few bits of new kit

[manick]

Hi all,

 

I have a small pot of money to start along the virtualisation path. I only have 3 physical servers which cater for around 1000 people (staff/students/admin with only 1 DB at present), I would obviously in the future add a few more to split out services.

 

Whilst I would love a couple of SANs and 3 hosts with VMWare essentials plus this is not possible with my budget (<20k). I have had a couple of cut back solutions presented to me that look viable but I've also noticed the SCALE HC3 offering and was wondering if anybody else is using this or what they thought of it.

 

Has anybody else had to virtualise on the cheap and if so what have you implemented and has it been successful?

 

Cheers

 

Manick

[Steve21]

Wouldn't waste the money on these "special" kit if I'm honest

 

For 20k you could easily get 2-3 brand new hosts with built in storage. Replication for free using HyperV etc.

 

What's the overall aim though? Did you want failover ability? Replication? Just "virtual"? Just a refresh/new kit? etc

 

Steve

[manick]

I don't know about wasting money, they seem a fairly cheap and straightforward product, however....

 

My main aims are I want to increase users storage space from a few hundred Meg to a few Gig and provide better failover capability for quick recovery with as little downtime as possible. Two of our current servers are end of life and are in desperate need of replacement the third can be re-used.

 

Manick

[JoHimself]

I don't know about wasting money, they seem a fairly cheap and straightforward product

 

It is my experience that you get what you pay for.

 

Stick with proper servers (Dell, HP etc) and proper software (ESXi, Hyper-V, XenServer at a push). It will be much easier in the long term if you have standard configurations for troubleshooting and advice.

 

Start by looking at what you have by way of CPU, RAM and storage (capacity and performance) and work from there. Obviously you don't want to be virtualising for the sake of it or spending £20k with no visible benefit to the business.

[Oaktech]

For 20k you will be able to do what you need to... I would be looking at 2 well specc'd hosts and some shared storage - if you speak nicely to dell you might be able to get supported pricing on an equalogic san with room to upgrade when finances allow. You could run it iscsi on a vlan provided you have a decent core switch.

 

I'm looking at doing a similar project with a similar budget.

[elsiegee40]

And don't forget that @ICTDirect_Dave can probably sort you some great warrantied refurb kit that will extend your tight budget

[ICTDirect_Dave]

Goodness! for £20K we can give you a pile of excellent servers

 

Quick example of what you can get for less than £1.5K from us:

 

Refurbished 2U HP SE326M1 Datacentre Edition Server

 

2 x Hex core CPU (12 Cores/24 threads)

96GB RAM (PC3-10600R) 12 x 8GB

25 x 300GB SAS HDD (Gives 3.6TB Storage in RAID10)

2 x PSU

P410i Raid Controller with Battery Backed Write Cache

Rails

2 Years Warranty.

 

We've got a number of these in use in schools across the country and they're proving to be amazing machines.

[apoth0r]

Cluster-in-a-box - Primergy In Action : Fujitsu UK

 

Exactly what you need a much cheaper than you think!

[manick]

Thanks for the replies guys, plenty to think about. It might be I do purchase refurb' gear I'll have to have a think about it.

 

I have had a couple of options put forward to me from firms but I'm doubting that what they are putting forward will be fit for purpose. We carried out an assessment of disk drive activity and it came back that over 3 servers we were performing at around 800 IOPS during the day and up to 1500 during the night (backup traffic). The solutions put forward are coming in at around 600 IOPS (2 hosts, 1 replicating the other and a backup device). I did give them the IOPS findings, Is there something I'm missing here?

 

Any thoughts?

 

Thanks

 

manick

[bossman]

@manick

 

We've just gone down this route our setup is as follows:

 

2 x Dell poweredge R710s Spec:

Dual Xeon 12 core, 196GB Ram, 2 x 146GB sas drives mirrored, Quad Nics teamed for failover

 

Backend storage:

Synology RS3614xs+ with 4 x 450GB SSDs for fast caching and 8 x 8TB sata drives for storage and have both units setup in a failover cluster with heartbeat.

 

Setup:

Server 2012R2 on both Dell servers:

Hyper-V cluster with DPM and offsite backup

 

Licensing:

EES data centre

 

Slightly more than £20k (£5k more) but should be able if you use refurbished servers from ICTDirect.

 

We do have another synology unit with 32TB extra storage which we have for backup storage and data.

 

IOPs galore with the SSD fast caching on the synology so that shouldn't be a problem.

 

Take a look at them: https://www.synology.com/en-uk/products/RS3614xs+

 

Have fun

[howartp]

As another option, I just bought Dell VRTX with 4 good hosts and storage for £18k.

 

We're running VMware Essentials on 3 of the hosts, and using 4th for Veeam SureBackup etc, but you could put HyperV on them.

[mikkydoos]

It is my experience that you get what you pay for.

