Replacement server :A G5 to a G6?
- Thread starter Destination
- Start date
- Joined
- 31 May 2009
- Posts
- 21,457
cpu (no issue)
disk queue (which got a slap also)
memory available (was always fine)
The database is SOE Exact database, so I am unsure what subset it belongs to, we do have mysql as part of 2003, but I don't know what DB Exact uses
disk queue (which got a slap also)
memory available (was always fine)
The database is SOE Exact database, so I am unsure what subset it belongs to, we do have mysql as part of 2003, but I don't know what DB Exact uses
The last time I dealt with SOE I had a major performance issue due the anti virus software scanning the database which if I remember correctly was SQL (it was about 8 years ago). Once I excluded the DB from scanning the server performed okay.
With regard to the quote - I think it is quite reasonable. After all the suppliers do have to make a few £££ so there is going to be margin on all the equipment. £500 per day is about standard for the industry at the lower end of the SME market. I use ML350 G6 as a base for running VMware (Essentials Kit - VS4-ESSL-BUN-C) using the Premium version of SBS2008. This permits multiple OS on the same hosts, so we separate the SQL from Exchange/Files Store. We also run a virtual firewall within the same host.
I would STRONGLY suggest adding 534916-B21 HP 512MB Flash Backed Write Cache for P410i, otherwise performance is going to be strangled.
With regard to the tape drive I would use an LTO 4 rather that 3. I would assume that the 4 x 300GB will be running RAID 5 permitting 900GB usable. An LTO 3 won't (probably) cope with more than 600GB compressed.
CA software is supposed to be really good but we don't use it - we use Symantec BUE for SBS which has all the agents included. I would also recommend Symantec System Recovery as it will restore the server can to a previous state with ease. It can also recover the server as a VM which is a real bonus in a DR situation.
Don't let anyone talk you out of the managed switch. With that amount of clients a managed switch is vital to support the network.
I hope this helps.
With regard to the quote - I think it is quite reasonable. After all the suppliers do have to make a few £££ so there is going to be margin on all the equipment. £500 per day is about standard for the industry at the lower end of the SME market. I use ML350 G6 as a base for running VMware (Essentials Kit - VS4-ESSL-BUN-C) using the Premium version of SBS2008. This permits multiple OS on the same hosts, so we separate the SQL from Exchange/Files Store. We also run a virtual firewall within the same host.
I would STRONGLY suggest adding 534916-B21 HP 512MB Flash Backed Write Cache for P410i, otherwise performance is going to be strangled.
With regard to the tape drive I would use an LTO 4 rather that 3. I would assume that the 4 x 300GB will be running RAID 5 permitting 900GB usable. An LTO 3 won't (probably) cope with more than 600GB compressed.
CA software is supposed to be really good but we don't use it - we use Symantec BUE for SBS which has all the agents included. I would also recommend Symantec System Recovery as it will restore the server can to a previous state with ease. It can also recover the server as a VM which is a real bonus in a DR situation.
Don't let anyone talk you out of the managed switch. With that amount of clients a managed switch is vital to support the network.
I hope this helps.
Just looked up that software "SOE EXACT" and it doesnt mention what database sits underneath it. From what aspirin says above it could just be a flat file type database. If it is then ensuring your virus scanner doesnt touch that file would be important.
- Joined
- 31 May 2009
- Posts
- 21,457
Thanks for the assistance.
We had previously identified lack of meory as issue, back 3 years ago, and we added more to the current server to bring it from 1GB up to 3GB. Then after 18 months usgae we identified as the DB got larger issues from the symantec virus checker searching the DB. The antivirus was then instructed to stop scanning the DB.
Now we're hitting the wall again, and I ponder if adding more RAM will help, we are seeking clarity if i can just buy matched pairs and add a GB myself, or if this will invalidate warrenty and support licence.
I know our currently drives are beginning to be low on space.
The main OS/programs and EXact and DB are based on 120Gb mirrored RAID, whilst our scans and radiographs were removed to a second mirrored RAID allowing 500GB storage.
The exact details of the new disk setup was yet to be discussed with the hardware folks who were recommending us the package. 4*300GB is unlikely to be suitable for long term needs unless they have something else in mind as radiograph storage.
I am assuming that ' SAS Small Form Factor 10 000 rpm 6GB/s Dual-Port' is a fast hard drive solution, but wondered regarding the entire tape backup system, and wonder if removal hard drives swapped each night and taken offsite would prove a more suitable method of backup, and if such a system would be faster?
