Letters to the editor

Java happens to be here now, and from what we hear it's pretty impressive. Perhaps you've heard differently. You will recall that the editorial points out that Sun has a long way to go between today's Java excitement and actual success, and we listed some examples of issues that have to be addressed and dangers avoided, including "second coming" hysteria, lest Java become, indeed, yesterday's news. And it could easily fail for other reasons. Your complaint of bias does not hold water.

The thrust of the editorial was that critiques of Java and its premise often miss the point.

Two questions and a comment for the Career Advisor

I am 25-year-old computer professional whose job responsibilities include supporting Microsoft's BackOffice products at a local Microsoft distributor. During the past five years I've been sending resumes to US computer companies, however I did not have any luck in finding a job in the US.

What am I doing wrong?

--Kevin Mizzi, (firm indeterminate)

Editors:

If system administrators are in such demand, why are there so few job postings? Do you have information on salaries for various levels of Unix system admin jobs by geographic location? I just moved to Colorado and want to know what I'm worth.
--(name and firm indeterminate)

Editors:

(Regarding Saadi's Career Advisor column: NT or Unix?) I really do not think so. Solaris is a true multi-threading and -tasking OS. But why NT is so popular? NT is catching on with its version 4.0 and pricing beats Solaris. You can get Solaris workstation for approx $1,200 but NT for $250; $650 for server. The C++ MFC priced at $450 but Solaris C++ is more than $1,000. Sun has an excellent product but high prices might kill it. As a network Admn, I think Sun will only shine with better pricing, or it might fade away.

--(name and firm indeterminate)

Performance Q&A column correspondence

Adrian:

Does the current edition of your book cover SunOS 5.5?

--(name and firm indeterminate)

Adrian:

Where may I find ruletool? (Enjoyed your article.)

--Stuart Cracraft, (firm indeterminate)

Adrian:

While the article on kernel and /etc/system parameters there are still several things missing. In Solaris 2.4 there was a fix in /etc/system

set nfs:nfs_portmon=1

/usr/ccs/bin/nm /kernel/misc/nfssrv | grep portmon
[29]    |      1676|       4|OBJT |LOCL |0    |3      |nfs_portmon

The manual page for nfsd(1m) also documents this change.

If the NFS_PORTMON variable is set, then clients are required to use privileged ports (ports < IPPORT_RESERVED) in order to get NFS services. This variable is equal to zero by default. This variable has been moved from the "nfs" module to the "nfssrv" module. To set the variable, edit the /etc/system file and add this entry:

          set nfssrv:nfs_portmon = 1

Adrian:

Hi, I have three or four SPARC 5 machines running Solaris 2.4 that have to be rebooted every two weeks or even every week depending on what the users are doing (e.g., every week if they continually run Matlab simulations along with the usual programs such as text-editing, email; longer for email and text-editing until they start running more resource demanding programs). What happens is that the machines slow down for a while and then completely freeze because they are paging in and out. The users could not do anything, even to just move the mouse from one window to another.

I ran some performance statistics using sar, ps and vmstat. On one of the machines which is a SPARCprinter II server, I was convinced that there was a memory shortage. So I added another 32MB RAM expanding the total RAM to 64MB. It still slows down when it's printing, but the system is able to recover from the paging activity unlike before when I had to reboot the system. In this case, I did not think of a possible kernel memory leak, so I did not collect sufficient statistics for analysis (sar -k). Here are the statistics output and notes on the analysis:

(extensive tables and statistics deleted)

I believe that all machines displaying the above-mentioned behavior are experiencing some kind of memory leak somewhere. I tried to set maxusers=40 in /etc/system, but that did not work. I would try setting shared memory (shmsys), but I could not find any helpful hints on it. Please help.

I appreciate any comments or thoughts on this issue.

--Cindy Doehr, (firm indeterminate)

It sounds like a kernel memory leak bug, have you tried loading the latest set of kernel patches for Solaris 2.4?

Adrian:

Most of your article on performance tuning is just fine, but I disagree when you state that so long as applications fit, swap size doesn't affect performance. I used to think that, too, but then I learned about fragmentation, and observed user complains going away when I increased their swap.

Also, you mention ncsize and ufs_ninode. These are discussed in Sun's old performance tuning overview. I'd be interested in an updated discussion of these as pertains to SunOS 5.3-5.

