QUESTIONS ABOUT 7135 RAID SYSTEM SPACE AVAILABLE
ITEM: RTA000034955
QUESTION:
I'm responding to an RFP requiring two machines, one with 20GB
of user file storage space and one with 10GB of user file storage
space. The standard disk in the RS/6000s being proposed will handle AIX
and we want to bid 7135 RAID subsystems for the 20GB and 10GB of user
space. The RFP stipulates that the 20 and 10 GB should be usable space
not including file pointers (which I take to mean inodes) and any other
"preallocation of file space". I have configured the 7135s as follows:
20GB system: 2 banks of 5*2GB drives = 16.0 GB data space
1 bank of 5*1.3GB drives = 5.2 GB data space
Non RAID parity space available: 21.2 GB
If they use JFS, I would allow 3% for inodes yielding 20.5 GB of space
available for user data.
10GB system: 2 banks of 5*1.3GB drives = 10.4 GB data space
Non RAID parity space available: 10.4 GB
If they use JFS, I would allow 3% for inodes yielding 10.09 GB of space
available for user data.
1) Is this a good scenario for the specified disk requests at the
lowest cost using 7135 RAID?
2) Also, how many bytes are actually available on the 1.3 and 2.0 GB
drives used in the 7135 and are those numbers based on
software GB (1GB = 1,073,741,824 bytes) or typical disk GB
(1GB = 1,000,000,000 bytes)?
3) If the disks are formatted using 4MB partitions in a jfs, how many
PP's are available for the 2GB drive and the 1.3GB drive in a 7135?
4) If the customer sets up the 7135 to be accessed in RAW mode by
a database, are there any general rules of thumb regarding space
required for pointers
(the database has not been chosen yet so comments related
to any major database like Oracle, Sybase, or Informix would be fine)?
5) Is there anything in AIX that would require us to respond to the
following question: "Offerors requiring preallocation of file space
should so state, provide an estimate of waste space thus created,
and propose adequate storage to compensate for this loss"?
6) I understand that the maximum transfer rate for the adapter is
10MB per second, but what is the expected average transfer rate for
a relational database and what assumptions should be stated (like
RAID-5, blocksize=xxx etc.)? If we attach a second controller to the
7135, when dual active controller configurations are supported in 1994,
I assume that each controller will be able to transfer at a maximum
of 10MB/sec and at the same average rate specified as a response to the
prior sentence. Is this correct?
---------- ---------- ---------- --------- ---------- ----------
A: I'm responding to an RFP requiring two machines, one with 20GB
of user file storage space and one with 10GB of user file storage
space. The standard disk in the RS/6000s being proposed will handle
AIX and we want to bid 7135 RAID subsystems for the 20GB and 10GB of
user space. The RFP stipulates that the 20 and 10 GB should be usable
space not including file pointers. I have configured the 7135s as
follows:
20GB system: 2 banks of 5*2GB drives = 16.0 GB data space
1 bank of 5*1.3GB drives = 5.2 GB data space
Non RAID parity space available: 21.2 GB
If they use JFS, how much space would would be used for inodes
after filesystem creation, hiw much space will we have for
available for user data.
>>>> Answer
jfs will use 3% for pre-allocation, the rest is usable.
The disk capacity is in powers of 10 (as stated in the Announcement
letters) and not in powers of 2. According to this, you will have
20.6 and 10.01 gigabytes, which puts you within spec.
1) Is this a good scenario for the specified disk requests at the
lowest cost using 7135 RAID?
>>>> Answer
I'm assuming that you are considering Model 110.
The announcement 193-195 has the following prices:
Three 1.3GB Disk Drive Modules,
First Bank of Disks Completion 271$ 15,600 $ 69
Five 2.0GB Disk Drive Modules,
First Bank of Disks Select 2723 24,600 79
Five 1.3GB Disk Drive Modules,
Full Bank of Disks 2715 26,000 115
Five 2.0GB Disk Drive Modules,
Full Bank of Disks 2725 35,000 125
Additional Array Controller 3050 13,150 80
Disk Drive Expansion Unit 4000 9,250 20
Selectable codes available with the 7132 Model REP
1.3GB Replacement
Disk Drive Module 2716 5,305 0
2.0GB Replacement
Disk Drive Module 2726 7,105 0
Even though it's not very clear, it appears that the unit price for
a 2GB disk is less than a 1.3GB diską I would not use the
1.3GB at all, not knowing other discounts which maybe applicable.
Consider the following comment:
#2723 is an additional cost to put 2 GB drives in the first
rank. #2725 should be used to calculate the cost of 5x2 GB
drives; #2715 should be used to calculate the cost of 5x1.3
GB drives. Thus the 1.3GB drives are $5.2K/drive or $4K/GB;
the 2GB drives are $7K/drive or $3.5K/GB.
2) Also, how many bytes are actually available on the 1.3 and 2.0 GB
drives used in the 7135 and are those numbers based on software GB
(1GB = 1,073,741,824 bytes) or typical disk GB(1GB=1,000,000,000
bytes)?
>>>> Answer
Provided above.
3) If the disks are formatted using 4MB partitions in a jfs, how many
PP's are available for the 2GB drive and the 1.3GB drive in a 7135?
>>>> Answer
The partition size comes in powers of 2. So, a 4MB partition is 2**22
which equals 4,194,304 bytes. This yields 238+ partitions, according
to my arithmetic. Do the same for the 1.3GB.
4) If the customer sets up the 7135 to be accessed in RAW mode by
a database, are there any general rules of thumb regarding space
required for pointers.
>>>> Answer
No. The application (or user) defines those.
5) Is there anything in AIX that would require us to respond to the
following question: "Offerors requiring preallocation of file space
should so state, provide an estimate of waste space thus created,
and propose adequate storage to compensate for this loss"?
>>>> Answer
I only know of the 3% indicated above for the inodes and also the
fractional partitions, which of course depend on how large they are
defined. For a 4MB partition size, 4MB - 1 is the largest "waste"
space that could be added to the inode space.
6) I understand that the maximum transfer rate for the adapter is
10MB per second, but what is the expected average transfer rate for
a relational database and what assumptions should be stated (like
RAID-5, blocksize=xxx etc.)? If we attach a second controller to the
7135, when dual active controller configurations are supported in 1994,
I assume that each controller will be able to transfer at a maximum
of 10BM/sec and at the same average rate specified as a response to the
prior sentence. Is this correct?
>>>> Answer
No performance measurements are yet available on the dual controller
configuration.
Alos,we don't have complete performance data but we have measured
RAID 5 performance at about 380-390 ops/sec for random 4K
reads and writes with 75% reads. It doesn't make sense to
talk about throughput for applications with small accesses
(such as 4K) since you'll be bound more by operations per
second than by throughput. To get high throughput, you
need large sequential accesses, such as in technical
applications.
---------- ---------- ---------- --------- ---------- ----------
This item was created from library item Q643144 CLDJS
Additional search words:
AVAILABILITY AVAILABLE CLDJS DATA ERROR FILESYSTEM HARDWARE IX NON
NOV93 OZIBM PARITY POINTERS QUESTION RAID RISCOHW RISCSYSTEM SPACE
SPACING SYS SYSTEM USER VERIFY 7135
WWQA: ITEM: RTA000034955 ITEM: RTA000034955
Dated: 04/1996 Category: RISCOHW
This HTML file was generated 99/06/24~12:43:13
Comments or suggestions?
Contact us