Thin Client and Virtual Machines Thread, Help!! urgently looking for comments for governors on a virtualised server environmet in Technical; Hi Everyone
I'm currently having a problem committee of are governors that we need to build in redundancy into our ...
23rd July 2012, 11:32 PM #1
- Rep Power
Help!! urgently looking for comments for governors on a virtualised server environmet
I'm currently having a problem committee of are governors that we need to build in redundancy into our new virtualised server environment. unfortunately this is delaying the project and he's asked me to get statements and recommendations on schools who are running a similar environment and why it is necessary. if anyone can e-mail me a comment in support of this spec and redundancy etc who are using a similar system in their school please e-mail me at. email@example.com
here are some of his comments.
"Please can you do research on similar schools with fairly recent new
installations, that will help confirm what we have specified is required. I
understand their may be some differences.
If you use this spec to go through with them so that we have some bench
"The problem is the quotes at the moment are about fixing a problem, which
is hypothetical and not based on measured capacity or usage."
"I accept the requirement for the virtual server, the backup server is more than the implied £5000, it is £9K with the duplicated licences and also the installation costs will also be higher as there is alot more to do and it is more complicated so it is a third of the quote. Yes I also accept that it should mean that there is never down time due part failure"
the basics that looking at is
HP DL380 G7 Tower Server 2
Hp SAN P4000 1
Windows Enterprise Server 2008 2
APC Smart-UPS 3000 RM 1
8TB NAS Device 1
Project Management Days 2
Installation Days 10
I've attached some notes I've made and feel free to check out the school website at http://baskvill.bham.sch.uk/ for information.
please help me with this as arena to get this project off the ground and running before September.
many many thanks
Most InportantPowerPoint Notes And Justifcations For Vertual And Redundancy and spec.docxBaskerville School new network prposal govners.docxSpec Draft 1.docx
23rd July 2012, 11:43 PM #2
The DL380 is only Rack mountable, the ML series are pedestal based.
The APC 3000 needs to be the RMXLI with runtime battery otherwise you will only have around 6 minutes to shut everything down...
23rd July 2012, 11:56 PM #3
- Rep Power
I'm after rack mount and checked all the other proposals and that correct I think. I'll have to look into the APC tomorrow with are suppliers
24th July 2012, 09:53 AM #4
Hmm, nice to see Governors properly checking out projects and asking questions, even if it is causing you headaches this time.
I'll post a few bits and bobs for you. Normally I'd email them, but I've had requests from other people for these docs so I might as well just post a shortened, sanitised version here.
I really pushed a virtualisation and consolidated storage project here a few years back. I think we've been running it for about three years now, and our current spec is:
3x Dell R710s with 2x E5620, 48GB RAM, 8x 1Gb ports, etc.
Oracle S7000 storage
Docs to follow...
24th July 2012, 10:00 AM #5
NOTE: December 2010, so may need updating.
Virtualisation offers the best solution to growing ICT requirements, a way to manage ICT infrastructure upgrades, and a better alternative to the traditional, expensive methods of solving these problems – particularly in a time of tight budgets. By virtualising ICT infrastructure, there are additional benefits of improved service resiliency and business continuity. This document aims to provide a brief outline of these points.
At present, introducing a new service (e.g. a web server, email server, networked exam software, SQL database) onto the network that cannot co-exist on an existing server will require the purchase of new hardware (may include a server, networking, power distribution, etc.), operating system licence, and the time to configure all of this equipment. Additional costs would also come in increased power and cooling requirements in the server room.
In a virtualised environment, deploying a virtual server requires no additional new hardware or networking (assuming there is free capacity on a virtual host), operating system licensing could already be covered in existing costs, and the time taken to deploy all of this can be vastly reduced through the use of virtual machine templates.
This principal scales as required until another virtual host needs to be purchased. Typical consolidation ratios of physical machines being virtualised to a host server are between 5:1 and 20:1, meaning there should be no requirement to regularly expand the number of physical hosts unless we are presented with a sudden requirement for a large number of new virtual machines.
By virtualising our server infrastructure and consolidating our storage requirements, we can resolve the issue of requiring new hardware each time we need to expand the network to meet new growth and requirements. In the process this will reduce the time and costs required for these upgrades.