--Anthony D'Atri, (firm indeterminate)

The only mechanism I can think of is that the pageout swapfs clustering could be limited if free swap space becomes fragmented into small blocks, this might delay the availability of free pages a bit, but if you are paging in and out at the same time to an overloaded disk you are already in a slow performance mode. Adding swap space on a separate disk helps.

Regarding ncsize and ufs_ninode, this is discussed in the book, and I have touched on it a few times in my Web articles.

From further discussion it seems that your experience of performance improvement may have been on SunOS 4.x systems. Solaris 2 has a completely different swap space layout policy that avoids this problem as far as I can tell.

Adrian:

Just wanted to drop you a quick email to add to your no doubt groaning emailbox, to thank you for the articles on performance on the web.

I am currently a system admin. at a small internet provider (but growing fast) in the UK, which means I do everything ;-)

Our Web server, sparc 20, sol 2.5, has just died performance wise over the last week.... (seem to be loads of time_waits) and whilst doing everything else to keep the provider running, Ive got to look at it urgently before all our customers leave :-(

Anyway, I asked for a list of /etc/system parms and was recommended to read your articles.. I've only just started going through them, but already I can see I have to order your book ;-)

I can see this week will be spent reading everything you've written, that I can get my hands on, to try and solve this problem, so thought I should at least email you to say thanks for the info.

--Keith Pritchard, (firm indeterminate)

Adrian:

Currently we have an installed base of Sun SPARC 4/110, 4/330 and 4/75s. I was wondering whether running Solaris 2.4/5 is compatible with these old hardware platforms, and what are the performance implications by running just Solaris and/or any simple X-based application.

--Ioannis M. Kyratzoglou, Mitre

Solaris 2.4 is the last release that works on a 4/110 or 4/260 or 4/330, or 4/490 -- we have upgraded our lab systems to 4/600 CPU boards with SuperSPARC modules that will run the latest OS.

All others run 2.5, the latest releases are faster and smaller than older Solaris 2 versions. Solaris 2.5 uses more RAM than 4.X, but most things are faster than 4.X, a few operations are slower because extra functionality has been added.

If you have enough RAM I'd upgrade them, if any are marginal don't bother. 32 MB is probably the minimum: you can't waste CPU cycles paging on a slow system, so you need more RAM to keep up with more recent hardware.

Adrian:

I have heard and read conflicting views of using SSA NVSIMM and Presto NVSIMM items together. While I understand the configuration problem of putting one logically before the other in order to have an orderly recovery after a system crash, I still am not clear that using a Presto-NVSIMM with an SSA-NVSIMM gives you anything over just the SSA-NVSIMM alone.

--Kerry P. Boomsliter, Knight-Ridder Information

The bandwidth to/from Presto NVRAM is perhaps 10 times that of SSA NVRAM. The CPU does more work with Presto as the data is copied to and from the Presto NVRAM, but doesn't bottleneck on the disk interface like the SSA NVRAM.

That is the main difference. Either one gets you most of the performance boost by changing disk accesses into memory accesses, but the higher bandwidth and capacity of Presto gets better response times, and the SSA NVRAM has higher throughput. Both together is the best option for maximum performance. The recent introduction of 16MB SSA NVRAMa (up from 4MB in older SPARCstorage Arrays) also helps performance for write-intensive applications.

Adrian:

(Your) performance tuning articles are excellent, but I wonder if anyone has any performance tips or gotchas for the vaguarities of the Intel platforms running Solaris x86? What special concerns come up on those hardware platforms?

--(name and firm indeterminate)

Adrian:

Is there a Solaris 2.4 kernel tuning parameter (like maxuprc) that would allow sysadmins to stop unfriendly programs from taking over a system? The problem we have sometimes seen is a poorly written program forking off infinite copies of itself until the machine dies or hits its process limit. We want to be able to limit a user's total to, say, 100 processes.

Is this possible under Solaris 2.4?

--Lance Nakata, Stanford University

The same maxuprc variable does this for you in Solaris 2.X

set maxuprc=100 in /etc/system and reboot

RAID feature feedback

I'm attempting to assist a client in baselining their current server environment. The client is not particularly technical, so I'm having to piece together a lot of my information. The topic of RAID has come up, and the client believes they have three 1-Gig drives with 2-Gigs of useable space and RAID-5. I thought I remembered something about RAID-5 requiring a minimum of four drives. Am I right? Is there an easy way to determine which RAID level is installed given the limited information the client provided about the number and size of their drives in combination with the amount of useable space?

