Servers and SAN

[lmrogers]

Good Afternoon!!

 

I'm looking at doing a server and SAN refresh.

 

I'm looking at 3 host servers running around 25 VMs and a SAN with around 20TB of storage (preferably SSDs).

 

Any suggestions would be greatly appreciated!

 

Cheers

[Norphy]

I would look at the Hyper-Converged devices.

 

Stone do a system which hosts from two to four nodes and storage in a 2U chassis called Integrum. @Roberto has two I believe. They work out considerably cheaper than the traditional Servers+SAN route and perform as well. It's something I'm going to have a close look at when my next lot of vhosts are up for renewal or if any schools get added to our trust.

[nicholab]

Datacore offer a solution on Lenovo kit or a reference architecture on a manufacture of your choose. It what I would do if I was starting to day. Also Stone and Dell have solution that could work well. The Dell VRTX has the option to use JBOD's as storage.

[TechieWils]

I had a 2 node Integrum at my last place and I loved it. A load cheaper than any other solution and did everything I needed with about 16tb of usable storage. Speak to Joseph Shore at Stone as he sorted mine out. If memory serves there was something about the Dell VRTX solution that put me off but I honestly cant remember now as it was a while back.

[Paid_Peanuts]

Good Afternoon!!

 

I'm looking at doing a server and SAN refresh.

 

I'm looking at 3 host servers running around 25 VMs and a SAN with around 20TB of storage (preferably SSDs).

 

Any suggestions would be greatly appreciated!

 

Cheers

 

I’m assuming you are using Hyper-V?

Otherwise these hyperconverged solutions from Stone/Dell are no good to you if you utilise VMware Essentials Plus using 3 hosts.

[ful56_uk]

I had a 2 node Integrum at my last place and I loved it. A load cheaper than any other solution and did everything I needed with about 16tb of usable storage. Speak to Joseph Shore at Stone as he sorted mine out. If memory serves there was something about the Dell VRTX solution that put me off but I honestly cant remember now as it was a while back.

 

Mind if I ask the rough cost of your setup as I am looking at our setup at the moment and more or less looking at the same usable storage

[TechieWils]

Mind if I ask the rough cost of your setup as I am looking at our setup at the moment and more or less looking at the same usable storage

Was around the £26k mark with tiered storage and 2 nodes. Can't remember how much ram as I don't work there anymore and brought it in 06. But it handled an exchange server, 2 dcs, sims, 4 application servers, print server, moodle, few test servers and other bits I chucked at it with plenty more space. Each mode could run everything and stone setup the instant failover for me as well as the 10gb fibre module for my core hp switch.

 

Honestly couldn't fault it and from what I've been told from my mate still there it still works perfectly.

[PotNoodleTech]

If I was looking at a brand new system now, SAN wouldn't be on the list. 3 x Hosts with Server 2016 Hyper V with Storage Spaces. A 4th host in a separate building as a Hyper-V replica. Veeam for backup.

[Paid_Peanuts]

If I was looking at a brand new system now, SAN wouldn't be on the list. 3 x Hosts with Server 2016 Hyper V with Storage Spaces. A 4th host in a separate building as a Hyper-V replica. Veeam for backup.

 

Can I ask why not? I am going through the process of pricing up our replacement kit (currently consists of 3 hosts and SAN using VMware Essentials Plus).

The indicative pricing I have received so far would indicate that 4 hosts in an S2D configuration, or 3 hosts using S2D with one as a replica is no cheaper then 3 better spec'd hosts connected to a SAN.

 

I appreciate the SAN may be the single point of failure in that setup, but i would like to bet you have many more single point of failures within your organisation which are far more likely to fail before a SAN does.

[lmrogers]

I'm going to be moving from VMWare to Hyper-V so possibly could look at any of these options.

 

What sort of pricing are you getting for the servers and SANs?

