IBM System Storage™

October 2014

IBM System Storage™ TS1150 Tape Drive

Non Volatile Caching

By Christine Knibloe
IBM Firmware Engineer, Tucson, AZ
IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 1

Introduction
Contents
The purpose of this whitepaper is to examine the benefits of non-volatile caching, a
function of the IBM System Storage TS1150 Tape Drive (TS1150 tape drive) that
Introduction...................................... 1 improves write performance and cartridge storage capacity.
TS1150 Overview ............................ 1
TS1150 Overview
Performance Overview .................... 2
Non Volatile CachingError! Bookmark not Thedefined.
TS1150 tape drive provides the highest levels of performance and cartridge
Backhitch Reduction ........................ 5 capacity. The TS1150 tape drive is compatible with the existing IBM TotalStorage
tape automation installations, thus allowing customers to protect their automation
Dataset Packing .............................. 6 investments.
Summary ......................................... 8
Supported environments include IBM System i, System p, System z and System x
servers, selected Oracle and Hewlett Packard servers; and Intel servers running
supported versions of Microsoft Windows or Linux:
The TS1150 tape drive includes a dual-ported 8 Gbit Fibre Channel interface for
attachment to supported adapters in supported servers or supported 1 Gbit, 2 Gbit, 4
Gbit, or 8 Gbit switched fabric or arbitrated loop Storage Area Networks (SANs).
The native data rate of the TS1150 tape drive is 360 MBps for type D media and
300 MBps for type C media (re-formatted). The native uncompressed cartridge
capacity is 10 TB (using JD/JZ media), 7 TB (using JC/JY media), and 2 TB (using
JL media). The actual data rate and cartridge capacity however may be substantially
higher depending on the environment..
The TS1150 tape drive also features technology enhancements that may help
improve performance. These include a larger internal buffer, dynamic digital speed
matching, increased search speed and Streaming Lossless Data Compression
(SLDC) algorithm
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 2

Performance Overview
The TS1150 tape drive features technology enhancements that may help improve
performance. These include a larger internal buffer, dynamic digital speed
matching, increased search speed and Streaming Lossless Data Compression
(SLDC) algorithm.
Key features of the TS1150 tape drive that are designed to improve performance
include:
 360 MBps native data rate (type D media)
 10 TB native data capacity (JD/JZ media)
 800 MBps burst data rate
 12.4 m/s maximum end-to-end high-speed search rate
 Encryption support
 IBM Linear Tape File System (LTFS) support
 Media partitioning
 Virtual Backhitch and high-resolution directory
 Compatible with existing IBM tape automation installations

e 2 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 3

Non Volatile Caching

Non-volatile caching has two major components that are designed to deliver
improvements in capacity utilization and net drive throughput when compared to
standard linear tape drives. These are ‘Backhitch Reduction’, which improves write
performance, and ‘Dataset Packing’ which optimizes capacity.
Tape drives typically have an internal, volatile buffer that is designed to match the
performance characteristics of the tape drive. The buffer stores the incoming
datastream until the tape drive can process the data and write it to tape. In the
TS1150 tape drive, this buffer is 2 GB.
This process is illustrated in the diagram below.

The tape drive’s internal buffer stores incoming data until it can be written to tape.

HOST BUFFER

TAPE DRIVE CARTRIDGE

After a certain amount of data accumulates, the buffered data is flushed to tape. As data
is written to tape, it is removed from the buffer.