--Steve Taylor, (firm indeterminate)

If they have three 1 GB disks but 2 GB usable, their overhead is 1/3rd, which is consistent with a RAID-5 configuration. In RAID-5, the overhead is the amount of disk space associated with one member disk, while other forms of RAID have differing overhead. For example, RAID-1 (mirroring) would be 50 percent overhead.

On the other hand, this could be a RAID-3 or RAID-4, since these have the same amount of overhead as RAID-5 -- the difference is where the parity information is located. Generally speaking, though, all commercial RAID-3's are actually RAID-5's in implementation since virtually all disks today are SCSI disks that do not permit partitioning at the sub-block level.

RAID-5 requires at least three disks. Although in principle you could build a two-disk RAID-5 volume (one data, one parity), this is much more work than simply mirroring, and provides advantages.

Editors:

I am a lost soul from New Zealand. Could you kindly help me with RAID-5?

1. I am setting up RAID-5 system (14 disks) on a Sun Sparc 20. Is the logging disk required?

2. if I don't use a hot swap, when a disk crashed, will the RAID system still be operable in terms of reading and writing to it?

--John, (firm indeterminate)

1. I presume from this question that you're using SPARCstorage Volume Manager on a SPARCstorage Array. The log is not required, but it's certainly a good idea. The case it covers is the one in which the failure of a disk causes a system crash. It's not supposed to happen, but lots of impossible things happen in computer systems. In that case, it is possible for the parity blocks of in-flight writes to be inconsistent with the data blocks on the disk, and without the log, you'll never be able to tell which is correct and which is not. You do need one and a half failures before you lose sleep, but it's something that Sun normally recommends except in unusual circumstances.

2. The whole point of RAID-5 is to provide data protection in the face of a disk failure. If you have a 5+1 RAID-5 volume (e.g., five data, one parity, one log) volume, you can sustain the failure of any single disk and have absolutely no interruption in service or data availability. (See my article for an explanation of why this is true.) this is similar to what happens in mirroring, except that you only have one or two disks worth of overhead in RAID-5, rather than five disks (in this case) for mirroring. Of course, there's a price to pay: RAID-5 writes are much slower than writes to a mirror.

   Hot swap and hot spare are two quite different things, and neither of them come into play until after some sort of storage protection (RAID-5, RAID-1 -- i.e., mirroring) is used to recover the data in the event of a disk failure. You don't need either one to improve your single-failure resistance. Hot sparing is used to improve the mean time to repair a disk failure (by automagically making a copy of the dead disk). Once you have hot sparing, hot swapping is an electrical technique that permits a disk to be physically removed and replaced without shutting down the system.

Brian:

I am faced with a problem in an area I have never dealt with before: we need an extremely reliable, Solaris 2.4 / 2.5 server system with full redundancy for all components. It just must never stop, no matter if a singe disk or processor or etc. may fail.

Do you have any good pointers where to look?

--(name and firm indeterminate)

There are a number of products out there, depending upon just what it is that you're doing. Sun offers Solstice High Availability-NFS and HA-Oracle, with future plans for Sybase and Informix. These all run either 2.4 or 2.5, in various states.

There are third-party products too, from companies like OpenVision.

Brian:

Thanks for your RAID article in SunWorld Online. It apparently aimed at dispelling some RAID myths, and it worked well for me. However, I still think my "mythological" view of RAID5 sounds pretty reasonable, and I wonder whether it is also another flavor of RAID or just a misunderstanding on my part.

I thought the granularity of RAID 5 was much finer, in the order of bits or bytes per block. So every write and every read would use all the disks. If this were the case, writing would not require reading, as all the parity info could be computed from the data to be written.

Of course, this would force all the disks to work in unison, with all heads moving in sync. The only advantage would be in throughput, not in head positioning. But it would still speed sequential operations.

Is this an available RAID option?

