you need a faster cpu for sure as mysql is maxing out the current cpu.

I'd first look at your mysql statements and tables and find out why mysql is maxing out your cpu (unless you're currently running on a P3 or something)

Ok.. So you are saying no on the Quadcore.. Go with the Xeon with the SCSI option..

Even though the current server you are seeing these results on are already a Xeon with SCSI.. I can forsee about the same reslts as what I posted above?

No?

Current server you are seeing these screen shots from is a

currently on:



Dual Xeon 2.4 Ghz
4 GB ECC REG Memory
73GB ultra320 15k SCSI HDD1
146GB ultra320 15k SCSI HDD2

Hey Phil..............

How is going from what it is currently on an:

Dual Xeon 2.4 Ghz
4 GB ECC REG Memory
73GB ultra320 15k SCSI HDD1
146GB ultra320 15k SCSI HDD2


to upgrade to

Dual Xeon 3.0GHz w/HT (2x 1MB L2 Cache)
4GB DDR ECC REG RAM
146GB 15k SCSI HDD1
146GB 15k SCSI HDD1


going to make much difference in a issue like this?? Wouldnt the 4 quads better address the issue? Seems like you are reconmending more a lateral move no?

http://www.boneprone4life.com/normal.jpg


I dont see much a improvement WITH WHAT HE ALREADY IS USING if he goes with the Xeon option. Its sort of what he already has.
Please explain.

why not get a quad core with scsi 15k drives ?

cant........

Have np idea, i am not a tech guy

get more 10k rpm hd's in the dual or quad core machine then and do raid 5 :)

I would say server A for the same reasons as everyone else above... HD is almost always your biggest bottleneck, there's no reason for most servers to have the latest and greatest processors with the most marketing behind them, when 2-3 year old workhouse technology is just as good when combined with SCSI drives.

I agree.. But have you seen my screenshots??

Those are with a SCSI HD server.

Dank,

Because the server is not a quad core :) If it was, I would say yes.. you'll likely get better performance with it according to the limited information before me. Again, before I'd feel comfortable offering a prospective customer an upgrade solution - I'd really like to be able to see output of iostat and vmstat at bare minimum.

Server option "B" is a single processor, dual core for a total of two CPU cores @ 2.4Ghz. Server option "A" is a dual processor, single core setup for a total of two CPU cores @ 3.0Ghz. The Core processor is better, but not by a mile. I would go as far to say as you'd see fairly similar performance characteristics in the real world given the two options, with the Core beating the Xeon by a tad.

And yeah, ideally here you'd simply go all out and get a server w/ both disk and CPU. Something like a Dell PE2950 or similar (e.g. 4 or 8 CPU cores total, and SCSI disk). Granted, that's expensive.

And the difference between an old 2.4Ghz Xeon and a 3.0Ghz will be decent. The CPU's on the two machines listed are "roughly" similar, so I'd go with what has the better disk subsystem.

-Phil

To reiterate what Phil21 has said... double check with the tech guy again to see how many processors you are getting.

server A: 2 cpu x 1 core each = 2 cores
server B: 1 cpu x 2 core each = 2 cores... or is it actually 2 x 2 core? (dual - Core 2 Duos)?

I havent seen a server run on 1 CPU in awhile.. maybe we're spoiled here at Silvercash =p

core2duo's are a desktop proc and can not be dual.

your screenshots don't help much thats just your current machine load. however your mysql is only using 1.9% of your memory.. you might wan to increase that.. i.e tune your my.cnf file

run the benchmark tool and you'll get a score the score will tell you everything you need to know.

the equivalent of a core2duo e6600 is a xeon 3060 and it is the **exact same** cpu, there is absolutely no difference. intel rebranded it and increased the price.

just fyi

If I got server B, I would do the upgrade the sales guy mentioned to it. Not sure if that helps a lot or not.

"ability to upgrade to Quad-Cores later (like Core 2 Quad Q6600:
4x 2.4GHz, 8MB Cache). "

also dont forget first server is probably better quality hardware second machine is problably a low end desktop board thats why scsi is not possible where the first machine is a server board.

its probably not scsi because the case can't fit the SCSI card... or because its matched pricing.

either the older dual xeon which they have in stock for a while but with scsi 15krpm or the new (so they have a capex) core2quad with raptors

Sales guy:

Upgrade to Core 2 Quad Q6600 (4x 2.4GHz, 8MB Cache) is $149 one time
setup.
Wow, if you upgrade to that Q6600 and setup RAID10 (4x 37GB 10k Raptor
SATA HDDs), your server would be a monster mysql server. :)

yup, raid-10 raptors would out perform 2 single scsi 15k rpm

fyi, the difference in cost between core2duo e6600 and core2quad q6600 is like $60

Ok.. All the way up till now my programer has been insisting on the SCSI Server A option..

Now he's changing his mind.. LOL. I think he likes the 4 hds option and the raid..