[nicholab]

The issue with hyper converged in Schools is that most systems do not offer enough storage at the right cost point. I would not buy a hardware SAN but would consider what Datacores or other software defined storage providers offer. https://www.datacore.com/best-price-performance-fastest-response-time and Storage Performance Council: benchmark_results_spc1_top-ten. Edited January 12, 2018 by nicholab

[Norphy]

The combined cost of the last lot of VMware hosts and SANs that we've bought was roughly £29,000 for three vHosts (16 cores, 256GB RAM, 6 10GBe NICs, 2 1GBe NICs, 2x300GB hard disks) and a SAN (Compellent SC3020 SAN with ~24TB usable storage, 4TB of which is flash). All of this is from Dell. However, we bought the servers just over a year ago and have only just bought the SAN so prices will be different by now.

[TechMonkey]

Otherwise these hyperconverged solutions from Stone/Dell are no good to you if you utilise VMware Essentials Plus using 3 hosts.

 

Sorry, could you expand on this? Is it just the number of processors allowed or some other licensing issue? I've seen a lot of trumpeting about VMWare and Hyper convergence but if they are making it tricky because of licensing then we would need to be careful.

[PotNoodleTech]

Can I ask why not? I am going through the process of pricing up our replacement kit (currently consists of 3 hosts and SAN using VMware Essentials Plus).

The indicative pricing I have received so far would indicate that 4 hosts in an S2D configuration, or 3 hosts using S2D with one as a replica is no cheaper then 3 better spec'd hosts connected to a SAN.

 

I appreciate the SAN may be the single point of failure in that setup, but i would like to bet you have many more single point of failures within your organisation which are far more likely to fail before a SAN does.

 

Maybe, but saying there are other single points of failure therefore lets keep adding single points of failure is not a good argument for me.

 

Some negative points of the SAN system for me (I could probably think of more if I sat down!):

1. Single point of failure. To configure it not to be is hideously expensive. Why spend £10k+ on a single point of failure component?

2. Overly Expensive. (Generally SANS are not cheap to buy, run, or replace, your already buying 3 fairly powerful hosts why not just put the disks in the hosts?)

3. High running costs. (usually, SAN warranty is steep, often with extra contracts or licencing involved)

4. Inflexible. (you've got your san, great but what if your storage requirements double all of a sudden, can you add more drives to it? More trays to it? Do your licences/support contract cover that?)

5. In the long term, your tying yourself into a solution that won't bend and flex like you need. If your using separate hosts and a next gen storage system such as VSAN or Storage Spaces then you can simply add extra hosts, remove old hosts, upgrade existing hosts as you please, on the fly, with no impact to your services.

6. Your at the mercy of your SAN manufacturer. If they decide to stop support, your a bit screwed, if they decide to triple support costs, your gonna have to pay. With commodity hardware and software, you can do what you want, when you want, how you want. If your budget gets chopped, you can buy more reconditioned servers, if a host fails, who cares it's a redundant system, etc.

 

Am I on my own here with this line of thought?

[NB9457]

Maybe, but saying there are other single points of failure therefore lets keep adding single points of failure is not a good argument for me.

 

Some negative points of the SAN system for me (I could probably think of more if I sat down!):

1. Single point of failure. To configure it not to be is hideously expensive. Why spend £10k+ on a single point of failure component?

2. Overly Expensive. (Generally SANS are not cheap to buy, run, or replace, your already buying 3 fairly powerful hosts why not just put the disks in the hosts?)

3. High running costs. (usually, SAN warranty is steep, often with extra contracts or licencing involved)

4. Inflexible. (you've got your san, great but what if your storage requirements double all of a sudden, can you add more drives to it? More trays to it? Do your licences/support contract cover that?)

5. In the long term, your tying yourself into a solution that won't bend and flex like you need. If your using separate hosts and a next gen storage system such as VSAN or Storage Spaces then you can simply add extra hosts, remove old hosts, upgrade existing hosts as you please, on the fly, with no impact to your services.

