Mmmmm Upgrades.......

[warmadmax]

damn... don't think the boss would shell out for that
would love to play with that sort of kit,
ours is running 2008 R2 ent. as well, local bog standard raid + 2.5" drives plus a NAS on iSCSI for extra storage.

only this year upgraded everything up to gigabit! (though servers have got ether channel on the go, 2x1Gig )


your i7 extreme is the "premium" end, there isn't much in it but at end of the day you've still got 6 cores, 2 up on anything else including ivy
only thing you'd gain going to Ivy is some wattage saved so don't fret

[BillyGoat]

I'm not bothered about burning a few extra lumps of coal if it means I get my work done much faster!!

Mmm, only thing I keep spotting when I'm out at customer sites is their RAID performance is p*ss poor. Seems most low-med servers come with one of the PERCi controllers and NONE of them come with a BBU. No BBU=disabled write cache=TERRIBLE performance.

There was some building going on that involved moving stuff around the drives (around 6GB per test) into build folders. Without BBU and WC disabled (automatic too) it was around 45-50 minutes!!! With, it dropped to 5 minutes!!!!!!!!!!!!!!!!!!!!!!

Why hardware partners don't make this clear to people who buy stuff is beyond me.

I suppose everyone doesn't need performance in the disks or have external RAID cards......?

BG

[root]

hmmm grumble grumble...... going off to rant to my boss

[warmadmax]

know what you mean

i've always specified battery backed cache, even seen some raid cards come with flash drives for storing the buffer in a power outage.

i've gone away from running raid 5 / 6 as it's just too slow from past experience,
easier just to add a disk or two and set it as raid 10 (raid 1 + raid 0)

reason they don't make it clear is it lets them put it on the options sheet and make the headline price cheaper,
if you are using purely external storage it is an unnecessary expense i suppose

prefer me HP proliants over Dell poweredges as well

(our poweredge 4400 had more drives fail than a little!, soiled reputation for me on a critical piece of infrastructure!)

[BillyGoat]

I think it was the DL380's and round those that use the Perc controller. All as bad as each other.

Problem with raid 10 is the ability to increase size. The system that the sites use requires storage and lots of it.

One of the guys I work with have this as their storage subsystem....

P400 SAS controller with 4x73GB 15K SAS drives in RAID 10 for the host OS (and a few other bits)

Storage is P800 SAS controller MAS 60 with 12x300GB 15K SAS drives in RAID5 (multipath)

Some of the bigger sites have massive cluster setups, the tools have desktop PCs. Take it as it comes I suppose.

They sit in the middle of some of the places I visit....but it just about gets them by. It's fast though Perhaps Root can borrow a few of their hot swap spares....

[warmadmax]

thought PERC was just a dell controller ,
just called "Smart array P400" in the DL's we've got

we've got 7x 15K 146GB drives for the DB on one controller (6 raid 10 + hot spare) and 3x 15K 146GB drives for the host OS (raid 1 + hot spare)
think they're DL370s, room for 16 drives in 2x8 bays

other one's just got a single bay full, runs our exchange server nicely.

they deffo need the space if they've got that many in RAID5, they do know it'll only cover them for one disk failure?
more disks you use, higher chance of multiple failures,
thats one of the other reasons for the RAID10, it's mirrored stripe, up to half the drives can fail, worst case 2 before the array fails.

for added paranoia i've put hot swap drives on

[BillyGoat]

They have another storage facility to back all their stuff up, so no problems with loss.....something most partners seem to have no shortage of :)

It's the data that's important for us, not the servers - they can be rebuilt very quickly.

BG :)