
    
  • 4 x 500GB SATA drives (all in RAID 10)
    
  • Purchased January 2015
    
  • Warranty expired January 2018
    

   

  [*]PowerEdge T430

  • 2 x Intel Xeon E5-2660 v4
    
  • 64GB RAM
    
  • 4 x 400GB SSD (RAID 10)
    
  • Purchased July 2016
    
  • Warranty expires July 2021
    

 

I'm getting concerned about the T420, it's already had multiple drive failures but at nearly 7 years old I'm concerned other, more serious, failures might occur soon. It's a lot noisier than it used to be, despite being cleaned a couple of times a year.

Due to the large drive capacity on the T320, this server hosts my file server. But it's also the least powerful server of the lot.

The T430 is fine for now.

 

I've been looking at replacing the T420 with a rack server, and then any new servers in the future would also be rack-mounted. This will save a lot of space in the very cramped server room. Initially I was looking at getting one with lots of SSD storage, but then I thought what about getting a SAN and having minimal storage on each future server? Has anyone else done it this way? It seems SANs are very expensive though. As a minimum I want two servers for redundancy etc. I was also thinking I could buy some more SSDs for the T430 to match the capacity of the T320, then that would satisfy my redundancy requirements wouldn't it?

 

What do you think?

[bobsmith]

I've seen people go both ways - Personally for a typical school I think a SAN is overkill - especially if you're considering dumping more stuff on the cloud.

 

There's a lot of performance improvements I've heard about when people have gone for nice fast RAID arrays using SSDs in their virtual hosts.

 

You can get the price right down by getting refurbs from somewhere like ICT Direct - we bought 2 of their HPs not that long ago to replace 3 ageing Dells and not had any complaints about them.

 

When I worked out the economics of it, if you replace based on age of manufacture rather than date of purchase, then buying refurbs and replacing them at the end of their warranty (typically 3 years) it's still cheaper than buying new.

[CHiLL]

Our configuration is all rack mounted and consists of 2x HP DL360 (each with Xeon E5-2630 v3 and 196GB RAM) VMware hosts connected to 2x HP 2040 SANa via 2x 10GbE trunk. The MSA has 12x 600GB 15K SAS drives (split into 2x RAID5 pools) that the server OS drives live on and 12x 3TB SATA drives (also split into 2x RAID5 pools) that our data drives reside on. We've had this configuration now since August 2015 and the performance is still good with nothing for me to complain about. If I needed to spec and replace this setup, I'd go down the same route with separate hosts/storage.

[dhicks]

I've been looking at replacing the T420 with a rack server, and then any new servers in the future would also be rack-mounted. This will save a lot of space in the very cramped server room. Initially I was looking at getting one with lots of SSD storage, but then I thought what about getting a SAN and having minimal storage on each future server? Has anyone else done it this way? It seems SANs are very expensive though. As a minimum I want two servers for redundancy etc. I was also thinking I could buy some more SSDs for the T430 to match the capacity of the T320, then that would satisfy my redundancy requirements wouldn't it?

 

For a recent hardware upgrade, we went for a SuperMicro server with 96GB of RAM and an HBA storage controller and disks all certified as being suitible for use with Windows Storage Spaces. We went for a tiered storage setup, with 2 x 1TB NVMe SSDs and some larger (8TB, I think) spinning disks. This gives us a single server with plenty of processing power, RAM and disk space to fit in all the VMs and file storage we need, with all the most common file access being to/from the SSD layer (haven't got usage statistics yet as we're still moving VMs over as yet). We then set up one of our existing, older servers as a fail-over device, replicating VMs with Veeam, ready to take over in case of hardware failure. We spent about £10,000 on the main server and another £5,000 on some more storage and Veeam and so on for the older server, just for an idea of rough costs. We used VeryPC, who gave us good advice throughout the process.

[Ditto]

It's going to get raised at some point, so I might as well go now. Have you considered a cloud based approach for some of your solution. Sure, it will need good internet performance and ideally some redundancy. It's not the right solution for all, but perhaps some of the service you are providing it could offer benefits - for example not needing to repeat the task in another 5 years time and it should address some of the cramped space issue. For me what is important is not being wedded to in-house or cloud only, but to show that you have considered all options and selected the most appropriate.

[JordanT91]

It's going to get raised at some point, so I might as well go now. Have you considered a cloud based approach for some of your solution. Sure, it will need good internet performance and ideally some redundancy. It's not the right solution for all, but perhaps some of the service you are providing it could offer benefits - for example not needing to repeat the task in another 5 years time and it should address some of the cramped space issue. For me what is important is not being wedded to in-house or cloud only, but to show that you have considered all options and selected the most appropriate.

I would love to consider a cloud-based solution. It's the future. However, at present, our broadband connection is not suitable. The most we get is 30 down 6 up, and we have been told there is no chance of Openreach changing the route our phone lines take back to the cabinet (100m line of site, over 1km in reality) In addition, our connection is provided by the LA and we often have drop-outs lasting up to a few hours.

 

In short, our connection isn't suitable for cloud storage.

[PotNoodleTech]

I would probably migrate all your VMs on to one good fast refurbished server with SSD storage such as a Dell Poweredge R730XD. You can actually get buy one get one free deals on these kind of servers which means you can set up the second one as a replica server for the "just incase" scenario.

Then I would use the T430 as a Veeam Backup and Recovery server (slap in some 4TB drives).

 

 

That should give you 99.9+ % uptime for very little outlay.

 

I am a fan of simplicity = reliability which is why I 100% never advise getting a SAN.

