Re: RAID and Partition

default is a stripe size of 64k


"John Fullbright" <fullbrij@xxxxxxxxxxx> wrote in message
news:%23hejeYDIGHA.3036@xxxxxxxxxxxxxxxxxxxxxxx
>A relatively large multiple of 4K.
>
>
> "Kerry Hoskin" <kjho@xxxxxxxxx> wrote in message
> news:%23elwtQDIGHA.2300@xxxxxxxxxxxxxxxxxxxxxxx
>> any recommendations as to the stripe size to use for the RAID 10 set?
>>
>> Kerry
>>
>> "John Fullbright" <fullbrij@xxxxxxxxxxx> wrote in message
>> news:%23WgCh3sHGHA.1288@xxxxxxxxxxxxxxxxxxxxxxx
>>> RAID 1 for OS/Page File.
>>> RAID 1 for Logs
>>> RAID 0+1 For DBs
>>>
>>> When you build out the 2800, don't use Server Assist. Server Assist
>>> starts with a FAT partition that is later converted to NTFS. This
>>> results in an allocation unit size, or cluster size, of 512 bytes. IO
>>> to the page file is 4K. When an IO request larger than the allocation
>>> unit size hits the file system driver, it splits the request into
>>> multiple smaller requests. In this case, each time a page is swapped
>>> that's 8 IOs instead of 1. You want a 4K allocation unit size on the OS
>>> partition, which by the way, is the NTFS default.
>>>
>>>
>>>
>>> "Kerry Hoskin" <kjho@xxxxxxxxxxxxxxxx> wrote in message
>>> news:%23UF9JyqHGHA.208@xxxxxxxxxxxxxxxxxxxxxxx
>>>> similair question.
>>>>
>>>> DELL PE2800 with 8*36Gb 15k disks around 180 mailboxes.
>>>>
>>>> Thinking of RAID 1 for OS,Pagefile
>>>>
>>>> and then either RAID 10 for the remaning 6 disks with partitions for
>>>> logs and DB or RAID 1 for logs and RAID 5+ hotspare
>>>>
>>>>
>>>>
>>>> "John Fullbright" <Fullbrij@xxxxxxxxxxx> wrote in message
>>>> news:OT3P76VHGHA.2212@xxxxxxxxxxxxxxxxxxxxxxx
>>>>>I like to keep the logs seperate. With only 100 users you may want to
>>>>>consider
>>>>>
>>>>> a mirror for the OS and smtp queue
>>>>> a mirror for the logs
>>>>> a mirror for the databases
>>>>>
>>>>> If we assume 10K drives, a mirror will give you:
>>>>>
>>>>> writes = 85*2/2 or 85 IOPS
>>>>> reads = 85*2 04 170 IOPS
>>>>>
>>>>> With a 3:1 read/write ratio for the database applied:
>>>>>
>>>>> (170*.75) + (85*.25) = 127.5 + 21.25 = 148.75 IOPS
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> "jc" <jc@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
>>>>> news:DD5FACAE-63CF-4F0A-AC93-5721FB28B39E@xxxxxxxxxxxxxxxx
>>>>>> Can some one confirm following configuration is effection for 100
>>>>>> Exchange
>>>>>> 2003 users?
>>>>>> - 6 ea SCSI 73G ea RAID system
>>>>>> - RAID 1 (disk0 and disk1) for OS, Logs and Queue
>>>>>> - RAID 10 (disk 2,3,4,5) for data
>>>>>>
>>>>>> Question:
>>>>>> - Should I partition RAID1 to C: (20G) for OS, D: for others, any
>>>>>> advantage
>>>>>> - Is it better to have RAID 5 (instead RAID10) with HotSpare, so I
>>>>>> have
>>>>>> better protection?
>>>>>>
>>>>>> Thanks,
>>>>>> JimmyChang
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>


.

Relevant Pages

  • Re: Gvinum RAID5 performance
    ... >> a single request, ... should preferably be a multiple of the ufs block size, ... > It seems we are holding back all requests to a currently active stripe, ... > about to be overwritten), and the old parity, and recalculate the parity ...
    (freebsd-current)
  • [md-accel PATCH 07/19] md: raid5_run_ops - run stripe operations outside sh->lock
    ... The raid5_run_ops routine uses the asynchronous offload api and ... copy data into request buffers to satisfy a read request ... recalculate parity for new data that has entered the cache ... the stripe lock, but it made sense to add it as an operation type ...
    (Linux-Kernel)
  • RE: Multiple requests to a webservice...???
    ... from client and perform some processing and write the output back to ... you're looking for a way to make multiple such requests ... package into a single http request so that you can make multiple such web ... Microsoft MSDN Online Support Lead ...
    (microsoft.public.dotnet.xml)
  • [PATCH 05/16] md: add raid5_run_ops and support routines
    ... Prepare the raid5 implementation to use async_tx for running stripe ... biofill (copy data into request buffers to satisfy a read request) ... biodrain (copy data out of request buffers to satisfy a write request) ... postxor (recalculate parity for new data that has entered the cache) ...
    (Linux-Kernel)
  • [PATCH 03/12] md: add raid5_run_ops and support routines
    ... Prepare the raid5 implementation to use async_tx for running stripe ... biofill (copy data into request buffers to satisfy a read request) ... biodrain (copy data out of request buffers to satisfy a write request) ... postxor (recalculate parity for new data that has entered the cache) ...
    (Linux-Kernel)