--Benjamin, (firm indeterminate)

Under this organization, the chunk size would be 1 byte or 1 bit. Back in the days before SCSI, this is in fact how RAID-3 and a few commercial mainframe RAID-2s were implemented. But today virtually all disks are SCSI disks, and SCSI defines a disk as a device that responds to requests for 512-byte blocks numbered from 0 .. n - 1. Therefore any RAID built out of SCSI disks necessarily has a minimum chunk size of 512 bytes, and once you get to that level you might as well choose something that suits your application.

Spindle sync is a controversial option. It makes the most obvious benchmark (reading and writing a single big file from disk) go faster -- a lot faster, actually. However, because the disks run in sync, they are now the logical equivalent of ONE disk, and the whole point of RAID-5 over RAID-3 (better random access performance) is destroyed.

A few RAID-3's continue to offer this; it's quite useful in RAID-3 because it reduces overall latency, especially for very wide stripes or RAIDs. But it kills RAID-5, and since virtually all RAID-3s are actually RAID-5s now, it isn't available very often.

CacheFs and AutoClient feature feedback

Editors:

How does (feature writer Brian Wong) derive the "NFSops"? (Number of NFS operations per second?) What tools are out there? Suggested usage and interpretation... This seems like the basis of another article: How to measure NFS / Client-Server / bootp / LAN-WAN traffic and the overall impact to the "system" (memory, CPU, response time, etc.). Or is this an old topic which I simply have overlooked/missed? Seems like this is an ongoing battle everywhere I have worked. I regularly use ping, traceroute and netstat; but it seems that there must be some "intermediate" level of tools and/or knowledge that allows old hacks like me to have some better clues as to what is going on in Ethernet.

--Steve Walker, Bell Atlantic

You have overlooked a couple of commands that are built into Solaris/SunOS. Specifically nfsstat(1) and snoop(1m). In particular, nfsstat shows the operation mix. With judicious zeroing of the counters it can also be useful in determining overall request or service rate. snoop can be used to collect more specific data about individual conversations, and there are whole products (such as AIM sharpshooter and HP NetMetrix) that are built upon the basic snoop-type functionality. (They also set the interface in promiscuous mode and watch every NFS packet.)

NFS version 3 feature feedback and question

Reading your article in the October SunWorld Online, I would like to pose a question.

We have seen miserable throughput using two Sun20's across 155Mb/s ATM using Solaris 2.4 and NFS. We have never bettered 20Mb/s and all the relevant vendors of hardware assure us and have reasonable evidence that the equipment is not to blame. The consensus is that NFS itself is not capable of delivering better than 20Mb/s in a real disk to disk transfer.

The word on the market is that while Solaris 2.5 and the "re-written" NFS will be better, the improvement is in the order of 15-30%, whereas we are is search of a climb from 20Mb/s to 80-120Mb/s. At this time Microsoft have announced that Cairo will support ATM native in the mid 96 release, I cannot believe that Sun would lag behind Microsoft by too much! Do you have any ideas on an NFS functional code which will bring the performance to ATM levels?

--Matthew Loxton, (firm indeterminate)

20 Mbits/sec = 2.5 MB/sec. You can definitely do better than this, although NFS Version 2 is not going to do a LOT better. If you have at least 32 nfsds on your server, add the following to your /etc/system on the CLIENTS:

set nfs_max_reads=16

The improvement from Solaris 2.4 to 2.5 is about 10-15% on the same test, but 2.4 and 2.5 has different inherent capability.

Solaris 2.5 and NFS Version 3 can go faster, but not as fast as you'd like. If you add

#set nfs:nfs_max_threads=6
set nfs:nfs3_max_threads=6
#set nfs:nfs_nra=6
set nfs:nfs3_nra=6
set nfs:nfs3_max_transfer_size=65536
set nfs:nfs_maxasynccount=256

to your /etc/system you'll get about 5.7 MB/sec (45 mbits/sec) between two SS20-71's on Fast Ethernet, and somewhat better than that using ATM.

Support for, and performance of, are two different things. The limitation in Solaris 2.5/V3 is the speed of the cpu, not the protocol or the ATM interface. Between two Ultra-1/140's we've gotten 8.6 MB/sec (68 Mbits/sec). I haven't measured it but I'd expect about 100 Mbits/sec between 200-MHz Ultras.