We had previously identified lack of meory as issue, back 3 years ago, and we added more to the current server to bring it from 1GB up to 3GB. Then after 18 months usgae we identified as the DB got larger issues from the symantec virus checker searching the DB. The antivirus was then instructed to stop scanning the DB.
Now we're hitting the wall again, and I ponder if adding more RAM will help, we are seeking clarity if i can just buy matched pairs and add a GB myself, or if this will invalidate warrenty and support licence.
I know our currently drives are beginning to be low on space.
The main OS/programs and EXact and DB are based on 120Gb mirrored RAID, whilst our scans and radiographs were removed to a second mirrored RAID allowing 500GB storage.
The exact details of the new disk setup was yet to be discussed with the hardware folks who were recommending us the package. 4*300GB is unlikely to be suitable for long term needs unless they have something else in mind as radiograph storage.
I am assuming that ' SAS Small Form Factor 10 000 rpm 6GB/s Dual-Port' is a fast hard drive solution, but wondered regarding the entire tape backup system, and wonder if removal hard drives swapped each night and taken offsite would prove a more suitable method of backup, and if such a system would be faster?
Do you know what the current memory configuration is at present?
I would guess that if you have 1GB originally you had the entry model which came with 2 x 512MB populating slot 1 & 5. Even if the other slots are using 512MB then you should have 2 free. From memory SBS2003 only came as 32 bit so it would be pointless inserting any more than:
1 GB FBD PC2-5300 2 x 512 MB Single Rank Kit 397409-B21
These should be available for sub £50.00 for the pair. These servers are designed for user upgrades so if you have a HP care pack there shouldn't be an issue.
If you are running short on disk space this could be causing your performance issues as you are limiting the page file size. BTW, SAS disks are excellent but again without:
534916-B21 HP 512MB Flash Backed Write Cache for P410i,
performance will be crippled. Also you can't run RAID 5 on the Entry Model with the embedded controller (Embedded HP Smart Array P410i/Zero Memory Controller).
I have a number of customers using external hard disks running Symantec System Recovery which is a brilliant product. It has the ability to extract native Exchange data built in. The only limiting factor is the restrictions on how much data is stored although we have can get a months worth of incremental's on one 1TB HDD (with about 350GB data to backup. It won't be any faster than tape as you are restricted to the speed of the USB 2 ports. That said, you only need a back up window of a couple of hours to archive all the present data.
I would guess that if you have 1GB originally you had the entry model which came with 2 x 512MB populating slot 1 & 5. Even if the other slots are using 512MB then you should have 2 free. From memory SBS2003 only came as 32 bit so it would be pointless inserting any more than:
1 GB FBD PC2-5300 2 x 512 MB Single Rank Kit 397409-B21
These should be available for sub £50.00 for the pair. These servers are designed for user upgrades so if you have a HP care pack there shouldn't be an issue.
If you are running short on disk space this could be causing your performance issues as you are limiting the page file size. BTW, SAS disks are excellent but again without:
534916-B21 HP 512MB Flash Backed Write Cache for P410i,
performance will be crippled. Also you can't run RAID 5 on the Entry Model with the embedded controller (Embedded HP Smart Array P410i/Zero Memory Controller).
I have a number of customers using external hard disks running Symantec System Recovery which is a brilliant product. It has the ability to extract native Exchange data built in. The only limiting factor is the restrictions on how much data is stored although we have can get a months worth of incremental's on one 1TB HDD (with about 350GB data to backup. It won't be any faster than tape as you are restricted to the speed of the USB 2 ports. That said, you only need a back up window of a couple of hours to archive all the present data.
- Joined
- 31 May 2009
- Posts
- 21,457
Currently 6 of the 8 slots are filled, it will only support 4gb, so i was thinking of a single 2*512 kit as you suggested.
I wonder reagrding the disk space, on a 120gb raid how little is too little?
534916-B21 HP 512MB Flash Backed Write Cache for P410i I shall have to look into this, as i know little of servers and its not something i recognise from the desktop world.
thanks for responding aspirin.
I wonder reagrding the disk space, on a 120gb raid how little is too little?
534916-B21 HP 512MB Flash Backed Write Cache for P410i I shall have to look into this, as i know little of servers and its not something i recognise from the desktop world.
thanks for responding aspirin.
You're welcome Hikari.
