Upgrading HP ML330 G6 Server

Forums › IT Pro and developers › Upgrading HP ML330 G6 Server

miked

207 posts

Master Geek

Trusted

#168566 18-Mar-2015 14:57

I work for a charity, and am looking to add a second CPU to our main server. It is an HP ML300 G6 with 1x E5620 (2.4GHz Quad core Xeon).

Retail the processor is still possibly $2000.

Looking for better / more cost effective options as I could by a whole second hand server for that price on TradeMe!

I've seen better prices internationally on eBay, but does anyone know of suppliers in NZ who might be able to supply at good price? Presumably I need the genuine part for the heatsink? Anything else I'd need to know to make sure I bought a part that would work with this server?

Are the 1st and 2nd CPUs different in anyway (eg different heatsink, connector, or some sort of dongle/daughterboard required for the motherboard to make the 2nd CPU work)?

Assuming I need the genuine part, is that HP Part Number: 603976-B21

Any tips on how to get this part without the high price tag?

Thanks

Mike

1 | 2

3866 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1261767 18-Mar-2015 15:18

Yes. Here: http://www.trademe.co.nz/computers/servers/servers/auction-859764758.htm

“Don't believe anything you read on the net. Except this. Well, including this, I suppose.” Douglas Adams

Referral links to services I use, really like, and may be rewarded if you sign up:
PocketSmith for budgeting and personal finance management. A great Kiwi company.

Ruphus

465 posts

Ultimate Geek

#1261858 18-Mar-2015 17:30

That would get you the CPU but you'd still need the processor board (536624-001). Might be better off getting an ML350 which supports dual CPUs instead. Would a second CPU increase performance or would it be better to add RAM and/or better hard drives?

Dynamic

3866 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1261877 18-Mar-2015 18:31

You could also add an after-market SATA card and an SSD for a performance boost to your most important applications.

“Don't believe anything you read on the net. Except this. Well, including this, I suppose.” Douglas Adams

Referral links to services I use, really like, and may be rewarded if you sign up:
PocketSmith for budgeting and personal finance management. A great Kiwi company.

miked

207 posts

Master Geek

Trusted

#1262067 19-Mar-2015 00:14

We currently have it configured with Server 2008 running Hyper-V to Host 2x Virtual Machines (SBS2008 and Server 2008 with SQL Server).
The server is running fine, and appears configured well, but I'm wanting to add a couple more VMs, so planning on upgrading The Host OS to get a newer version of Hyper-V on there (to make better use of the hardware). Then add 2 new VMs:

One new VM running Server 2012 / R2 for our database application (currently running on the SBS2008 instance, but no longer supported on SBS2008 in new version).

The other new VM for ESET remote admin as a pre-packaged VM from them (Linux based image).

We have 12GB RAM in there already and 8x SATA drives (as 4x separate RAID 1 arrays).

I had been pondering about adding (or swapping) in some SSDs with this round of upgrades, but definitely will be bumping the RAM.

Equally the Machine isn't breaking a sweat ever, so could probably add both new VMs without any new upgrades just fine...

The RAM would be the most important upgrade though. Then probably CPU as the next most urgent. SSDs would be the finishing luxury treatment!

Thanks everyone.

miked

207 posts

Master Geek

Trusted

#1263600 19-Mar-2015 17:48

Ruphus, (& Dynamic)

I finally had the server case open today, and you are correct, I would need to add a processor board to be able to install a second CPU! That puts that idea outside the realms of being a "cost effective upgrade", unless I happen to find a suitably priced processor board second hand.

As you say, it probably means I'd be better off looking for a ML350 fully stocked as a replacement ready to go (either with a single E5620 so I can shift this one over, or with Dual CPUs of some sort already installed). I could swap over some/all of our disks (depending on how the second hand one came configured). There are a few on TradeMe with 15K SAS disks already in, which would presumably trump our SATA disks for speed. I could also then either keep our current ML330 as a backup on the shelf, or run it as a second unit in parallel, or flog it on TradeMe to recoup some of the cost of the ML350.

For reference, we have 6 DIMMS in there to make up our 12GB (unknown if they are UDIMM or RDIMM). If I wanted to bump our RAM quota with no processor add-on, that would give us 3 free slots (unless we have UDIMM if I am reading the specs correctly, in which case, we can't add any more?). However, if we get the processor daughterboard, we'd get another 9 slots of course.