Data accumulates in the buffer until it reaches the write threshold. At that point the
The presence of this buffer facilitates drive begins to write the buffered data to tape. While writing, the drive moves tape
streamed writing at a constant speed. This is called a ‘streamed write'. Speed matching devices, such
as the TS1150, have multiple speeds available for streamed writing.
As data is written to tape it is removed from the volatile buffer. Under certain
conditions, the data buffer may empty. This could occur if the effective host data
rate is slower than the drive's native data rate, or if a synchronizing command is
received.
A slow data rate can cause streamed With highly compressible data, even a relatively fast host can have an effective data
writing to cease rate that is less than the drive's slowest native data rate. In such cases, the amount
of data being written to tape, and therefore being removed from the buffer, is
greater than the amount of data entering the buffer from the host interface. Any
time the data buffer empties, streamed writing must cease.

e 3 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 4

Synchronizing commands cause streamed Any non-immediate write-type command is considered a synchronizing command.
writing to cease Synchronizing commands require the drive to store data on tape prior to returning
command complete status. This differentiates synchronizing commands from
immediate write-type commands where the drive can return command complete
status as soon as data is stored in the volatile buffer. A SCSI ‘Write Filemark’
command with the ‘Immediate bit’ equal to zero is an example of a synchronizing
command.
To maintain the order of data when a synchronizing command is received, all data
previously stored in the buffer must be written tape ahead of any data associated
with the synchronizing command. Therefore the entire buffer is flushed to tape.
Even if no data is associated with the synchronizing command, such as a non-
immediate Write Filemarks command with a transfer length of zero, the entire
buffer is still flushed. Following the flush, the amount of buffered data is well
below the write threshold.
When streamed writing ceases, tape motion does not halt instantaneously. For
example, if tape is moving at 6.22 m/s when the data buffer empties, an entire meter
of tape can go unwritten in 160 milliseconds. Lengths of unwritten tape reduce
capacity. Typical tape drives use a backhitch to prevent this type of capacity loss.

A backhitch is a mechanical reposition Backhitch operations are performed in order to eliminate lengths of unwritten tape.
of tape The process is rather intricate.
• First the drive must decelerate to halt the movement of tape in the forward
direction.
• The tape drive then reverses direction to the point where previously data
was written
• At that point, the drive decelerates to halt the movement of tape in the
reverse direction.
• Now tape accelerates in the forward direction so that the write operation
can continue.
Each step of the backhitch operation must be precisely controlled so that the tape
reaches streamed writing speed before the first portion of unwritten tape is reached.
Even with this exactitude, the time required to complete a backhitch can
significantly degrade performance.

e 4 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 5

Writing from buffer.

Buffer empties. Stop writing. Decelerate to stop.

STOP

Reverse the direction.

Access previously written data.

STOP

Reverse the direction. Accelerate to write speed.

Resume writing.

Backhitch Reduction
Non-volatile caching, or NVC, is a TS1150 tape drive feature which can greatly
improve write performance. If the TS1150 determines synchronizing commands
and the resulting backhitches are sufficiently frequent to degrade write
performance, the drive invokes NVC mode. NVC mode improves write
performance through backhitch reduction. The transition between standard writing
and NVC mode is transparent to host applications.
In order to perform backhitch reduction, NVC temporarily reserves portions of
physical tape for cache areas. Data received from the host is written to the volatile
buffer as usual, and also to the non-volatile tape caches areas.

e 5 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 6

Writing data to tape on a continuous basis eliminates the overhead normally

incurred when flushing the entire buffer on a synchronizing command. As an added
benefit, all data written during NVC mode has a copy in non-volatile memory,
namely tape.
While writing to the cache areas, the restrictions on dataset spacing are relaxed.
Larger spaces are allowed between datasets in the cache areas. This has a
temporary impact on capacity, but virtually eliminates the need for backhitch
operations

Allowing larger spaces in the cache areas virtually eliminates the need for backhitch
operations.

Standard Portion of Tape

Tape Cache Area

The temporary capacity loss is easily recouped. While data is written to the tape
cache areas, the same data accumulates in the volatile buffer. When certain criteria
are met, NVC mode ends. At the end of the NVC cycle, the data in the volatile
buffer is streamed onto the standard portion of tape.
The tape cache areas are then released so that they may be overwritten. Rewriting
to the data portion saves time when compared to the alternative backhitches. The
result is improved write performance.