6. Your at the mercy of your SAN manufacturer. If they decide to stop support, your a bit screwed, if they decide to triple support costs, your gonna have to pay. With commodity hardware and software, you can do what you want, when you want, how you want. If your budget gets chopped, you can buy more reconditioned servers, if a host fails, who cares it's a redundant system, etc.

 

Am I on my own here with this line of thought?

 

Nope.

 

My next Server setup is looking like 3x beefy hosts with local storage, DPM backup and one of our old servers as a Replica for file services. The setup will be incredibly simple and comes in at either £13.5k with mech drives or £19k with SSDs. And that's 128GB DDR4 RAM, 2x8 Core Xeons, Minimum 8TB Storage per host.

 

We have a SAN and blade system currently, it cost a lot to buy, it costs a lot for warranty, it's pretty complex for what it actually does and it's gone wrong badly at least once.

 

Personally, a SAN for a school isn't really necessary, what is necessary though, is having documentation, good backup and plans.

[Paid_Peanuts]

Maybe, but saying there are other single points of failure therefore lets keep adding single points of failure is not a good argument for me.

 

Some negative points of the SAN system for me (I could probably think of more if I sat down!):

1. Single point of failure. To configure it not to be is hideously expensive. Why spend £10k+ on a single point of failure component? Dual controller, dual connectivity, dual power supplies. OK, single back plane/motherboard and disks are only points of failure. Disks just as likely to fail no matter what hardware they are in. Back plane just as likely to fail in a host - who has ever had a back plane/motherboard fail on infrastructure kit from a tier 1 manufacturer? I don't see a SAN as any less reliable then any other method being talked about on here.

2. Overly Expensive. (Generally SANS are not cheap to buy, run, or replace, your already buying 3 fairly powerful hosts why not just put the disks in the hosts?) False. The quotes I have received to date are all competitive and I wouldn't say they were expensive. To achieve the resiliency you all crave, putting disks in hosts requires the purchase of more disks? 3 hosts for example, S2D, 3 x number of disks to create replicas.

3. High running costs. (usually, SAN warranty is steep, often with extra contracts or licencing involved)False. Purchase it originally with 5 years warranty and licensed features required = nothing else to pay. After 5 years warranty extensions might be expensive, but who wants to rock 5+ year old infrastructure kit?

4. Inflexible. (you've got your san, great but what if your storage requirements double all of a sudden, can you add more drives to it? Yes More trays to it? Yes Do your licences/support contract cover that?) Yes

5. In the long term, your tying yourself into a solution that won't bend and flex like you need. Based on 3 host VMware Essentials Plus, each host consisting of dual 18 core, 384GB, 10GB SFP+ connectivity, how much more "flex" do we need? If your using separate hosts and a next gen storage system such as VSAN or Storage Spaces then you can simply add extra hosts, remove old hosts, upgrade existing hosts as you please, on the fly, with no impact to your services.Which typical secondary school requires a 4+ S2D setup?

6. Your at the mercy of your SAN manufacturer. If they decide to stop support, your a bit screwed, if they decide to triple support costs, your gonna have to pay. False. Pay upfront when you purchase the hardware for the warranty. The likes of Dell, IBM, HP publish their support road maps for these products so you know you'll be supported in 5+ years With commodity hardware and software, you can do what you want, when you want, how you want. If your budget gets chopped, you can buy more reconditioned servers, if a host fails, who cares it's a redundant system, etc.

 

Am I on my own here with this line of thought?

 

I don't believe everything you said below to be true. My thoughts to your statements above.

[Paid_Peanuts]

Nope.

 

My next Server setup is looking like 3x beefy hosts with local storage, DPM backup and one of our old servers as a Replica for file services. The setup will be incredibly simple and comes in at either £13.5k with mech drives or £19k with SSDs. And that's 128GB DDR4 RAM, 2x8 Core Xeons, Minimum 8TB Storage per host.