 

Stick with proper servers (Dell, HP etc) and proper software (ESXi, Hyper-V, XenServer at a push). It will be much easier in the long term if you have standard configurations for troubleshooting and advice.

 

Start by looking at what you have by way of CPU, RAM and storage (capacity and performance) and work from there. Obviously you don't want to be virtualising for the sake of it or spending £20k with no visible benefit to the business.

 

He's got it.

 

I have a sub 20K budget and have VMWare and Veeam running no problem. It's not that expensive. I splashed out 6K on a decent HP host then its maybe 2k / year to license it and Veeam.

 

Or go the Hyper-V route and add the cost to your MS licensing. Again, it won't break the bank.

 

Oh, and you can just start with one host. I run one Esxi host and it's never failed in 3 years. As soon as I get to virtualise our physical MIS server on it, I can turn that machine into host no. 2.

Edited November 25, 2015 by mikkydoos

[mikkydoos]

Cluster-in-a-box - Primergy In Action : Fujitsu UK

 

Exactly what you need a much cheaper than you think!

 

Y, Fuji servers are good... and cheap !

[PotNoodleTech]

20k is a decent budget! Nowadays you can get an immensely fast immensely reliable system for well under 20k. I'd go for something like:

 

2 x Dell R730 (you won't need to go mad on the spec, 2 x 6 core and 32GB each would be plenty for you by the sounds of it) local storage on each I suggest 6 or more SSD Drives in each.

1 x single CPU server for Veeam backup server and to act as an emergency hyper-v server

2 x APC UPS 3000

3 x Win2012R2 Datacentre licences

2 x 2 CPU Veeam Backup Essentials for education licences

 

You'll get all that for under £15k.

 

If you do want to spend the full 20k, instead of using local storage, go shared storage to do clustering in hyper-v and get a SAS JBOD Array as the shared storage (it's much more reliable and faster than iscsi/san and HyperV can use it natively in storage spaces with advanced features like SSD caching).

 

Really there's no need to spend £50k on school server systems any more IMHO. The people telling you to spend that are probably the people selling it to you.

[PotNoodleTech]

Also, personally I avoid cluster in a box type solutions. The whole point of a cluster is to spread out you're services over multiple hardware(s) for maximum reliability. Cluster in a box takes that idea and then crams it all into once piece of hardware which IMHO is totally backwards, putting all you're eggs in one (expensive) basket you'd better pray you're cluster in a box chassis doesn't suffer a catastrophic failure of a b0rked firmware update. Wheras if you had 2 or 3 separate hyper-v servers? One's died, who cares the other two are still up, no problem.

[ITGURU]

Physical or virtual is user preference and no one solution fits all. I still believe in having physical servers with local storage for critical systems, i.e SIMS, Exchange, File Access. If you lose one server for any reason your only losing part of your network.

 

I'd have to be convinced to change to anything else.

 

I have 2 hyper-v hosts but those only run non-critical apps such as WSUS, McAfee EPO, Paxton Access etc so if have to take down or have hardware issues they have no effect on end users.

[JoHimself]

Conversely your mission critical systems are best protected by the HA systems available in virtualised environments. And technology such as DRS can help to improve the performance of key systems. And virtualisation means, even in the event of hardware failure, key systems can be brought back up quickly.

[PotNoodleTech]

Conversely your mission critical systems are best protected by the HA systems available in virtualised environments. And technology such as DRS can help to improve the performance of key systems. And virtualisation means, even in the event of hardware failure, key systems can be brought back up quickly.

 

Yeh I would completely agree. All virtualisation does is give you many different benefits that can increase speed, reliability, disaster recovery. There is literally no reason to go with a physical system, particularly if you are buying new hardware anyway.

 

Physical system = less reliable, less disaster proof.

[ITGURU]

Yeh I would completely agree. All virtualisation does is give you many different benefits that can increase speed, reliability, disaster recovery. There is literally no reason to go with a physical system, particularly if you are buying new hardware anyway.

 

Physical system = less reliable, less disaster proof.

 

 

A physical system is no less reliable than virtual. In 13 years (touch wood) I have never had a server fail. Had the odd hard drive fail and still using servers that are 6 years old.

 

Why put all your eggs in one basket or have to double up on everything for redundancy/failover i.e 2 x SANs.

I once had to manage a network that had a single SAN/2 x ESX systems and spent more time sorting their problems out and if they lost their SAN, kaboom that's the entire network down.

 

I have a happy healthy reliable network and no reason to change it.

[JoHimself]

A physical system is no less reliable than virtual.

 

Again, conversely it is no more reliable than virtual.

[PotNoodleTech]

A physical system is no less reliable than virtual. In 13 years (touch wood) I have never had a server fail. Had the odd hard drive fail and still using servers that are 6 years old.