I think I'm most likely to dabble with re-configuring what we currently have with no upgrades at all, and see what happens to performance. If we then decide we go the ML350 route, I'd rather not buy more RAM immediately, as most of the ML350s on Trade me have plenty of RAM included!

However in the short term, that leaves me tight for RAM (which is the most important upgrade), until I decide whether I'll need the extra CPU power, or will stick with this current box.

How much RAM should I be leaving for the HOST OS (Server 2008) running Hyper-V? Currently that has 2GB, as does the VM instance running Server 2008 with SQL 2008. The SBS2008 VM has 8GB allocated. Am I likely to get into trouble if I tweak those numbers to allow some spare RAM for a development VM?

Much appreciating the input. I'm much wiser on this beast's capabilities already, and the upgrade options open to us (and those not so readily open!).

Thanks

Mike

networkn

Networkn
32349 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1263602 19-Mar-2015 17:57

I have a bunch of ML350's left over from upgrades for clients. What is your total budget?

With what you are proposing I'd have at least 24GB memory, 16GB for SBS Alone :)

miked

207 posts

Master Geek

Trusted

#1263605 19-Mar-2015 18:02

Thanks for the offer. I'll flick you a PM to discuss (either tonight or tomorrow as I'm just heading out the door now).

Thanks

Mike

Equinox IT

Cloud spending continues to surge globally, but most organisations haven’t made the changes necessary to maximise the value and cost-efficiency benefits of their cloud investments. Download the whitepaper From Overspend to Advantage now.

timmmay

20575 posts

Uber Geek

Trusted

Lifetime subscriber

#1263609 19-Mar-2015 18:06

If your CPU isn't being pushed then just add RAM. I'd have thought 32GB would be a good interim amount of RAM, but it's not unusual for servers with many VMs to have hundreds of GB of RAM.

toyonut

1508 posts

Uber Geek

#1263610 19-Mar-2015 18:07

What RAM does the ML330 G6 have?
Something like this might be the go.
http://www.trademe.co.nz/computers/servers/servers/auction-860911427.htm

Check Ram compatability, you may be able to strip what you have out of the 330 and put it all into the 350. I have found the 350 to be a solid server and you can spec one into an absolute beast. If you are lucky, the CPU and RAM should drop right in instantly giving you twice the server in one unit.

Otherwise one like this:
http://www.trademe.co.nz/computers/servers/servers/auction-859997163.htm
is already specced out to more than you need.

Prices are pretty good on these servers second hand.

Try Vultr using this link and get us both some credit:

http://www.vultr.com/?ref=7033587-3B

Ruphus

465 posts

Ultimate Geek

#1263612 19-Mar-2015 18:14

I'm pretty sure the RAM and CPUs is interchangeable between the ML330 and ML350. Both servers run DDR3 10600R. I'm not so sure about the heat sink though.

Dynamic

3866 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1263847 20-Mar-2015 09:30

I've got a couple of ML350's here that I would offer, but they only have dual-core CPUs so it does not get you any further ahead. If networkn can get you some machines at a decent charity price, we might be able to help with some spare RAM or HDDs.

I'll respond to your PM tonight. Today is manic.

“Don't believe anything you read on the net. Except this. Well, including this, I suppose.” Douglas Adams

Referral links to services I use, really like, and may be rewarded if you sign up:
PocketSmith for budgeting and personal finance management. A great Kiwi company.

miked

207 posts

Master Geek

Trusted

#1273402 30-Mar-2015 13:33

Thanks for the help everyone.

We are now the proud owners of an ML350 G6 from TradeMe... and now it comes to setting it up!

However, the new server is SFF with only 3x 146GB 15k SAS drives. Our old server LFF drives and I hoped I could bring across the backplane cages to re-use these old drives in, but the cage racks are the wrong size to fit the server chassis. That means I'll need to source some extra SFF drives. To do that I need to make a plan....

