Rules of thumb for calculating tape capacity 1

[eddysteenbergen]

We are running AS 2000 under Win2000. But the 40/80 DLT tape (compressed format) seems to fill up after only about 53gb of data. We find this a little disappointing.

1. Do others get this sort of performance?

2. Does anyone know how to increase tape capacity in AS2000? e.g. by changing blocking factors

Eddy Steenbergen
Team Leader Internal Systems
Corporate IT
Dept Primary Industries Water and Environment
Tasmania, Australia
ph: (03) 6233-3838

 

[jimpames]

http://support.ca.com/techbases/as61/MediaFull.html

http://support.ca.com/products/nt/16004.pdf

http://support.ca.com/techbases/as61/NAS60378.html

http://support.ca.com/techbases/asnt/NASNT598.html

------------------------
The advertized capacity of your media is based on the assumption that your data will be compressable at a ratio of 2:1.
It appears that the data your are backing up is only 1.3:1 compressable.
Raw text is highly compressable.
Some files can not be compressed, they are already compressed.
The drive uses hardware compression.
The correct block size is 64K.
Do not change the block size if it is 64K.
------------------------------------

http://support.ca.com/techbases/as61/10022.html#18

18. Why does not my tape drive hold as much data as advertised by the vendor ?

There are four factors that may affect the data capacity on tapes:

Data compression ratio.
Backing up files already compressed.
Providing steady and sufficient data flow to the tape drive.
Recoverable write errors.
These are explained in the following passages:
Data Compression Ratio Take a closer look at the brochure of the tape drive you purchased. You will probably notice a small asterisk next to the capacity amount, or it might use a figure like 4/8 or 8/16. The first number represents what they call the native capacity of the drive. That is how much the drive can really hold. The second number is the "perceived" capacity of the drive if Data Compression is factored in. Also look at the bottom of the page in small print where it tells you the asterisk is a disclaimer saying it assumes a 2:1 compression ratio and that the amount you get may vary. The amount a file will compress depends on what type of data is contained within that file. There is no way to know exactly how much data will fit on tape until you hook up the tape drive and see how well the data you are backing up compresses. Thirty to fifty percent above the native capacity is about average.
Backing up Files already compressed One of the major features that was introduced with NetWare 4.x was data compression on the server volumes. ARCserveIT is designed to keep a file in its compressed state when it delivers it to the tape drive. This is the fastest and most efficient way of backing up the data. Since the data is already compressed the tape drive is not going to be able to compress it further. An example of this can be found in the lines from the Activity Log below. This is what you will see in the log for the beginning of the backup: Approximately 8,301 files (s) 1,001,017 KB will be processed. This is the estimation that is done before the data starts backing up. The amount of data represents the amount of data there is to backup in its uncompressed state. Then at the end of the backup you see something like this: 8,301 files 562,980 KB written to @26,472 KB/min. This represents the amount of data that is actually written to tape. This is the same data that the estimation made at the beginning of the backup, but the actual amount of data written to tape is the compressed value. Like we stated before, ARCserveIT leaves the data in its compressed state. Therefore since the tape drive can not further compress the data it is backing up then the total amount of data that is calculated as being written to tape is going to be much closer to the native capacity of the drive rather then the maximum compression value.
Providing steady and sufficient data flow to the tape drive The SCSI host has to maintain a steady and sufficient data flow to the tape device to optimize the utilization of the space on the tapes. When the host adapter does not provide data to the drive fast enough, the drive has to flush its cache buffer and stop the tape as it waits for data to fill up the cache buffer. Once the buffer is filled, the tape is started again and the write process continues. Since most tape drives heads are designed to write a big chunk of data during each write shot (e.g.: 256Kbytes), if there is not 256K worth of data available in the buffer, the tape drive will pad the data with blank signals and still write 256K per shot, part of which will be wasted space. So, if the SCSI data transfer speed is constantly unable to catch up with drive write speed, the drive will need to continually flush the partially-full cache buffer, write the available data with padding, stop the drive, wait for more data, and then continue another cycle. This can greatly affect the utilization of the space on the tapes.When the SCSI host continuously sends a steady and sufficient data flow to the tape drive, the drive always flushes a full cache buffer (cache buffer size = n * 256 Kbytes), so that no data padding is required (and no tape space is wasted) until the end of the data transfer. The key to fast, efficient tape backups is that the tape server and host adapter maintain a steady and sufficient data flow of data to the tape device. This prevents tape drive stoppages, optimizes the utilization of the space on the tapes, and provides the fastest possible write speed. Tape drives writes data to tape in units called frames. For example, on an HP DAT tape drive, 1 frame is 256Kbytes. A frame is the smallest amount of data that a tape drive can write to a tape. If a frame is not completely filled with data when a tape drive needs to flush its internal cache buffer, the tape drive automatically pads the frame so that it is completely "filled." The padded data is transparent because the tape drive automatically filters out the padding when the tape is read. Only the actual data is received from the drive during a tape read. In the worst case scenario, a frame will have only 1 byte of real data while the rest is padded with filler when the frame is written to tape, thereby filling the tape more quickly with less real data. These tape drive stoppages (called "shoe-shining&quot occur when the tape server software or the SCSI adapter cannot send data to the tape drive quickly enough to keep the tape drive streaming (that is, writing continuously). If the tape drive needs to pause (because it is not receiving data quickly enough), and exceeds the timeout for the retention of data in its internal cache buffer, the drive will write whatever data is in the buffer to tape. If the last frame in the buffer is not completely filled with "real data" then the drive will pad out the rest of the frame and write it to tape. If the tape drive constantly "shoe-shines" during the entire backup job, then there will be many incomplete frames. If the tape drive is streaming, then the internal buffer will nearly always be full, and there will be fewer incomplete frames, or none at all.
Recoverable write errors. When write errors occur, the tape drive automatically rewrites the data. If the rewrites are successful, the tape drive continues with next data writes. So, write errors are recovered from at the expense of tape space. When excessive write errors occurs (possibly caused by dirty read/write heads, or a faulty tape device), the data capacity of the tapes can be greatly affected.

Jim P. Ames

http://www.jimpames.com