[Jcx500]

We Built out a new hyper-v host last year , dl380 gen10, hosts all of our vms on a raid 6 array , so glad i didn't go down the san route

[JordanT91]

Okay the SAN idea is out.

 

The idea of putting all my VMs on one host, even if there is another redundant one readily available, makes me feel uneasy. It seems to be the thing to do though!

 

So lets say I purchase a new server with a large amount of RAM and SSD storage. I get rid of the T420. I move all the VMs over to the new server. I'm left with a not very powerful but loads of storage T320, and a quite-powerful T430. I suppose I could use the T320 as a backup server (I currently have Veeam on a VM) and get some SSD storage for the T430 and make that the redundant server. Just thinking out loud here.

[ultima170387]

We have 3 R740s all with local storage in a cluster using storage spaces direct. All backed up to another server with storage in a seperate location using Veeam. Simple to setup, no hardware raid. Never had any issues yet.

[hallb15]

Okay the SAN idea is out.

 

The idea of putting all my VMs on one host, even if there is another redundant one readily available, makes me feel uneasy. It seems to be the thing to do though!

 

So lets say I purchase a new server with a large amount of RAM and SSD storage. I get rid of the T420. I move all the VMs over to the new server. I'm left with a not very powerful but loads of storage T320, and a quite-powerful T430. I suppose I could use the T320 as a backup server (I currently have Veeam on a VM) and get some SSD storage for the T430 and make that the redundant server. Just thinking out loud here.

 

That sounds like a good plan!

[PyROm]

Do people go for read intensive SSD or mix use SSD for generic vm host/file server? Im looking round at dell and the prices for 4x 1.92gb SSD Mix use seems mad at £1992.00 per drive.

[chris11256]

I stuck with HDD's during out server/san refresh. I've got two HPE DL360 G10's connected via 12Gbps SAS to a HPE MSA 2050 SAN with 42TB raw storage. Each DL360 has two 10 core Xeon 4114's with 192GB ram, the hosts run about 15 vm's each including our virtual firewall and normally sit at about 80-95GB ram usage during normal running. Was deliberately specced so that one host can run everything if the other host died.

 

Almost all our production servers are vm's. The only physical servers we have are for Veeam and a domain controller. Veeam runs backups to a Synology NAS. All other servers are vm's and uptime has been almost 100% since the new servers went in during July. Only outage was caused by me accidentally not realising a power cable was loose(was connected & working), but a tiny nudge by me then power cycled the SAN(one power supply had failed, so I was waiting for the new part). So was my fault completely rather than random hardware failure.

 

Just for an idea on the costs, the hosts were £11K overall including 5 year warranty. SAN was about £12K also including 5 year warranty. Of this the actual MSA 2050 chassis was £3600, warranty £3000 and the 24 1.8TB 10K SAS drives were £230 each.

Edited January 12, 2020 by chris11256

[Guest Guest]

I went for a mix of mix use SSDs and spinning rust. User Doc's etc on the HDDs and OS' and databases etc on the SSDs

[localzuk]

The way we did it here was to have two servers as VMWare hosts, and a pair of QNAP NAS's filled with 2TB SSDs for storage. We then have a third QNAP NAS filled with 8TB HDD's elsewhere onsite for "quick recovery" (eg. if a NAS fails, if someone loses a file, or if a VM becomes corrupt). We then back up to Azure for disaster recovery.

 

Specifics would be easier with knowledge of what it is you're running (how many VMs, how many client devices etc...).

Edited January 12, 2020 by localzuk

[MS2011]

The way we did it here was to have two servers as VMWare hosts, and a pair of QNAP NAS's filled with 2TB SSDs for storage. We then have a third QNAP NAS filled with 8TB HDD's elsewhere onsite for "quick recovery" (eg. if a NAS fails, if someone loses a file, or if a VM becomes corrupt). We then back up to Azure for disaster recovery.

 

Specifics would be easier with knowledge of what it is you're running (how many VMs, how many client devices etc...).

 

Are you replicating VM’s onto azure ? How did you configured your DR ?

 

Thanks

[JordanT91]

The way we did it here was to have two servers as VMWare hosts, and a pair of QNAP NAS's filled with 2TB SSDs for storage. We then have a third QNAP NAS filled with 8TB HDD's elsewhere onsite for "quick recovery" (eg. if a NAS fails, if someone loses a file, or if a VM becomes corrupt). We then back up to Azure for disaster recovery.

 

Specifics would be easier with knowledge of what it is you're running (how many VMs, how many client devices etc...).

 

I'm running 12 VMs in total. The file server (1TB storage) is on the T320. The T430 is running a SharePoint VM. I would like to migrate to SharePoint Online in the next year or two. I'm thinking that I could purchase more SSDs for the T430, and use this alongside the T320. I'm just just concerned that the T320 isn't massively powerful, and it's getting on itself.

 

I have two NASs, one in the server room and one in a separate building. They are used for Veeam backups.

[themightymrp]

I'm looking to refresh our servers this year. Currently running 3 old Dell PowerEdge R610's with 96Gb RAM in each plus a HP P2000 SAN holding 24x 600Gb SAS HDD's. I've been here 4 years and these were already servers previously used for other things and then had a RAM upgrade to be made into a cluster.

 

We are running around 15 VM's but I can see that going up towards 20 quite easily as I have a small number of physical servers I want to get rid of. The key servers are backed up to the cloud (Redstor) and the less important ones are backed up using Veeam to a couple of 4Tb NAS's.

 

I don't know whether or not to go down the SAN route or not based on this thread? I haven't used the storage spaces feature so not too sure how well it works with regards redundancy?

 

Anybody got any thoughts or recommendations on what to go for? I could do with increasing the storage total to probably at least 20Tb I reckon.