 

If you're physical system fails, you're screwed. Can you grab another server and auto-restore you're backup on to that within an hour? No. Because physical systems rely on specific hardware and drivers, you'd have to set the system up from scratch, install all the drivers, then start manually restoring all of the files from you're backup. And you have to do this manually, the'red be a lot of testing needed at the end to make sure all the shares, services, permissions are working well. I hope you're not ill the day that happens.

 

Virtual? Even the most basic free version has the ability for you to just copy the image on to a totally different PC and just "press the play button" - hey presto, system up. Can be easily automated too, incase you're ill.

 

Why put all your eggs in one basket or have to double up on everything for redundancy/failover i.e 2 x SANs.

I once had to manage a network that had a single SAN/2 x ESX systems and spent more time sorting their problems out and if they lost their SAN, kaboom that's the entire network down.

 

Completely agree with you there! But here's the secret, you don't have to go mad and spend £50k on double SANS to get almost all of the reliability/disaster recovery benefits of virtualisation. You an do it all on a £1.5k server. In an emergency, you could literally go to an IT suite, grab 5 x core2duo machines, repurpose them with the VMsoftware of you're choice, and restore you're virtual machines to those. Hey presto, network up, easily, for free, until you can get some kind of serve repair/replacement.

 

I have a happy healthy reliable network and no reason to change it.

 

Excellent. That's all all of us are doing here it trying to keep it that way!

 

All virtualisation does is give you more flexibility, more and faster options on how to get dead hardware back up. More often than not it can be done for free with Free ESXi or nearly for free with EES Windows Server licencing.

Edited November 25, 2015 by PotNoodleTech

[m25man]

I think you need to look carefully at all of the benefits of a virtualized system before dismissing them or engaging them completely.

 

Rackspace is expensive to fill, power and cool and those alone make a good case for virtualisation.

Then you add the convenience of recoverability and efficient use of disk space through shared storage be it NAS/SAN/JBOD

 

The unexpected will strike everyone at some point but the cost of redundant everything is simply unrealistic in a school.

I've seen serious public institutions being run on server infrastructures less robust than many primary schools!

 

Over the years I think (no I Know!), that the education sector has inherited a lot of quite experienced IT staff from various business sectors such as finance whom cant sleep at night without having spent 3 yrs worth of IT budget on enterprise solutions that will never be put to task.

 

We all want the easiest ride through our working life and a balanced view on everything is what's important.

 

New technologies like Hyper-V replica now mean we are not dependent on an expensive SAN however we don't want to drift back to having dozens of servers with under utilised CPU's RAM and islands of un-inhabited storage.

 

We are currently living in a Hybrid world of IT with a lot of stuff moving to the cloud and a lot of stuff we would rather keep locally.

 

Before jumping in and blowing the budget on a Cluster and a SAN see what you can offload to O365 or Google first.

You may be surprised at just how little you actually need!

 

Virtualising your critical stuff and replicating VMs in near/real-time might be a viable solution to a full blown failover cluster.

[ITGURU]

If you're physical system fails, you're screwed. Can you grab another server and auto-restore you're backup on to that within an hour? No. Because physical systems rely on specific hardware and drivers, you'd have to set the system up from scratch, install all the drivers, then start manually restoring all of the files from you're backup. And you have to do this manually, the'red be a lot of testing needed at the end to make sure all the shares, services, permissions are working well. I hope you're not ill the day that happens.

 

 

Virtual? Even the most basic free version has the ability for you to just copy the image on to a totally different PC and just "press the play button" - hey presto, system up. Can be easily automated too, incase you're ill.

 

 

 

Completely agree with you there! But here's the secret, you don't have to go mad and spend £50k on double SANS to get almost all of the reliability/disaster recovery benefits of virtualisation. You an do it all on a £1.5k server. In an emergency, you could literally go to an IT suite, grab 5 x core2duo machines, repurpose them with the VMsoftware of you're choice, and restore you're virtual machines to those. Hey presto, network up, easily, for free, until you can get some kind of serve repair/replacement.

 

 

 

Excellent. That's all all of us are doing here it trying to keep it that way!

 

All virtualisation does is give you more flexibility, more and faster options on how to get dead hardware back up. More often than not it can be done for free with Free ESXi or nearly for free with EES Windows Server licencing.

 

 

 

Yes, all my servers are the same make/model and always have a spare.

Depends on the system. 2012 R2 is pretty good with drivers to get a basic server online in 20 minutes.

Depends what the issue is, we all have backups so even if the local storage failed/SAN failed you wouldn't be able to access the data to copy.

[clockend25]

If you're looking for bang for buck, have a look at Server Source. We grabbed a couple of Dell C6100s from them a few years ago which scored us 64 cores and 768gb of RAM in total for about £10k. I'm still on their mailing list so see the offers they have on the go each week, and they're somewhat tempting to say the least.

[ITGURU]

Even the biggest of companies are meant to have resilient systems in place, but how often do they work when they are needed and don't failover properly.

 

Each to their own really!