"Hmmmm. Thats intresting
37 gb will be 37 * 2 gb for one volume

and we will have 2x 37gb hdds for mysql & system boot

and 37*2 gb for thumbs maybe.. Hmm. This could be good"

you dont need 2 hd's for thumbs i have done 80mbs on a celeron 2.4 with a normal ide hd using less then 20% cpu and less then 200mb of ram.

Thumbs on one hd and shit on the other.

dont do what he is saying, then it wont be raid 10.

get 1 RAID-10 array using the 4 x 36GB Raptors and you will smoke the Xeon in HDD speeds. The RAM and CPU is already better on Server B.

Dont do what who is saying?

your programmer. splitting it up like he wants will not get you better performance than a RAID-10 array

if you can get:

Intel Core2Quad Q6600
[ 4 x 2.4Ghz, 8MB L2 Cache ]
2GB DDR2-667 RAM
4 x 36GB Raptor 10,000RPM SATA
Hardware RAID-10


Then this server will kill the Xeon in all factors, and this discussion is over. Server B, with the above specs is the winnar.

What is the price they are charging you monthly for the above server? how much bandwidth is included? just wondering

500.00 per month with 30mpbs

Non cogent

The salles guy and my programer are now arguing. My programer doesnt want raid 10. Sales guy is explaing to him we do.

""RAID0 is not something we should do. stick with your original plan or SERIOUSLY consider RAID10. It is
best for your server and will actually out perform. It not only gives you fast I/O disk, but also
protect you from hard drives failure. However, if there are some reasons
that you can't go with RAID10, you can go with 2x single drives without
RAID cause RAID0 is not something we should do."

Programer is saying "raid10 can't match the speed of dual raid 0"

thoughts?

raid 0 is good if you make daily backups and dont worry about hd failures :)

4 drives in raid 0 would rock :)

But it would be a little risky :)

guys

raid-10 IS A STRIPED AND MIRRORED ARRAY. RAID-10 is the same as getting RAID-0 but also getting the benefits of RAID-1.

So you get striping and mirroring.

Your programmer is clueless on servers, he should stick to programming lol

RAID-0 is stupid when you have RAID-10 available as an option.

No sandman. this is what my programer is sayin.


"dude

not 4 drive raid 0

Im saying
two seperate raid 0 volumes

2x hdd raid 0 + 2x hdd raid 0"

boneprone, raid-0 is the most risky raid you can run.

whatever drives you put into raid-0 array will act as 1. so if just 1 drive dies, everything is lost.
this basically doubles the risk of just running 1 drive because you have the liability of both.

the only time to do RAID-0 is if you want speed but dont care about your data and downtime WHEN a drive fails (not if).

RAID-10 will give you both the benefits of RAID-0 (stripping) and RAID-1 (mirroring) in 1 package. the only reason people dont normally do it is because it requires 4 drives minimum and they dont have the budget for it.

and FYI, having 4 x hdd raid 0 would be faster than 2 x 2hdd raid-0

Yeah... You don't run RAID0 in a production environment unless one of these case are met, and even then you very much should question that decision. RAID0 exponentially increases your chances of a filesystem failure as you add each drive (e.g. 4 drives = 4 times likely to suffer data loss than a single drive).

The cases I would see as acceptable..

1. Scratch space. Meaning temporary disk storage space. Only there for doing things like resizing mpegs or whatnot. No data loss if it fails, and application should be smart enough to know it failed and not try to utilize it.

2. Very specific hardware load balanced setups, where individual machine failures are normal and expected.

4 drive RAID0 *will* be faster than 4 drive RAID10, but not by very much. You still will get roughly "4 times" the performance of a single drive in either case, minus overhead. The speed difference is negligible, and the chance of outage is guaranteed with RAID0. RAID10 you can sustain two drive failures, if they happen to be the correct drives. The single advantage RAID0 will give you, is the size of your array. (RAID10 will eat half the space in "overhead" - RAID0 has none)

Your sales guy is right, listen to him :) Anyone recommending RAID0 for a production server simply does not have any experience in the matter.

-Phil

he needs to stick to programing i was joking with the raid 0 on 4 drives 2 raid 0 arrays is kind of stupid makes no sence at all.

what i dont understand is why you want to serve thumbs from the same box much easier to unload thumbs on a seperate server.

Why exactly are you getting a new server?

Based on the screenshots, the current server looks to be doing the job just fine. One should be wary of "Top" stats - they can often paint a misleading picture of the situation.

Is the load average spiking above 2 a lot? and/or warnings/errors are showing up in the logs? and/or the server is often unresponsive? and/or are you expecting that server utilization to greatly increase very soon? ... if all no, then my view is leave things alone and save the money for something else... run with what you have for as long as you can.

Ron

You are no fun...

All sales guys boo you.


Come on guys... Say it.. "booooooooooooo"

I would tell your programmer to go and optimise the sql queries and tables. If mysql is constantly maxing out cpu on your current config, then there's something wrong. If the cause for this hogging is pretty much table reads, you need to start doing some table archiving so mysql can use the cache for the majority.

