Terry Lemons wrote:
Hi Matthew
Check out the NetWorker Performance Tuning guide for your release. In
particular, look into using bigasm.
Funny that you bother mentioning the Performance Tuning Guide. This
guide (in all of its versions, 7.4 included) is quite useless.
1. It contains references to hardware and software which is obsolete.
"The Ampex DST tape drive writes only 3 MB per second with 64 KB SCSI
transfers and 20 MB per second with 992 KB SCSI transfers" (7.4 PTG page
42).
"SCSI bus performance depends on the technology deployed. Some
specifications for
different types of SCSI technology include the following:
? SCSI-2 single-ended, 10 MB/s
? FWD (fast-wide-differential), 20 MB/s
? UltraSCSI, 40 MB/s
? Fibre
Note the following:
? The rated speeds of SCSI buses are defined as "best case."
? The selection of UltraSCSI-compatible devices is limited."
(7.4 PTG page 22).
"Upgrade to faster LAN media, such as 100 MB Ethernet or FDDI, to improve
network bandwidth. Note, however, that TCP packet overhead can reduce the
total transfer rate."
(7.4 PTG page 31)
"Note: For Solaris 2.6, the parameter is check tcp_close_wait_interval."
(7.4 PTG page 25) Solaris 2.6 is not even supported with Networker 7.4.
2. It contains vague and meaningless statements.
"Note: Install sufficient computer and network hardware to support a
storage application.
Failure to meet minimum system requirements, particularly for memory,
free space, and CPU
speed, can have an obvious and immediate negative impact on the storage
management
processes."
(7.4 PTG page 16)
3. It lacks references to recent hardware (such as LTO).
4. It lacks concrete information about the expected performance which
will help us do basic server sizing. I would like to know how much LTO-3
drives I can expect to run at full speed with a typical Xeon. I would
like to see performance numbers for typical configurations. Statements like:
We were able to get two LTO-3 drives at their max speed with a dual
core 3.0Ghz, with 2Gb running Windows 2003 SP1 and Networker 7.2.2.
would allow me to estimate which server I need to buy before I
actually spend the money. This is relevant for both new and old
installations. Currently we need to guess and if we guess wrong we end
with a suboptimal and unbalanced configuration (such as trying to drive
an LTO-4 drive with a 486 over a Fast Ethernet). I would like to see
performance data for five typical configurations. I would like to see
performance figures for AFTD devices as well (the AX100 comes to mind as
an example for such a product).
5. OS related issues are also not discussed. Performance comparison
of UFS vs ZFS (or ext2/3 vx XFS) could let us decide which to choose and
what to expect with each configuration. Performance comparisons of the
Leadville SAN stack and Emulex/Qlogic drivers for HBAs might also be
useful.
It seems that this guide was written at 1995. It doesn't seem as if
someone bothered to read it carefully before approving it for the next
version. To summarize, it leaves a bad impression.
Or maybe EMC thinks that paying customers should not be left to make
any performance tuning or sizing decisions, but rather pay more for EMC
professional services to do the sizing for them.
tl
--
-- Yaron.
To sign off this list, send email to listserv AT listserv.temple DOT edu and type
"signoff networker" in the body of the email. Please write to networker-request
AT listserv.temple DOT edu if you have any problems with this list. You can access the
archives at http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
|