Understanding Hyper-Converged...

[CyBeRkId2002]

Hi all,

 

We are currently beginning work on a server refresh for our current ~7ish year old hardware. Currently we are running a 3 Node VMWare solution with EMC VNXe SAN.

 

What we are looking for is a simplified, easy to manage but most importantly reliable solution which has led me to look at 2 Host Hyperconverged solutions. I am happy with Host/SAN configs but less so with converged solutions. I have a couple of quick questions to check my understanding is correct:

 

• A converged solution is one where a Hypervisor sits on the same physical hardware as the storage
  
• Storage is pooled across multiple servers and is resilient in the event of both single server failure (Hypervisor failover) or drive failure (RAID).
  
• In S2D a two host solution results in 25% data storage efficiency with 2 host nested resiliency.
  
• If we go down the Server2019 / S2D route specific hardware is required. Is there an official microsoft compatibility list available?
  
• Other options for Hyperconverged worth investigating are Dell HPE, Starwind - Happy to hear others
  

 

Is there anything else major I have missed. At the same time we are looking at switching from VMWare to Hyper-V. Are there any considerations surrounding hardware solution to support this?

[jslate1980]

Hi all,

 

We are currently beginning work on a server refresh for our current ~7ish year old hardware. Currently we are running a 3 Node VMWare solution with EMC VNXe SAN.

 

What we are looking for is a simplified, easy to manage but most importantly reliable solution which has led me to look at 2 Host Hyperconverged solutions. I am happy with Host/SAN configs but less so with converged solutions. I have a couple of quick questions to check my understanding is correct:

 

• A converged solution is one where a Hypervisor sits on the same physical hardware as the storage
  
• Storage is pooled across multiple servers and is resilient in the event of both single server failure (Hypervisor failover) or drive failure (RAID).
  
• In S2D a two host solution results in 25% data storage efficiency with 2 host nested resiliency.
  
• If we go down the Server2019 / S2D route specific hardware is required. Is there an official microsoft compatibility list available?
  
• Other options for Hyperconverged worth investigating are Dell HPE, Starwind - Happy to hear others
  

 

Is there anything else major I have missed. At the same time we are looking at switching from VMWare to Hyper-V. Are there any considerations surrounding hardware solution to support this?

 

 

[*]If we go down the Server2019 / S2D route specific hardware is required. Is there an official microsoft compatibility list available?

 

With S2D or Azure HCI Stack Here is Microsoft's link to there hardware compatibility site https://hcicatalog.azurewebsites.net/#/.

[supportman]

Hyper-Converged = Over Engineered in my opinion. Just like Sans really.

 

Local SSD storage + a Hyper-V or Vmware and a good local as well as cloud backup solution should be all you need to run any size school really.

[CyBeRkId2002]

@supportman - This is another option I amconsidering. What do you do in terms of redundancy / disaster recovery then? Is it manually managed and recover from a snapshot image of VM's in the case of catastrophic server failure. Are your servers specified well enough to run all VM's (Storage / RAM etc. in the eventuality that they need to)? I suppose my only thinking surrounding this is if the servers are spec'd well enough to do this, why not just go hyper-converged and sleep a little better at night.

[moonbeam]

Local SSD storage + a Hyper-V or Vmware and a good local as well as cloud backup solution

 

I am currently speaking to schools about this kind of solution. As supportman says, local SSD storage disks + Hyper V or VM Ware is a good way to go. Maybe worth looking at Synology DS+ series device with disks running RAID 5/6 mode. Servers VM mode so that a snapshot image can receover quickly in event of failure. Backup local data to Acronis Cyber Protect, cloud backup

 

Happy to have a chat to help you further.

[TechMonkey]

Hyper-Converged = Over Engineered in my opinion. Just like Sans really.

 

Local SSD storage + a Hyper-V or Vmware and a good local as well as cloud backup solution should be all you need to run any size school really.

 

With local storage if a node goes down doesn't that just take all VMs out on that node? I thought the benefit of Hyperconverged was that you used local storage but the magic was the vms were spread about so if a node went down everything still carried on. So the benefits of a SAN without the cost? OR have I got confused along the way?

[supportman]

With local storage if a node goes down doesn't that just take all VMs out on that node? I thought the benefit of Hyperconverged was that you used local storage but the magic was the vms were spread about so if a node went down everything still carried on. So the benefits of a SAN without the cost? OR have I got confused along the way?