For example, I've a mysql dbase with 16 tables where one contains over 6 million rows, and another 1.2 million rows. The webserver mainly does db writes, which causes no spikes at all. When the db is being read, it's to pull out stuff based on days/weeks/months etc. Turn the mysql cache off and the server cpu is hogged by mysql. So, anything < today is archived, as this is never going to change by the db writes. Hence db reads for anything < today either use the cache, or very well optimised keys. This means I can hit mysql with a ton of reads and never hog more than 1% cpu. DB reads on "today" still caused big spikes, so all I did was inactivate reads on "today" and roll the website to be 1 day out-of-date. To the user, it doesn't make any difference, as yesterday's data becomes today's data, and they're none the wiser - all relative.

Like I said, dunno what you're doing, but you've gotta be able to hone your sql script somehow so as to not hog cpu like that on your current setup.... mysql optimisation isn't trivial, but it's always possible to keep on tweaking to shave a few cpu cycles.

between sata and scsi, i prefer sata in raid. unless you are going with SAS. :)

looks like i overlooked those stats. between server A and B, i'll go with A, because of scsi. unless you are running raid-10 with 15k rpm drives. if you have some money and time, try SAS. :)

check out this link: http://en.wikipedia.org/wiki/Serial_Attached_SCSI

if your hdd is doing swaping memory all the time, got more ram. :)

RAID-10 w/ SATA Raptors wins over the SCSI's, we've already established this

good thread, I learned a lot.

I'd go with A, but I know nothing so ignore the Fred

Ok after three days of Discussion this is what we are going for.

1U SuperMicro (brand new, 4x hot-swap bays)
Core 2 Quad Q6600 2.4GHz (quad-cores, 8MB Cache) - $0 setup(waived)
4GB DDR2 ECC REG RAM - $0 (waived) setup
150GB 10k Raptor SATA HDD1 - $0 (waived) setup
2x 36GB 10k Raptor SATA HDDs - $99 setup
30Mbps Dedicated Unmetered Port
APC Remote Reboot Port
DDoS Protection
MRTG from switch


A three HD system.

My programer does not want RAID10. Its like it is a Sin or something.

Seems stable this way.
And a great price. So no SCSI and No Raid10... LOL. Talk about not going with the masses!

programmers taking you down. a programmer codes software, you cant rely on them for system administration, they are two different related technical fields.

Good choice Boneprone - whats the monthly rub on that? I was hoping you wouldn't do server A because it was pretty much a lateral move... and server B wasn't all that either - server C was the right option the whole time, the only way to have bettered that would have been to go with SAS drives.

With regard to the comment earlier that the Western Digital Raptors are consumer drives thats not true at all. They are enterprise quality and come with a 5 year warranty. The only reason companies like Dell don't put Raptors in servers, or at least even low or mid grade servers, is because they like to sell more expensive drives. Additionally, they never came up with a compelling offering of larger sizes... like the 300gb and now 500gb SAS drives. If you read performance reviews on http://www.tomshardware.com it's an excellent reference for processor comparisons and HDD benchmark performance. To answer the other earlier question, all Raptors are Sata 150 none are 300.

Also, a side note, the 15k drives aren't all that they seem at face value. Manufacturers like to let you just simply assume that they up the performance ante by 50&#37; when that's not at all the case. They typically benchmark out only slightly faster than the 10k sisters... and depending on the application, manufacturer model, you're going to see really varied results.

Brad

Thanks my man..

Its going to be 489/month for 30mpbs. Non cogent..

This is not my primary server which I of course have hosted at Jupiter (two servers there level 3) or even my secondary server which I have at reflected.. This is a script server that hosts some thumbs (non crucial ones at that) in which I need something that will be fast and not lag while processing my network's custom scripts. Yes they may not be the most efficient scripts but they were custom made to my needs and desires so naturally they may use some resources. Thats why im making this move.

I hope we made the right Choice. Seems there was tons of advice on going with RAID. My programer kept insisting it was not needed. He licked his lips some on the SCSI drives but he said in reality the three SATA raptors would be just fine spreading things across them. And with the Quad cores and 4 GIG Ram he was excited to get started with optimizing the scripts.


Hope all is well Brad.. See you in Vegas I hope!

Mojo 4-Life

Hey I want to give some serious props and thanks to all the people who helped out on this tread. It was very very very educational.. Even though I went with what my programer insisted on I still learned a lot..

Huge props for all you hosting people who posted here as well. And its extra classy none of you even spammed! It would have been perfectly fine if you did ya know!! But Ive taken note as im sure many others here have of this thread. Some of you who put their 2 cents in have been noticed and very helpful. Ill be sure to make a note and remember you all in the future.

Great thread..
And good to see there are so many well informed hosting options out there now days.
BP4L

Go with server B the dual cores and the SATA with the DDR2 ram will preform better.

Later,