Redundancy, Business Continuity and Disaster Recovery
At present the majority of our services on the network are tied to one physical server. If that server fails, then we lose that service until the server can be recovered. This may be a time consuming process as it could involve having to procure a new server, install the operating system and applications, then restore data from the last backup. Even with next business day warranties, it is likely that this process would take several days until the service was available again.
Virtualisation removes the direct link between the server hardware and the operating system and applications running on it. It does this by running virtual machines on a hypervisor that presents ‘virtual hardware’ to the operating system on the virtual machine. As the hypervisor is the same on all the physical hosts, it is possible to move a virtual machine from one host to another without any reconfiguration. The virtual machines themselves can be stored on mirrored storage devices that are independent to the servers, and so can be accessed by multiple hosts.
Using virtualisation it is possible to reduce the downtime involved with a server failure from days or weeks to minutes or hours. In certain scenarios it can even be possible to have a physical host fail and the virtual machines running on it suffer no downtime at all. The benefits here are clear, and in a time of ever-increasing ICT dependence they are more relevant than ever.
A virtualised infrastructure is also far easier to recover in the case of a disaster than a traditional, physical one. While backing up and restoring data is fairly simple, backing up an entire physical server in a form that can quickly be restored is quite difficult due to the variations in hardware and the way the operating system uses them. With a virtual machine, the ‘virtual hardware’ is standardised and it is possible to back up a virtual server simply by copying a few disk and configuration files while they are in a stable state, rather than the thousands which make up a physical server and are difficult to guarantee the state of. If a situation occurs that requires restoring the majority of the network from backups, this will be far easier to achieve with virtual servers than physical ones.
Storage is the fastest growing requirement on our network. A traditional infrastructure would involve purchasing additional storage servers and splitting the data between them. This makes growth and management difficult, and requires regular upgrades which involve data migration.
By using consolidated storage devices (usually referred to as a SAN [Storage Area Network] or NAS [Network Attached Storage] – however the device we currently use offers Unified Storage, which means it can be used as both a SAN and NAS) we can have a physical device that contains far more capacity than a typical server, is easier to manage, and can also be expanded without resulting in data fragmentation. These devices are also essential to server virtualisation as they provide the central data storage for the virtual machines to run from. It is important that this storage is not on the virtual hosts themselves as the virtual machine data needs to be available even if the host fails, so that another host can take over.
By consolidating our storage we vastly improve management and scalability, but it is important not to create a single point of failure in the storage device itself. To resolve this we can replicate between two storage devices so that if one fails, another can take over. This, along with server virtualisation, can help remove the usual points of failure within a network and so greatly improve our ICT infrastructure reliability.
The initial costs involved with a full virtualisation and consolidated storage deployment are fairly large. However, over time there should be a cost saving due to the reduction in physical hardware and upgrades required.
Assuming a single-site virtualisation deployment, we could reduce our physical servers from twenty-two down to approximately seven. When hardware replacements or upgrades are required there is an immediate saving simply because there is less hardware to replace. Ongoing costs are reduced as there are less server warranties to keep active (although in exchange there are other virtualisation licensing subscriptions) and less power is being consumed – both by the servers themselves and the air-conditioning unit.
It may also be possible to reduce Microsoft server licensing costs by virtualising. At present we buy one Standard or Enterprise licence of Microsoft Server for each physical server. Enterprise Server grants the ability to run four virtual machines, so four Enterprise licences would cover 16 virtual machines. It may work out cheaper to by a few Enterprise licences (or a Datacenter license – unlimited virtual machines) rather than individual licences of Standard.
Deploying a physical server can take a considerable amount of time. It usually involves specifying and purchasing the hardware, waiting for delivery, installing the operating system, then installing and configuring the desired services. In contrast, deploying a virtual server can be completed in a matter of minutes with proper preparation. A virtual machine template can be created which contains the hardware configuration, operating system install and basic service configuration. All that is then required is to clone the template and perform configuration specific to that machine.
Although, as a school, we do not deploy production servers on a regular basis, the ability to do so quickly will always be a benefit when new services are required on short notice. The ability to rapidly deploy servers is also incredibly useful for testing and development purposes, freeing up time ICT Support staff would otherwise spend installing and re-installing operating systems and software to test it.
It is possible to continue running the ICT infrastructure in its current format, however doing so will require a large financial investment on a regular basis and reduce manageability as the network grows. We will also struggle to provide a backup and disaster recovery that is appropriate to the dependence on ICT in school. Virtualisation and consolidated storage is not a flawless solution and it does require a large initial investment. However, considering the growing ICT requirements, limited budgets, and the benefits it offers, it is a greatly preferable option.