The old server has the 8x drives arranged as 4x RAID1 Mirror arrays. These were allocated:

HOST (OS & Hyper-V install)
VHD for SBS2008 (OS, Exchange & Data)
VHD for SQL (OS)
VHD for SQL Data (Database Data)

This doesn't strike me as the best allocation, so looking to make a better plan going forwards, and then accordingly work out what drives I might want to add to the new server to make the new configuration possible.

My hunch is that whilst the above arrangement guarantees that no OS can be slowed down by any other OS drive access, it also means that there is no speed advantage from sharing spare capacity or from any drive striping. I am presuming that since the loads will vary, it will be better to allocated the drives into bigger, faster arrays, and then give each OS access to that array, that way whilst one system isn't using the drives much, then the others can use the spare capacity and get a speed increase. This might mean occasionally a small delay is induced during heavy drive access, but the rest of the time everything should be faster.

However I don't know if that is true, or whether my logic is flawed?

It also means that increasing or decreasing the number of Virtual Machines running on the server instantly de-optimises the setup until everything is re-jiggled around.

I'm thinking of one or two big fast RAID 10 or RAID5 arrays with the VHDs stored on them. What would you recommend?

Mike

Dynamic

3866 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1273406 30-Mar-2015 13:43

RAID5 writes slowly. We mostly use RAID10 these days.

“Don't believe anything you read on the net. Except this. Well, including this, I suppose.” Douglas Adams

Referral links to services I use, really like, and may be rewarded if you sign up:
PocketSmith for budgeting and personal finance management. A great Kiwi company.

miked

207 posts

Master Geek

Trusted

#1273585 30-Mar-2015 17:51

Yeah RAID 1+0 is my preference I think. But depends how many disks I can fit in there and in what configuration... I suspect there could be quite a few scenarios where RAID6 works out better:

RAID6 array on 5 disks would equal 3x drives of data and 2x drives of parity for parallel processing. To match/exceed this in Raid 1+0 would need 6x drives for 3x drives of data, and 3x drives of mirroring. eg one cage slot extra.

Equally if I compared the 5x drive RAID6 with 4x drive Raid1+0, I would expect the RAID 6 to beat it in most scenarios of read and write speed (due to the extra striped drive parallel write).... even with the slower write performance overhead of RAID5 or RAID6.

If I had 6 free cage slots, and could run RAID6 on it instead of RAID 1+0, it would give me 4x data drive stripes and 2x parity.... compared with only 3x data stripes and 3x mirror stripes on RAID1+0.

So if I want to maximise throughput and I am limited on cage space depending on how I allocate, RAID6 might be faster overall for a given number of drives even allowing for the write performance hit.

Like say I do 6 drive RAID6 for most data and a 2 drive mirror (for relatively little used data). My hunch is that might be faster setup than a RAID1+0 across those same 6 disks (also with a 2 drive mirror).

Equally I could do a RAID6 array on 5 drives, and a RAID5 on 3 drives.

Of course, if I'm not trying to get 2 arrays in there, then I could do an 8 drive RAID1+0 (4 parallel data stripes) vs 8 Drive RAID6 (6 parallel data stripes).

There are other advantages / disadvantages to running RAID6 compared with RAID1+0. (array rebuild time

I guess actually the main question I need to answer is: Are multiple separate arrays better or one bigger faster one for multiple VMs? If it depends on various factors, what are those defining factors?

This will then allow me to work out how I will divide up my space (cage slots), and therefore work out which RAID spec to go for.

Dynamic

3866 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1273841 30-Mar-2015 22:31

More spindles reading at the same time = more read speed. This makes RAID5 one of the quicker arrays for reading. Writes are slower because of the checksum calculation.

RAID10 has greater write speed at the cost of sacrificing half your disk space. RAID10 also gives you greater redundancy, as the more pairs of drives you have, the greater the chances of surviving a multiple-disk-failure scenario.

For ourselves on an 8 disk system we have a mirrored pair of SSDs for our virtual host and our primary production VM that runs our client management system, and a 6 disk RAID10 for lesser-used stuff. Prior to this we have been running systems on multiple older servers with single SSDs and have been fortunate enough to have no failures.

“Don't believe anything you read on the net. Except this. Well, including this, I suppose.” Douglas Adams

Referral links to services I use, really like, and may be rewarded if you sign up:
PocketSmith for budgeting and personal finance management. A great Kiwi company.

1 | 2