Dataset Packing
The smallest complete unit of information written to or received from the TS1150
tape drive is called a dataset. The size of the TS1150 dataset is fixed. Dataset size
is not related to the size of records transferred by host applications. Instead host
records are fit into the constraints of datasets. Certain conditions allow datasets to
go unfilled, thereby reducing capacity.
As previously stated, synchronizing commands issued during standard tape drive
write operations cause the entire buffer to flush to tape. When the flush occurs, data
is committed to tape in the same order it was received.
Normally the data associated with a synchronizing command is also the last data
included in a dataset. Even if the dataset is not full, the next data written to tape
goes into a new dataset. Frequent synchronizing commands result in many unfilled
datasets. The effect is reduced cartridge capacity.

e 6 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 7

This diagram shows how traditional writing can cause unfilled datasets. Individual host
records are depicted with distinct colors and letters.
Small records may fit completely within a single dataset, like Records A, C, E, and F,
while larger records will span multiple datasets, like Records B and D.
Because the datasets containing B2, D2, E and F are unfilled, we can assume Records
B, D, E, and F were associated with synchronizing commands.

Unfilled datasets created with traditional writing.

A B1 B C D1 D2 E F
2

NVC reduces the number of unfilled datasets through dataset packing. In the tape
cache area unfilled datasets are rewritten with additional data. This creates multiple
versions of a single dataset. Each version contains more data than the version that
preceded it. The process repeats until a full, or packed, version of the dataset is
created. The process then repeats for the next dataset.

Multiple dataset versions in the tape cache area.

A B1 B B C D1 D2 D2 E D2 E F
2 2

Copy 1 Copy 1 Copy 1

Copy 2

Writing multiple dataset versions creates redundant copies of some records in the
tape cache area, but the redundant copies are never written to the standard portion
of tape.
While data is written to the tape cache areas, the same data accumulates in the
volatile buffer. When certain criteria are met, the packed version of each dataset is
written to the standard portion of tape from the volatile buffer. Writing only one
version, the packed version, eliminates partially empty datasets in the standard
portion of tape. Writing packed datasets also optimizes capacity.

Packed Datasets in the Standard Portion of Tape

A B1 B C D1 D2 E F
2

After copying the packed versions to the standard portion of tape, the tape cache
areas are released so that they may be overwritten. This eliminates the redundancy.

e 7 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 8

Summary
The two components of non-volatile caching provide major performance and
capacity improvement over standard LTO writing. Dataset packing improves
overall tape capacity through improvements to individual datasets. Backhitch
reduction decreases the frequency of costly mechanical repositioning. Non-volatile
caching provides an innovative approach to increasing both capacity and write
performance.

e 8 of 10
 IBM System Storage TS1150 Tape Drive Performance Whitepaper
Page 9

© International Business Machines Corporation 2014

IBM Systems Group
5600 Cottle Road,
San Jose, CA 95193

All Rights Reserved

IBM, the IBM logo, eServer, iSeries, pSeries, zSeries, and

TotalStorage are trademarks or registered trademarks of
International Business Machines Corporation in the United
States, other countries, or both.

Other company, product and service names may be trademarks

or service marks of others.

Product data has been reviewed for accuracy as of the date of initial publication.
Product data is subject to change without notice.

This information could include technical inaccuracies or typographical errors. IBM

may make improvements and/or changes in the product(s) and/or programs(s) at
any time without notice.

References in this document to IBM products, programs, or services does not imply
that IBM intends to make such products, programs or services available in all
countries in which IBM operates or does business.

THE INFORMATION PROVIDED IN THIS DOCUMENT IS DISTRIBUTED

“AS IS” WITHOUT ANY WARRANTY, EITHER EXPRESS OR IMPLIED. IBM
EXPRESSLY DISCLAIMS ANY WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE OR INFRINGEMENT.

IBM shall have no responsibility to update this information. IBM products are
warranted according to the terms and conditions of the agreements (e.g., IBM
Customer Agreement, Statement of Limited Warranty, International Program
License Agreement, etc.) under which they are provided.

e 9 of 10