Client/server column comments and questions

Your introduction on Web client/server is interesting, but I think of other possibilities like Sybase web.sql (you mentioned Oracle's Wow) and also Java applets at the client side, which may be more flexible & powerful than JavaScript or other LiveScripts for real clients (i.e., not too dumb clients). Someone is also working on a promising SybJava tool, it's available, but very uncomplete today (to my eyes).

But, one thing is sure: this kind of technology is going to invade the market of three-tier client/server. The Web really needs more interactivity with real world apps.

--Bernard Van Haecke, UCL - University

Editors:

Noticed that there was no mention of Web Objects and the accompanying EOF from NeXT. I'd appreciate your comments on the exclusion of these products as they seem to be germain to the subject.

I am currently faced with the situation as described and at least for the moment, am testing Perl and sybperl. Interesting position to be.

Good article on a subject that needs airing so that we may be able to get some really good and easy-to-use products.

--Tom, (firm indeterminate)

Editors:

I saw recent letter-to-editor asking about Java and C code sizes. I'm finishing up translation from C to Java of one of our internal molecular modelling feeder programs and those numbers are

• C source 5800 lines
• C *.o files (-O) total 110 kbytes
• Java source (flatfooted tenderfoot translation) 5100 lines
• Java *.class files (default javac, Beta 2 JDK) 73.3kbytes

--(name and firm indeterminate)

Editors:

I was dismayed to see that products such as Spider (Spider Technologies) and products from Bluestone, VPE, ... were not even mentioned in the article. I was also surprised that the author failed to mention about LiveWire and LiveWire Pro from Netscape Communications.

Who is interested in looking at stuff that was done back in 1992?

For a magazine that is on-line should reflect the technology/products that are available in the marketplace (public domain as well as commercial).

--(name and firm indeterminate)

Security column answers

I am a neophyte Unix sysadmin. Thanks for your security column, it's great!

Where is the best source for a comprehensive overview of Unix security for a novice? Most of what I have read so far assumes a great deal of knowledge. How about a overview of Unix administration for the novice!

--Kyrie, (firm indeterminate)

For book recommendations, I suggest the Nemeth Et Al Unix Administration Handbook for general administration stuff. To get started in security, check out the O'Reilly book series, including Practical Unix Security (to be updated soon), and Computer Security Basics.

Peter:

I enjoyed your article, especially since Trusted Information Systems is mentioned in a number of places. Just a few corrections.

1. One cannot purchase support for the TIS Internet Firewall Toolkit (FWTK) from us, or anyone. It is supported -- on a "best effort" basis -- via electronic mail to fwtk-support@tis.com.

2. The firewalls mailing list is not really the best place for FWTK support. The address above is, and there is also a users list -- fwtk-users@tis.com.

3. People should be aware that the FWTK is not a complete firewall, in that they should also employ kernel modifications (such as we do with our Gauntlet Internet Firewall) and/or a screening router. The proxies alone will not protect against IP spoofing, source routed attacks, etc. And, of course, if they are not experts they are much better off with a commercial firewall -- anyone's commercial firewall -- rather than thinking that all they need to know is how to type "make" at the Unix prompt (and believe me, some people have written to us and weren't even sure what that means).

--Trusted Information Systems

Sysadmin column off the mark

You mention crypt and pgp as adequate security tools in certain situations. Crypt is not adequate. It is crackable using brute force methods in reasonable amounts of time using inexpensive (i.e., desktop PC) equipment.

--(name and firm indeterminate)

Do you enjoy protection from the US Secret Service? The term "secure" is relative. For some settings, the Unix crypt command offers enough protection. For others, PGP and other measures are appropriate, just as most of us go through life unescorted while heads of state live with armed bodyguards.

See Hal Stern's Sysadmin column this month for more about crypt and PGP.

Connectivity column lacked detail

Your review (of SoftWindows and WABI and SunPC) should have included WINE.

--(name and firm indeterminate)

Rawn:
An interesting general article on emulators on Sun. however, you missed a few points.

1. Wabi 2.1 supports (in a limited fashion) WFW.

2. Wabi & MAE give the user sound. SunPC and Insignia do not.

Yes, this is important to some users. I actually would have bought SoftWindows 2.0, if they had Soundblaster emulation. but they don't, so i didn't. (and bought MAE 2.0 instead :-)

--(name and firm indeterminate)

The column was meant as a general overview, not point-by-point review. SunPC and SoftWindows support simple audio, but don't emulate the SoundBlaster.