 

We have a SAN and blade system currently, it cost a lot to buy, it costs a lot for warranty, it's pretty complex for what it actually does and it's gone wrong badly at least once.

 

Personally, a SAN for a school isn't really necessary, what is necessary though, is having documentation, good backup and plans.

 

I am currently looking to replace our existing servers which are almost of the same spec, dual 6 core, 192GB. Only thing mine don't have is internal storage.

 

Can I ask how many VM are you planning on running on this proposed spec?

Are you going to run Hyper-V, assuming you are, are you planning on running storage spaces? If you are, how are you configuring the drives and how many additional drives will you have to purchase as S2D on 3 hosts has a 33% efficiency rate?

Which make/model of SSD are you planning on using? Going from £13.5k to £19k is £5.5k difference across 3 hosts. Less than £2k per host for SSD upgrades? Will these be enterprise mixed use SSD?

[NB9457]

I am currently looking to replace our existing servers which are almost of the same spec, dual 6 core, 192GB. Only thing mine don't have is internal storage.

 

Can I ask how many VM are you planning on running on this proposed spec?

Are you going to run Hyper-V, assuming you are, are you planning on running storage spaces? If you are, how are you configuring the drives and how many additional drives will you have to purchase as S2D on 3 hosts has a 33% efficiency rate?

Which make/model of SSD are you planning on using? Going from £13.5k to £19k is £5.5k difference across 3 hosts. Less than £2k per host for SSD upgrades? Will these be enterprise mixed use SSD?

Was just planning on running local storage, around 8 VMs per host. The failover features aren't essential to us, file servers will be replicated and we will have VM backup, so the way we planned it, lose a host, fail back file services to replica and we live without some none core services for a day and copy the VHDs from the core services to another host if they're needed. I just want to keep it as simple as possible and I keep everything as well documented as I can, so if something happens to me, long or short term, somebody could walk in and get a handle on things pretty quick.

 

I've worked at quite a few High Schools now, one had several SANs and over 20 servers, 10Gb switching, it was hugely overkill and was even configured as optimally as it could be, in fact, whilst the backend was all singing and dancing, the clients were P4s and Pentium Duals predating 2010 and this was only going back last year. So £££s spent on the backend, yet the client machines were as slow as anything, how they got away with that, I'll never know.

 

Personally, the way I see it, the money saved not buying a SAN can be spent on bettering the end users device and experience. If all your clients don't have SSDs, that's where the money should be going. I often see threads like this pop up and as stated by myself and @AButters on many other threads, it's overkill, technically a single point of failure and may not even provide the best performance. As I said, documentation and planning is key.

 

SSDs I was looking at were the Kingston DC400 (DC Standing for Data centre)

[PotNoodleTech]

I don't believe everything you said below to be true. My thoughts to your statements above.

 

All fair points! Well I guess I've got my opinion and you've got your opinion!

 

[lmrogers]

So looking at Hyperconvergence, how exactly does this work and is it a good way forward?

[Sephiroth]

Just out of interest, does anyone here use VMWare's vSAN? I'm in a position where I need to look at storage and have been looking at this, but can't see how it actually works.

[Tefters]

Just out of interest, does anyone here use VMWare's vSAN? I'm in a position where I need to look at storage and have been looking at this, but can't see how it actually works.

 

I could be wrong but I believe vSAN is being discontinued next year.

[Sephiroth]

Hmm... Might be a short conversation with VMWare next week then!

[Tefters]

Hmm... Might be a short conversation with VMWare next week then!

 

Like I say, could be wrong but I know they are shelving some products.

[mavhc]

In the end the hardware is all the same, either in a SAN or a Server. So it's mostly about the software and cost. Software moves fast so you have to reevaluate the whole system every time you upgrade.

 

What couldn't you do with the old system? What bottlenecks did your monitoring find? As it's 5 years later you should be getting the same thing 4x cheaper, or something 4x better, or hopefully, something 2x better for half the price.