I would suggest that if you have less than 10% disk space the performance will be restricted. You can always force the OS to use another disk for its page file if necessary just like a XP/Vista/Win 7 client.
Read page 3&4 on:
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00687518/c00687518.pdf
and you will understand why I'm so adamant about write cache. We have completed our own workshop test when we recently sold 150 HP ML330's to a nationwide retail outfit for EPOS servers and there was a massive performance increase with the BBWC. Such a difference that we convinced them to buy the additional controllers
I would suggest that if you have less than 10% disk space the performance will be restricted. You can always force the OS to use another disk for its page file if necessary just like a XP/Vista/Win 7 client.
Read page 3&4 on:
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00687518/c00687518.pdf
and you will understand why I'm so adamant about write cache. We have completed our own workshop test when we recently sold 150 HP ML330's to a nationwide retail outfit for EPOS servers and there was a massive performance increase with the BBWC. Such a difference that we convinced them to buy the additional controllers
The OS & DB being on the same spindles is the main thing I notice. Is the pagefile there as well? RAID 1 or RAID 10? MS recommends RAID 10 for MSSQL. What are you doing for backups at the moment? If you intent to take HDDs off-site, don't forget to encrypt them.
I suggest getting a proper DBA in as a contractor to make recommendations on suitability of your current system, and to do some performance checks. MSSQL has some good tools to monitor the performance of every single query. You could perhaps do this yourself if you did a lot of research.
If you're looking for recommendations, I know a very good DBA (not me) who works full-time but might be open to a bit of consulting.
I suggest getting a proper DBA in as a contractor to make recommendations on suitability of your current system, and to do some performance checks. MSSQL has some good tools to monitor the performance of every single query. You could perhaps do this yourself if you did a lot of research.
If you're looking for recommendations, I know a very good DBA (not me) who works full-time but might be open to a bit of consulting.
- Joined
- 31 May 2009
- Posts
- 21,457
As said, we are currently unable to establish if the DB is MSSQL, I will contact SOE next week and request the information.
We're in mirrored raid currently.
2 mirrors, primary array is backed up to the tapedrive system within the G5, tape changed each morning and stored offsite. Secondary array is backed up to a removable harddrive, of which we have 2, one swapped for the other each day and again stored offsite.
We would stand to lose little if an array failed, or a days information if the building burned to the ground. This is acceptable as loss of one days infomrtaion could be recovered in the period of time it would take to rehouse a 9 surgery dental practice (months).
We're in mirrored raid currently.
2 mirrors, primary array is backed up to the tapedrive system within the G5, tape changed each morning and stored offsite. Secondary array is backed up to a removable harddrive, of which we have 2, one swapped for the other each day and again stored offsite.
We would stand to lose little if an array failed, or a days information if the building burned to the ground. This is acceptable as loss of one days infomrtaion could be recovered in the period of time it would take to rehouse a 9 surgery dental practice (months).
In the world of server disks it pretty much goes as follows:
Slow > Fast
Cheap > Expensive
SATA 5400rpm > SATA 7200rpm > Near Line SAS 7200rpm > SAS 10000rpm > SAS 15000rpm > SSD (although you dont see these in servers currently)
Its perfectly reasonble to spec a server with multiple RAID arrays as per your current server and use disks for each RAID set which goes with the intended usage. So you could put some 300Gb SAS 15k disks in for the database and some 7200rpm SATA ones for the radiograph storage.
Also bear in mind that the choice of RAID level has a bearing on the speed of the disks, and the RAID controller is yet another variable. Again spend more get more.
*Very rough*, but total "speed" = number of disks divided by the following RAID IOPS numbers.
RAID 0 – 1 IOPs (No benefit to redundancy. Capacity is same as original drives)
RAID 1 – 2 IOPs (Data is mirrored across 2 disks, hence 2 writes for every logical write. Capacity will be 50% however.)
RAID 5 – 4 IOPs (Data is broken into chunks and a check digit is calc'd and written, hence the number of writes. Capacity is total number of drives minus 1)
RAID 6 – 6 IOPs (Same as above but has 2 check digits IIRC - nobody ever uses RAID 6 IMHO)
RAID 10 – 2 IOPs (Basically its RAID 0 and RAID 1 combined. Very fast, but you only get 50% of the original drive capacity)
As you can see, the fastest setup you can have is RAID10, especially on a transactional database, but RAID5 is OK too, so long as you have a good controller card; as remember all those check digit calculations will be happening on that controller (or worse in the driver!).