• Expanding traditional ICT infrastructure normally requires the purchase of new hardware. With virtualisation it is usually possible to continue expansion without requiring new physical servers or storage. When these purchases are required they will be more cost effective than when upgrading a traditional, physical network.
• If a server fails in our current network, we will lose the service (e.g. email, website, SIMS) running on that server. Virtual servers can be migrated from one physical host to other with little downtime in the case of hardware failure. Virtual machines are also far easier to backup and restore than physical ones and we would be able to recover the network in a far shorter time in the case of a disaster.
• By consolidating our storage we provide a central data store that enables virtualisation, and we are able to expand and manage our data without causing fragmentation. Although the storage devices are not cheap, they are far more powerful and flexible than individual storage servers. To eliminate a single point of failure, it is possible to replicate data between two storage devices.
• Although the initial costs involved with virtualisation are relatively high, the reduction in physical hardware (and associated warranties, power and cooling) and improved cost effectiveness of upgrades can result in a reduced long-term expenditure compared to maintaining a non-virtualised network of the same size and performance.
• In many ways a virtualised network infrastructure is far easier to manage than a physical one. There are time savings that will allow an improved response to requests for new service deployments, vastly reduced time required for software and operating system testing and development, and a huge improvement in the time taken to recover from a critical server failure.
• Overall, virtualisation and consolidated storage offers us the best way to maintain our status as a Computing Specialist school that leads the way with up to date software, services and resources. Considering our ever-increasing requirements for network growth and new ICT features, it is important that we find a solution that is both financially viable and will continue to meet these needs in the long term – virtualisation and consolidated storage can offer us this. After discussion with a large number of hardware and software vendors and resellers, as well as other schools in the UK, our plans as outlined are agreed to be a positive way for the school to maintain its strong ICT position.
DISCLAIMER: I am not liable for any errors in this document, use it at your own risk. Feel free to use the information in here and copy short sections if you find it useful, however please do not copy it in its entirety.
(let me know if this is any use as I have more)
24th July 2012, 10:03 AM #6
In a virtual environment with no redundancy, you literally are putting all your eggs in one basket. In a standard physical environment, you may have a 15% chance of failure, per server - but a single server outage affects no other systems.
A non redundant virtual system may have the same 15% failure chance, but a hardware outage affects *all* systems, potentially to greatly damaging effect.
A redundant system bypasses this concern, by allowing virtual machines to migrate between hosts as needed. this also produces other benefits, such as allowing maintenance/upgrades to be done within normal working hours
That'd be my starting point anyway..as for specs, we're not a school, but what you've specced is close to what we use as a test bed in one of our satellite offices, if you need any details, just shout
24th July 2012, 10:08 AM #7
Here's the costings plan, may be more useful to you. I've removed the actual prices and the suppliers we worked with, although the usual props to Andy @ WTL Ltd, Andy and Kim @ Cutter Project, and Wayne @ CCS Media.
Virtualisation Business Proposal
NOTE: October 2010 – will need updating.
Physical VMware Hosts
These are the physical servers that run VMware ESX/ESXi to host the virtual machines. They are high-spec dedicated servers with lots of RAM, NICs, no internal HDDs, etc. Initially we will buy three servers which will provide enough resources to host the vast majority of our current servers, and as these older servers are virtualised they will become available to use as VMware hosts for expansion or a DR site.
Dell PowerEdge R710
2x Xeon L5520 (2.26GHz)
48GB (12x 4GB)
Dual Port Broadcom Gb NIC
Quad Port Intel Gb NIC
Dual 870W PSU
On-board SD Card (1GB)
Ready Rack Rails (no CMA)
3 Year Support
VMware vSphere 4 and vCentre Licences
VMware vSphere (ESX/ESXi 4.0) is the hypervisor that is installed on the bare-metal host servers. It acts as a layer between the hardware and virtual machines. vCenter is the tool that manages the vSphere cluster, controls configurations and monitors the hosts. vSphere is licenced per-socket so we will need 6 licences initially (3x dual-socket servers) and will need further licences if we wish to re-use our old servers as VMware hosts.