Praise & question for connectivity columnist

Your articles on PC-to-Unix interconnection is very helpful and informative. I am connecting Windows NT 3.5 to a few Unix operating systems such as, AIX, HP-UX, OSF/1, SunOS, and Solaris. I am using Beame & Whiteside NFS Client on the NT machine and pcnfsd on the Unix OS. I am also using an X server (Hummingbird eXceed) from Hummingbird Communications for X Window display. The interconnection is a success.

As for e-mail, I can run a text base mail package on the Unix machine through terminal emulation programs or an XDM (X Display Manager). Another way to setup mail is to use Beame & Whiteside Mail for Windows and an SMTP and a POP2/3 server on the Unix machine. Currently, I run BW mail on the NT and University of California POP3 server on the SunOS. The e-mail works seemlessly. However, I would like to find a POP3 server for AIX, HP-UX, OSF/1 operating system.

--Tom Ceman, (firm indeterminate)

From our reader support dept.

I am just curious about what tools are used to put together this magazine. You seem to have one of the most comprehensive magazines around. What do you use to manage it all? How do you collaborate to put it all together? Thanks.
--(name and firm indeterminate)

Thank you for your kind words. The editors use their favorite text tools (vi, textedit, etc), and we gin-up the html pages using templates for each section (i.e., news, features, columns). We've tried a few html word processors, and found them too slow, buggy, and rigid.

We worked together before on a paper-based magazine, so we reused a simple copy-flow scheme.

Editors:

Good mag.

Although I prefer to be able to take a mag home, put my feet up and peruse in a more comfortable, relaxing atmosphere, there are things that make online mags nice. I have put many of my favourite columns my netscape bookmark list -- kind of like cutting your favourite articles out of all those old magazines before pitching them in the garbage.

Performance Tuning by Adrian Cockcroft ranks right up there as the best in my opinion.

Keep up the good work.

Editors:

Is it true? Has Sun bought Apple Computer? You have an answer for me.
--Andrew, (firm indeterminate)

Editors:

Please be nice to the Macintosh. I love the Mac. I use PC's, Unix, but I love the Mac. Use Apple and the Macintosh for good, not evil.
--Scott Flowers, AMMG, Inc.

More about Java

I work for a financial software company and we are examining how we might use Java to front-end our server systems. Also we might have a requirement to locate people versed in writing Java apps. How does one (learn more about) these types of issues? Are there relevant news groups? thanks in advance.

--(name and firm indeterminate)

Editors:

If 80% of all webmasters use the Macintosh platform, why are the Mac java tools always the last to be developed?

--(name and firm indeterminate)

It's unfortunate many organizations have simply dumped their paper-based marketing collateral on their Web servers. No doubt much of this originated from Quark and FrameMaker files in Macintoshes. Calling this repurposed content Macintosh-originated is true but nothing to brag about.

That said, I think Java ISVs that ignore the Macintosh forfeit opportunity. --Mark Cappel, editor-in-chief

Editors:
Is there an on-line magazine just for JAVA freaks? If not, there should be.

--Jim, (firm indeterminate)

We agree there should be a magazine for Java freaks. That's why we're launching JavaWorld!

JavaWorld will be the first publication to serve the rapidly growing Java community. The independent monthly Web-only publication will feature hands-on tutorials for both novice and advanced Java programmers as well as profiles of businesses that use Java for key applications and coverage of Java-related news and events. JavaWorld will also focus on the business-related information needs of the Java community.

JavaWorld's first issue will be available February 15. Keep reading SunWorld Online (and subscribe, if you haven't already!) to make sure you know where and when to find our fulfilling dose of Java how-to articles, news, and other stuff that's bound to satisfy.

You can also send an e-mail to jwinfo@javaworld.com to keep informed.

Feel free to send your suggestions my way.

If you have problems with this magazine, contact webmaster@sunworld.com

URL: http://www.sunworld.com/swol-02-1996/swol-02-letters.html
Last updated: 1 February 1996

If you have technical problems with this magazine, contact webmaster@sunworld.com

URL: http://www.sunworld.com/swol-02-1996/swol-02-letters.html
Last modified: