Link to home
Start Free TrialLog in
Avatar of TheSonicGod
TheSonicGod

asked on

What Raid Set/Drive Size is best for a new 2008 R2 x64 Data server & Esxi or local

Hi Everyone,

Just wanted some opinions for what raid set is best for a new Data server I am building for a 15 user office.

I have a new HP Proliant DL 360 G7 with 8 Drive bays and 16GB of Ram I am replacing a current 5 year old Data server with 1TB Raid 5 array (4 drives) and 4GB of Ram. The current 1TB is about 60% full

I want to get the best disk performance I can but I need to keep as much space as I can to grow since it is a data server

Here are my options in the price range I have:

option 1:

8x500GB SAS-2 7200rpm
1) Raid 10 - 2TB storage and min 1drive redundancy

2) Raid 50 - 3TB storage and min 1drive redundancy

3) Raid 60 - 2TB storage and min 2drive redundancy

option 2:

8x300GB SAS-2 10K RPM

1) Raid 10 - 1.2TB storage and min 1drive redundancy

2) Raid 50 - 1.8B storage and min 1drive redundancy

3) Raid 60 - 1.2TB storage and min 2drive redundancy

option 3:

8x1TB SATA-300 7200 RPM

1) Raid 10 - 4TB storage and min 1drive redundancy

2) Raid 50 - 6B storage and min 1drive redundancy

3) Raid 60 - 4TB storage and min 2drive redundancy

I am also trying to decide if it is worth it or if there will be any performace lose to have Esxi installed on the server them build the 2008 R2 x64 server as a VM or just load windows on bare metal

Please let me know your thoughts and thanks in advance for your replies,

TheSonicGod
SOLUTION
Avatar of David Spigelman
David Spigelman
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
SOLUTION
Avatar of Andrew Hancock (VMware vExpert PRO / EE Fellow/British Beekeeper)
Andrew Hancock (VMware vExpert PRO / EE Fellow/British Beekeeper)
Flag of United Kingdom of Great Britain and Northern Ireland image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
for 15 users, your likely overengineering your solution.  while I agree with everyone that mirroring is faster than RAID5 (or it's derivatives)... at the end of the day, it's only 15 users!   Get SAS drives, put a single RAID5 in  (with monitoring so you know if a drive fails) and call it a day.

The 8 drives at 7200 will outperform the 8 drives at 10k, but you are sacrificing performance for capacity.  I'd go for the 7200 for now...

I'd take a pair of drives, RAID1 for the OS, create the rest as a RAID 5 array.  If down the road you need more space, use the HP array utility to do online array expansion, and buy the new, bigger, faster drives 2 years from now and expand the volume.  No brainer :)
What sort of Data are you storing?  How often and frequent is the access?  Those are what you need to answer before you determine your RAID system.  Databases have different storage needs than just having lots of small files or a bunch of large files.
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Avatar of TheSonicGod
TheSonicGod

ASKER

Thanks everyone for all your responses and I have reviewed everyone's replies (a couple times now).

I am going to try to answer everyone's questions and ask a couple more:

1) d0ughb0y - thanks for your reply - yes I assumed that 2.1 was the fasted at the cost of Space - but is it overkill for what I need given the 15 user office and if I do not use VMware?

2) Thethical - thanks for your reply - and thanks for the idea of having a hot spare versus using all of them in the Raid configuration

3) hanccocka - thanks for your reply - and thanks for the battery idea - that is something I will need to investigate for the current specs and add to the card if required. We have 2 physical servers running 4 current VM servers so I thought for flexibility purposes building via Esxi would be a good idea.

4) SysExpert - thanks for your reply - and thanks for your 4 drive idea as this may open up additional options. I was so focused on as many drives as possible that I did not consider the less drive more space option

5) FastFngrz - thanks for your reply - I think you might be right here that I am over engineering this solution. So I think you are saying faster drives with Room to grow - i.e. try to build a 2-3 year solution versus a 5-10 year version is the best approach? Also do you have any documentation on the HP array expansion - I would really like to look into this as perhaps I will build this option into my future plans for this server

6) serialband - thanks for your reply - the data we are storing is mostly office documents and PDF. Most of this data is medical research data and patient information there is a ton of data but the individual file size is not very large - usually each user may have 15 to 20 documents each open at once and could be editing all of them simultaneously at once

7) andyalder - thanks for your reply - I am now aware of the raid change on the fly possibility - do you have any links for this. Also, I have not been able to find the $300 license that you refer to - is it possible the unit I have shipped with this included for Raid 6 as it is an option in the cards Bios config

8) millardjk - thanks for your reply - Are you advising that 16Gb of Ram is not enough for a 15 user data/file server? They have other servers in their network handling their exchange and AD.

The additional questions I have for everyone is: Is there a significant performance gain from using 10K SAS versus 7200RPM and how big a loss is it to drop to 7200RPM SATA from 7200RPM SAS?

Thanks again for everyone's responses,

TheSonicGod
I noticed this post because I am having a similar discussion around my SQL transaction server. I have to strongly disagree with the thought of using 7200 RPM drives; be they SAS or SATA. The performance is horrid as transactional load increases.

For SQL 15k SAS drives are non-negotiable. 10k performance is fine until the load increases. But if budget is a problem go for 10k - I've been running a server with 10k drives for 3 years and it is only now a problem with the DB sitting at over 100Gb. SATA - that is asking for trouble.
Battery-backed cache is really good to have. It costs the equivalent of a single 15k drive and allows the drives a little breathing space when the load is on.

My ideal SQL server (the one I am going to buy today) has 8 drives; 4 pairs in RAID 1; OS, Data, Logs, TempDB. Battery BackUp Cache will be as big as I can afford. RAM will be what I can afford.
If you are going with a DL360 get the SAS-Extender (preferably now, or when you need more space) and that allows you to add a further 8 drives. You can then add mirror pairs and spread your data files over multiple RAID 1 pairs with no downtime at all.
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
I'm not saying that 16GB isn't enough; you indicated that you plan to install Windows 2008 R2, which has a recommended 4GB minimum, and 8GB gives nice performance for most of the built-in features. No: I'm saying that this OS will give you better file server performance with higher RAM because it will use excess RAM as cache for the file system.

Your later post indicates that the handful of users (15) having a handful (15-20) of files open at once gives a potential for 300+ files to be simultaneously in cache for read/write operations instead of physical I/O. Upon first opening a file, it will come from disk at the speed of that subsystem, but further operations can be at cache speed if the server memory isn't being constantly paged out for other duties.

One other thing strikes me about your use case, tho. If this research data is being edited on a time-bound basis: new data comes in, users scrub the data, enter it into another system for statistical analysis, etc., and then the files are then stored more-or-less untouched for archival purposes, you might consider two tiers in your server.

If you get a pair of SSD drives (RAID 1) for the most recent stuff, along with my previously recommended 6-spindle setup for completed work, you can use the Scheduler to automate a batch script to move files from the super-fast tier to the high-capacity tier as time goes by.
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
According to http://h18004.www1.hp.com/products/quickspecs/13598_na/13598_na.html, the model of G7 will determine default RAID controller memory.. something you want more than 0 of!

The Array controller at http://h18004.www1.hp.com/products/servers/proliantstorage/arraycontrollers/smartarrayp410/index.html supports array expansion, see http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c00729544/c00729544.pdf

I'm a big fan of HP servers - the idea that you can upgrade all the drives in an array with a larger capacity drive (one at a time) without downtime and then take advantage of the additional space is just COOL!
ASKER CERTIFIED SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Thanks Again everyone for the replies - a lot of valuable information here that I hope will be useful for others too.

I think from what everyone is telling me I have concluded that this may be the best options to do:

1) 2x500GB 7200rpm - Raid 1 for C; drive - total space 500GB
2) 5x600GB 15Krpm - Raid 5 for Data drive - Total space 1.8TB (with 1 hot spare for 2 drive redundancy)
3) Add 512mb with battery for HP Smart Array P410i
4) 16GB of Ram should be good enough

This setup would leave me an extra drive slot to add a drive if required and because of the HP controllers ability to upgrade on the fly if required I could always add larger drives to the Raid 5 array as per FastFngrz/millardjk suggestions - I still have future-proofing ability

And to clarify - this is a file and print server (data server) serving documents/PDF/media files - it will not run SQL databases / Exchange database etc. - sorry for the confusion if something else was implied

Please let me know if you think this is a adequate configuration

thanks again,

TheSonicGod
Just a file server???  Really? Really??  Throw a handful of sata disks in the box, raid 5 'em and call it a day.
At the size of the documents we are dealing with 500 to 4000 page PDFs with detailed images from medical researchand and the number of files open and being edited at once 10 to 15 per user, plus throw in 15 users 4 to 8 gb archive pst files all open and being searched - I am not sure you can call it 'just a file server'

My comment above was advising that I will not be running any SQL or exchange or other databases from it (or at least that is the current plan) so technically it would be called a file server but it's load I think is a lot larger than an average file server would be with an office much larger than this one

Hopefully that explains it better

Thanks

TheSonicGod

Sorry for the snippy reply, :)

The I/O behaviour of a box serving up large files is significantly different than a database server doing 4k random page I/O... In your case, even though the files are large, it's only 15 users, (15 users x15 files per=225) so the likelihood of having thousands of simultaneous files open is slim.. Your best bet would be to make sure to:

1. Ensure you defrag on a weekly basis (default with Server 08)
2. Optimize for read's over writes (I assume a greater than 3:1 read:write ratio that is typical)
3. Ensure you have a FAST antivirus solution as to not dog the server down (ie: Not Symantec!)
4. Make sure your Outlook clients have "instant search" turned on, so they're not scanning the archive PST's off the server for search (Assuming you are using Outlook 07/2010)

Cheers!
Huge medical images? You might want to look at the specs for the HP Medical Archive system to get an idea of what they use for the hardware, note the use of midline SAS disks (which are basically SATA disks inside the assembly) on some storage nodes - http://h18000.www1.hp.com/products/quickspecs/12179_div/12179_div.pdf
Yes, huge medical images, some inside or embedded into the PDF but also the raw images like the inside of your Liver for example in a resolution that can be poster size without degradation

These reports are created for submission to the FDA or CDA or ???? and the images can not be compressed as other doctors/scientists and medical people need to view and manipulate the images as requires so these files are very large and so are the PDF's and other docs they get inserted into.