Academic VMware vSphere 4 Enterprise for 6 CPU
Academic VMware vCenter Server 4 Standard for vSphere
1 Year VMware Vendor support Mon-Fri 9-5pm for 6 CPU vSphere Enterprise and vCenter Server
Veeam Backup Software
The Veeam backup software integrates with VMware and allows us to snapshot, backup and restore entire virtual machines. Restoring a VM is extremely simple and in terms of disaster recovery this will be one of our biggest improvements. Veeam is licenced per-socket for the physical hosts, so we will need to buy further licences if we add more hosts at a later date.
Veeam Backup for 6 CPU VMware Environment
1 Year Support (Academic)
Professional Services and Support
In order to ensure we follow the best practices for setting up a virtualised environment and to assist with the initial install and physical to virtual migrations, we will use the services of <supplier> who have extensive knowledge in this area, and are also familiar with our Sun S7000 storage device. <supplier> will provide documentation and training for the install, and on-going support. Rather than offering generic VMware install and training services, <supplier> will be able to provide expertise that is specific to the type of install we are deploying.
Plan, install, configure and test VMware vSphere, vCentre and Veeam backup for 6 CPU
1 Year - 1st Line support of VMware vSphere Environment
Mon-Fri 9am -5pm
3 x vSphere Enterprise Servers
1 x vCenter Server
Our Supply Requirements
In addition to the above, we will need to provide the following:
Install Dell servers in rack
Connect sufficient UPS
Dedicated storage network via HP ProCurve 1800-24G
Sufficient bandwidth to curriculum network via Core Switch
All network cabling
List of IP addresses and hostnames available for use
Access to Sun 7410 (and 7120 if purchased) which will be on latest stable firmware release
A server to use for Veeam backup and access to shared storage other than the Sun 7410 (or 7120 if primary)
The new Oracle S7x20 storage range has just been released and I am awaiting confirmed pricing. This business proposal covers a ‘single-site’ virtualisation solution. In order to provide a fully redundant solution, we require a ‘second-site’ location that will host a second SAN and set of VMware hosts.
DISCLAIMER: I am not liable for any errors in this document, use it at your own risk. Feel free to use the information in here and copy short sections if you find it useful, however please do not copy it in its entirety.
Last edited by Duke; 24th July 2012 at 10:09 AM.
24th July 2012, 10:13 AM #8
Big one this is the industry stand way of supporting this number of user!!
24th July 2012, 10:28 AM #9
Useful advice above... I'll only offer advice for anyone "having problems with governors"
The Governing Body's role is strategic, not operational. The school's SLT manages the school and the GB is there to challenge those decisions ensuring that the school receives Best Value ... which doesn't necessarily mean cheapest. The Governing Body is not doing its job if they don't ask questions and Ofsted is looking for evidence of that challenge in the GB documents. The GB is now judged by Ofsted under a single leadership heading, not separately. It seems to have awoken some SLTs to working more closely with their GBs than before.
You are being asked these questions because it is a governor's job to challenge decisions being made at the school. To do that, the governor(s) need to understand that you have done all your homework and that this really is offering best value for the school.
If everything is properly explained... and assuming the money physically exists in the budget... the governors will agree with school decisions.
With reference to the budget, as a governor I've had to agree with the decision not to purchase a load of PCs this summer. It was painful, but the money was needed more urgently elsewhere.
24th July 2012, 11:31 AM #10
Can I just say you want to be careful saying that it's a redundant solution because it's not. You are relying on a single SAN, a single UPS and have you sized it correctly so that the servers can be run on one server if the other one fails? You also need to look at your disaster recovery plan and discuss expected recovery times. Look at your critical infrastructure services such as Domain Controllers, File Storage, Email, MIS and get maximum downtimes for these systems and acceptable amount of data loss. You then need to provide a solution based around these times and don't be over optimistic on the speed it takes to recover as it takes a lot longer than you expect. For instance, what happens if the SAN fails, what if you lose the server room to fire/electrical problems, what if you're run over by a bus?
Hardware fails, software has glitches, expect the worst. As I have commented on another thread, we've been bitten badly 3 times in the last academic year, with some large outages due to fire, hardware failure and a SAN/Switch software glitch. We are now installing a full blown High Availability solution across 2 server rooms which is costing us a serious amount of money, but SLT/Governors were prepared to spend this to get the required recovery times.
24th July 2012, 12:54 PM #11
Here's 2 pence...
