How complex is your network?

[howartp]

One chassis to rule them all, once chassis to find them, once chassis to fail catastrophically, and in the downtime bind them.

Nice rhyme, but if you've got another £40,000 for me to spend then I'll gladly have it.

 

We have sufficient DR in place to run the estate live elsewhere in school, so system failure of that nature isn't a concern.

[mavhc]

What you needed hardware to do 10 years ago is done in software today.

 

First step is to make sure you're not locked into a system, can expand by buying someone else's drives/servers/etc. If you have to throw something away because it's incompatible with the future it's not great.

 

Define what you want, how much risk mitigation will be, and how likely it is to occur. Will you have times in the future when you can have downtime or is everything 24/7 from now on?

 

Make sure you can test your emergency procedures, and then test them on purpose. Turn off that switch that's supposed to be redundant, check it is. Restore from a backup, does it really work the way you assume?

[PotNoodleTech]

Nice rhyme, but if you've got another £40,000 for me to spend then I'll gladly have it.

 

We have sufficient DR in place to run the estate live elsewhere in school, so system failure of that nature isn't a concern.

 

You've hit the nail on the head when you said DR is so good nowadays that you can restore your entire estate to any server you want within a couple of hours, so why are you wanting to spend £35,000 - £90,000 on a three host ultra system?

 

Not having a go at you, just trying to understand, as here in Wales we've been forced by 20 years of austerity to think very differently. My IT budget for the whole secondary school is barely £90,000 every 5 years.....so spending that on just servers seems obscene to me.

[howartp]

My IT budget for the whole secondary school is barely £90,000 every 5 years.....so spending that on just servers seems obscene to me.

Mine is approximately four times that per annum - IT hardware, software, tablets and licensing - so £35,000 becomes more in perspective.

[mhosker]

Very interesting to read through the responses on here.

 

With regards to the single SAN design, I agree that you may as well just have local storage on each host if your only using a single SAN. Means you have increased redundancy and less management headache!

 

If your using two SANs with replication and then failover clustering that makes sense.

 

That restore to Azure sounds very much like a feature of Datto, great product, but the contract term puts me off.

 

Questions to all of you that use VLANs :

 

A: Do you inter VLAN route on your switches or UTM / Router / Firewall?

 

and

 

B: How do you ensure redundancy with your VLANs? More specifically aimed at those using layer 3 switches for inter VLAN routing. What happens if your core switch goes down?

 

Mike

Edited June 19, 2019 by mhosker

[Duke5A]

With regards to the single SAN design, I agree that you may as well just have local storage on each host if your only using a single SAN. Means you have increased redundancy and less management headache!

 

• 5k students
  
• 7 buildings
  
• Private fiber
  
• 70 switches
  
• Wireless - 180 access points
  
• CCTV - 190 cameras
  
• 600 Chromebooks
  
• Lots of Windows machines
  

 

Cluster consists of:

• Three hosts - Dell R710
  
• One SAN - MSA P2000 G3
  
• Two switches for storage
  

 

I honestly don't see the need for multiple SANs. The P2000 I'm running has redundant controllers with redundant links to the storage switches. Hosts have redundant links back to the switches. Only way I could see using local storage on the hosts is if you're using vSAN, or whatever MS's equivalent to it is. If you're in the VMWare crowd then you should already know vSAN is nothing but a headache. Can't speak to MS. All three hosts are diskless and boot from an SD card. Through upgrades over the years one host has enough resources to handle all the virtual machines.

 

Veeam runs backups each night that are in turn replicated to one of the other buildings. Got an offline backup server that gets switched on every week for jobs on top of that.

 

If there was a major failure with the SAN getting back up wouldn't be quick, but honestly, we've got a four hour support contract through HP for it.

 

 

I do still keep a physical domain controller and two more on another offsite host.

[PotNoodleTech]

Mine is approximately four times that per annum - IT hardware, software, tablets and licensing - so £35,000 becomes more in perspective.

 

Not jealous. Not jealous at all...............

 

 

..........ok jealous, very jealous

[Jcx500]

Very interesting to read through the responses on here.

 

With regards to the single SAN design, I agree that you may as well just have local storage on each host if your only using a single SAN. Means you have increased redundancy and less management headache!

 

If your using two SANs with replication and then failover clustering that makes sense.

 

That restore to Azure sounds very much like a feature of Datto, great product, but the contract term puts me off.

 

Questions to all of you that use VLANs :

 

A: Do you inter VLAN route on your switches or UTM / Router / Firewall?

 

and

 

B: How do you ensure redundancy with your VLANs? More specifically aimed at those using layer 3 switches for inter VLAN routing. What happens if your core switch goes down?

 

Mike

 

Using my core l3 switch for inter vlan routing, I guess my redundancy is my hp warranty which is 24 hour call out.

[MicrodigitUK]

Very interesting to read through the responses on here.

 

With regards to the single SAN design, I agree that you may as well just have local storage on each host if your only using a single SAN. Means you have increased redundancy and less management headache!

 

If your using two SANs with replication and then failover clustering that makes sense.

 

That restore to Azure sounds very much like a feature of Datto, great product, but the contract term puts me off.

 

Questions to all of you that use VLANs :

 

A: Do you inter VLAN route on your switches or UTM / Router / Firewall?

 

and

 

B: How do you ensure redundancy with your VLANs? More specifically aimed at those using layer 3 switches for inter VLAN routing. What happens if your core switch goes down?

 

Mike

 

We have 2 Cisco routers for L3 between VLANs and applying ACLs. They are set up to use Virtual Router Redundancy Protocol (VRRP).

[Tefters]

Simple 3 node VMware cluster with a mirrored SAN however its hitting 8 years now and showing its age.

 

Looking into HCI solutions with Server 2019 for next year however not impressed with HP or Dell's offerings as its old gen kit (gen 8) that is overpriced and then they just come in and install windows and configure the cluster.

For the same price they are touting I can do it myself and just buy the hardware from the like of Broadberry but get all the latest gen kit and so much more value in the hardware (more ram, NVMe's etc...) for pretty much the same cost they want to come in and install/support the whole lot.

 

I'm confident in setting up the HCI solution from a software level so its just a case of weighing up the DIY vs vendor installed scales with SLT.

[DrCheese]

Everything I read on reddit/sysadmin et all gave horror stories about self rolled Storage spaces direct - The hardware requirements are really specific and MS will just laugh at you support wise if you don't have a supported solution. May be fear mongering but...

 

It's quite difficult to self test with whatever kit you have laying around as well, due to the above h/w requirements

[Tefters]

The hardware from Broadberry is all certified kit (supermicro) and on the support hardware MS list, mellowex 25gb throughout etc...

[DrCheese]

ah right, for some reason I didn't expect supermicro gear to be on the this list...

 

https://www.microsoft.com/en-gb/cloud-platform/software-defined-datacenter

 

Turns out it is!