Sure it goes down... but with a good backup system you can recover in minutes. I've done it myself with our finance server and hardly anyone even noticed.

[supportman]

@supportman - This is another option I amconsidering. What do you do in terms of redundancy / disaster recovery then? Is it manually managed and recover from a snapshot image of VM's in the case of catastrophic server failure. Are your servers specified well enough to run all VM's (Storage / RAM etc. in the eventuality that they need to)? I suppose my only thinking surrounding this is if the servers are spec'd well enough to do this, why not just go hyper-converged and sleep a little better at night.

We backup all VM's every night with Veeam, so just restore a VM to another host if it ever goes down in a disaster. Our 3 Hyper V hosts and Veeam server are in 2 different physical locations.

 

I sleep far better at night knowing I can restore a server in 15mins to any host in the school than having a complicated hyperconverged setup in a single location. I can even restore the server temporarily to the Veeam backup server in seconds if needed in an emergency, then migrate later in the day when people go home to a permanent host.

Edited April 27, 2021 by supportman

[TechMonkey]

Sure it goes down... but with a good backup system you can recover in minutes. I've done it myself with our finance server and hardly anyone even noticed.

 

Are you backing up minute by minute? Hardly anyone isn't no-one.

I'd be interested in a comparison of costs because if you are backing up enough for a mid day outage to not lose any data there must be a fair amount of storage, at which point does it actually save money? Also it sounds a very manual process, rather than being an automated process.

[Rob_D]

Are you backing up minute by minute? Hardly anyone isn't no-one.

I'd be interested in a comparison of costs because if you are backing up enough for a mid day outage to not lose any data there must be a fair amount of storage, at which point does it actually save money? Also it sounds a very manual process, rather than being an automated process.

 

Surely if you're backup up that regularly you'd use rolling backups with a fairly short history (probably pared with a weekly/daily parked backup), which should cut down on storage requirements.

[supportman]

Are you backing up minute by minute? Hardly anyone isn't no-one.

I'd be interested in a comparison of costs because if you are backing up enough for a mid day outage to not lose any data there must be a fair amount of storage, at which point does it actually save money? Also it sounds a very manual process, rather than being an automated process.

 

Your right the data for the morning would be lost as the backup is only done nightly. I've heard of people doing it hourly, but to me that's overkill in a school scenario.

 

This has only happened once in the last 7 years when a mechanical disk went on a host after an air conditioner caught fire. I can only speak from personal experience where it wasn't an issue for our finance department at the time. They were actually very thankful to be up and running again so quickly after a disaster scenario. It would have been quicker if I had done an instant restore on the backup server, but I choose to do a full restore to a new host.

 

I guess you need to balance uptime vs cost and simplicity based on your organization.

Edited April 27, 2021 by supportman

[RobD]

• A converged solution is one where a Hypervisor sits on the same physical hardware as the storage
  We have our OS on local storage on each host - seperate from the shared storage
  
• Storage is pooled across multiple servers and is resilient in the event of both single server failure (Hypervisor failover) or drive failure (RAID).
  yes, storage DRS basically.
  
• Other options for Hyperconverged worth investigating are Dell HPE, Starwind - Happy to hear others
  We use Cisco, its fine.
  

 

Something we didn't consider:

Each host needs a greedy little VM that controls and monitors the VMs, hosts and storage. This means if a host has 1TB of RAM then 100GB immediately will be used up.

Updating esx becomes more complicated, you can only use a supported version and generally only the tools that are provided.

 

 

Hyperconverged is a really good solution if you're looking to replace hosts, storage and their networking as its a all in one solution. Its also great if HA and IOPS are important to you, if your business can cope with some downtime while you restore from backups then I'd price everything separately potentially with vmware storage DRS.

[Davit2005]

Backup is a backup, raid is no backup so whatever storage you use either local or shared it is still going to be treated the same way when it comes to backups though.

 

For instance when it comes to redundancy if you are running Domain on site with DNS and DHCP make sure there is one DC on each host providing DNS and DHCP just to have 2 DCs as a minimum.

 

INMO There is a cost vs need discussion to be had at the end of the day. Appetite/ability of the org/business to invest for shorter downtimes and redundancy.

Edited April 27, 2021 by Davit2005