We set up a windows network last year and went virtual for our service servers, by that I mean everything except the AD servers.
We have two virtual hardware servers. Both identical, they run Server 2008 Data center on 52gb RAM with 2 quad core CPUs
They are connected via two switches spliced so that each server connects to each switch and the switches are then in turn spliced into the Data SAN.
The SAN hardware has two controller boards which each have 8 ethernet ports meaning that you can loose "half a SAN" and still have full data connectivity. Each board is served it's own power and network so they are independent.
The way it works out is that server1 is connected to Switch1 and Switch2 which are both then connected to the SAN, each switch is connected to each controller on the SAN (See the attachment)
This gives almost total redundancy: If server 1 completely goes offline all the virtual servers auto-migrate to the working server. If SwitchA goes both servers route through SwitchB. If the SAN controller 1 (ctrlr1) goes offline ctrl2 can route all the iSCSI data itself so there is no less there. Essentially we can loose a whole server, a whole switch and a full controller board (IE it's PSU and it's 4 port NIC and it's own dedicated motherboard and on board firmware) and then with the raid we can loose one of the 6 hard drives completely too yet still retain 100% efficiency with very minimal user downtime, something on the order of 5 minutes while each of the servers migrates. Due to the AD servers being on dedicated hardware sessions and logins are all kept so it is just simply things like loss of connection to the SIMS database for a few minutes then it catches up where it left off.
The hardware of the servers is powerful enough to run all systems on one set of hardware for (in theory) an indefinite time. This means we can install updates, reboot, change hardware etc.. with little to no impact on staff in the middle of the day. The servers run about 8-9 services including multiple SQL instances for things like SIMS, Sharepoint, SCCM, Biometrics etc.. so they are real workhorses. I can't remember the hardware models we have but I can check if you want. I know they are HP Rackmounts designed for virtualization.
There are really only a few threats to uptime, one is fire/flood/physical destruction of the room since the servers all sit in the same cabinet in the same room. The other is power, they do have two UPS's (again spliced so each server is in each UPS) but they are just there to prevent a hard shutdown, they wont keep the servers on for anything more than about 10 minutes and the last is a catastrophic software corruption. We do have backup for that situation of course but it wont be an instant restore. And of course if the rest of the network, EG a switch, goes down then those users loose connectivity.
Server 2008 Data center gives us unlimited virtual servers so as we did for the biometrics we can just partition up the SAN and get on with it, no worrying about licensing and what not which is actually a lot more useful than it sounds... With the biometrics the usual process is to install the server stuff on to the school's existing SIMS server, knowing how finicky SIMS can be we just setup a brand new virtual server that just hosts the biometrics. Same for printing, a stupidly small server but as it doesn't cost us anything it has a whole virtual server to itself. Really gives you some granularity when it comes to updates and backups etc... as each server is sandboxed.
The AD servers (of which we have two) are each on their own hardware. As they are the backbone of the network and knowing what windows is like it made sense and as it would happen one of our AD servers has been offline for the last 2-3 months until this summer when we are going to rebuild it. Again the hardware is powerful enough that we can provide a good level of service based on just one server. It's not been perfect but it has kept us going.
It cost around 30k including all the hardware and licensing. But that price is just for the 4 servers, two SANS (One for data which is user documents, files, DHCP, DNS and the second for the VDIs) and a cheap cabinet. We didn't touch the rest of the network and run our own cabling in the room itself and did everything ourselves. If you outsource any of it then you'll obviously pay for it.
25th July 2012, 01:05 PM #12
- Rep Power
thank you everyone the help and information it's been really useful. I'm currently putting the final case to the governors they are about 95% convinced so hopefully this will push it all over the edge. post back and let you know how you think is gone.
25th July 2012, 06:10 PM #13
By Lora in forum Educational IT Jobs
Last Post: 18th May 2012, 10:31 AM
By sujonbappi in forum Internet Related/Filtering/Firewall
Last Post: 14th October 2011, 02:57 AM
By jamin100 in forum Our Advertisers
Last Post: 22nd April 2009, 02:22 PM
By openhgs in forum Network and Classroom Management
Last Post: 2nd October 2007, 09:57 PM
By beeswax in forum Other Stuff
Last Post: 25th March 2007, 09:36 AM
Users Browsing this Thread
There are currently 1 users browsing this thread. (0 members and 1 guests)