A few questions about Linux server
 

Hi, I am new to server as I have always used Desktop and consider myself to be intermediate in terms of Linux, a few questions here, please forgive me for my little knowledge:

1). For a rack server such as Lenovo ThinkServer RD330, my configuration is Intel Xeon E5-2407 Processor (2.2GHz, 1066MHz, 10MB, 80W) x2, 2x8G memory, Primary RAID 0 and 2 HDDs, two 10Gb NIC: does this mean there will be two sockets (two CPUs), since this is quad-core, that means I will have 8 cores, am I right?
2). If I want to have a high performance server, for example, I buy 4 of them, then I will have 8 sockets with 32 cores, my question is: do I install RHEL on each one of them or I should use some kind of fast-connection switch and assemble all four of them into one super-machine with 8 Xenon processors? and one of the server will be the master server where I will install RHEL? since each server has 2 NIC, how shall I connect this thing to the internet? in that case, I will have 2x8x4=64GB memory
3). I am checking RHEL price, it says: 4-sockets with up to 4 virtual guest, $2398, so in my case, I will not able to use the machine because there are 8 sockets? or shall I buy two liecenses? also 4 virtual guest? this means I can only install 4 kvm vms? I certainly don't like this kind of limitations?
4). If I install VMware ESX5i, I plan to install 15 VMs, each with 2 cores, does that make sense?

Sorry, I have never even touched a server, so I don't know, any quick answer is appreciated!

Yes. The server has two sockets (according to the technical specifications), and if you buy a model with 2 quad-core CPUs, the system will have a total of eight cores.

There's no way to connect multiple machines to make them behave as one system. The closest you can get is a cluster (which is how supercomputers are created), but they are used for very specific tasks. Ordinary programs and services would not be able to utilize the computing power of a cluster, and would end up running no faster than if they were installed on a single node.

Whichever way makes the most sense. Rackmount servers are usually equipped with 2 or 4 NICs (plus one for remote management/IPMI), and you can use these as separate NICs, or aggregate them into teams if your switches support it. One common scenario is to connect one NIC to the LAN and another to a seperat ethernet segment dedicated to (typically iSCSI) storage.

You do whatever makes sense in your environment.

I'm no authority on RHEL licensing, but you will need at least one RHEL license per physical server. I don't know how RHEL deals with VMs, but why don't you just give them a call?

Sure, you can do that. I believe ESXi requires a separate license once you have nodes with more than 32 Gb RAM. Again, you can just give them a call.

A server is built around the exact same architecture as a regular PC. The main differences are:

• the motherboard is typically optimized for general I/O rather than graphics performance (the difference can be substantial)
• midrange to high-end systems usually come with an integrated SAS storage controller that can deliver significantly higher throughput than your average onboard SATA controller in a PC
• server usually have an array of internal sensors for temperature and voltage monitoring
• most servers have a dedicated subsystem for remote management, often (but not always) including a separate NIC
• redundant power supplies are common on servers
• server motherboards have a lot more DIMM sockets (the RD330 maxes out at 192 Gb, which is actually not a lot compared to similar models from other manufacturers), and they typically need fully buffered or at least registered ECC memory

I have to say I'm a bit puzzled by the fact that you go from never having touched a server to planning a setup like this. Have you considered hiring a consultant for a few hours? It could pay off if it keeps you from designing a solution you later find to be less than optimal for your needs.

http://www.atjeu.com/wp-content/uplo...1/IMG_3310.jpg

so the picture here shows 4 individual servers? this means I have to get 4 license of server OS? either Windows 2008 RC2 or RHEL 6.4? I always think you can combine them into a single one

Definitely I will hire a consultant!

..additionally:

Unless your setup specifically calls for using RHEL, and if you can (hire) support, do note there's also un-branded RHEL in the form of CentOS and Scientific Linux.


If you're solely looking at how to evenly distribute resources, yes, else (if different VMs have different purposes): maybe not.

I agree you should hire a professional. It may cost you money now but it may well save you money later on.