Comtech EF Data CDM-840 - Manual (Page 208)

CDM-840 Remote Router Revision 2
Appendix D MN-CDM840
D–2
widely available as an open-source software tool; it does not require the use of a software
license.
Deflate-compressed blocks are wrapped with a header and footer to become GZIP files.
Typically, when a classical, single, general purpose CPU performs GZIP compression, either the
compression performance is scaled back to maximize data throughput speeds, or the CPU runs
slow. To negate either deficiency, an efficient solution is to offload the compression task to a
hardware-based GZIP function, as is accomplished with the CDM-840. Hardware-based GZIP
compression offloads lossless data compression and frees up the system’s main CPUs. This
allows the compression functions to operate not only independently, but also at much higher
data rates if needed. The ASIC takes in uncompressed input data, compresses it, and outputs the
data in compressed form. The compression hardware does many tasks in parallel, only offloaded
from the central CPUs of the CDM-840. This effectively eliminates the multi-pass and iterative
nature typical of a classical, single, general purpose CPU that is over tasked with executing the
Deflate algorithm.
D.1.2 C ompres s ion P erformance
Compression performance is classically measured by two metrics – size reduction and data
throughput:
• Size reduction is usually reported as a ratio of the uncompressed original size divided by
the compressed size.
• Data throughput is measured in bytes per second (bps) as measured on the
uncompressed side of the GZIP ASIC.
Data complexity has no effect on data throughput. Easy-to-compress data files that compress
with a high ratio pass through the co-processor at the same high data rate as very complex data,
which achieves lower compression ratios.
Table D-1 provides Comtech AHA GZIP compression ratio results as compared with the Calgary
Corpus and Canterbury Corpus industry standard file sets and algorithms. The HTML file set is
from a collection of Internet dynamic content; LZS (Lempel-Ziv-Stac) compression results are
based on publicly available descriptions of the LZS algorithm.
Table D-2 outlines the comparison the effects of the CTOG specification for current operation,
based on a session-based compression for which the current performance specifications are
given.
Table D-1. Comtech AHA GZip Performance Comparisons
File Sets Comtech AHA363-PCIe LZS ALDC
Calgary Corpus 2.7:1 2.2:1 2.1:1
Canterbury Corpus 3.6:1 2.7:1 2.7:1
HTML 4.4:1 3.4:1 2.65:1