home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.protocols.nfs:2727 comp.sys.sun.admin:8099
- Path: sparky!uunet!sun-barr!cs.utexas.edu!swrinde!sdd.hp.com!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!agate!doc.ic.ac.uk!uknet!mucs!pq
- From: pq@cs.man.ac.uk (Paul Quare)
- Newsgroups: comp.protocols.nfs,comp.sys.sun.admin
- Subject: SUMMARY - eNFS/Cache
- Message-ID: <6701@m1.cs.man.ac.uk>
- Date: 9 Nov 92 20:21:03 GMT
- Sender: news@cs.man.ac.uk
- Followup-To: comp.protocols.nfs
- Organization: Dept Computer Science, University of Manchester, U.K.
- Lines: 191
-
- A couple of weeks ago I asked for comments from sites using eNFS/Cache from
- Interstream Inc, I didn't receive many replies but here is what I did get.
- I haven't included the straightforward "me-toos". I've put some contact
- information etc at the end. Thanks to all who replied.
- Paul Quare
-
- Original posting
- ----------------
- >There is a product eNFS/Cache from Interstream which will cache read side
- >NFS requests on a local disc. We have recently upgraded many of our clients
- >to have a small (105M) disc for root/swap and a small amount of software.
- >eNFS/Cache looks to be a useful product in that it doesn't require any
- >intervention by the system adminstrator in terms of decisions about what to
- >cache (real usage dictates that) and removes the possible need to alter the
- >filestore layout to reflect what you might decide to store on the local disc
- >rather than access via NFS.We have obtained an evaluation copy from Interstream
- >and it seems to perform without any problems,(on a single machine).
- >
- >Is there anyone who is using this product in a large network who can comment
- >on its performance/reliability etc. I will summarise to these newsgroups if
- >you email me. I understand that Sun may have a similar product in a release of
- >Solaris2.1 for later next year,one of the factors affecting whether we purchase
- >will be the date at which we upgrade these machines (in teaching laboratories)
- >from SunOS4.1.X to Solaris2.X,we have a limited budget!
- >
- >Paul Quare
- >pq@cs.man.ac.uk
-
- --------------------------------
- From: petsche@hawk.siemens.com (Thomas Petsche)
-
- PLEASE forward any information you get about eNFS/cache. I asked
- about this a while ago and didn't get any feedback (other than
- this sort of note asking a copy of information).
-
- BTW, the announcement I saw about enhanced NFS from Sun said that
- some of the enhancements would be available next year, but I don't
- think local caching was one of them. If you have more solid
- information, I appreciate it.
-
- Thanks!
-
- Tom
-
- [[See [1] below]]
-
- ----------------------------------
- From: David Barr <barr@pop.psu.edu>
-
- I'd be very interested in the results of your summary. I'm
- currently evaluating AFS, which also has a local cache. You might
- also want to check out IFS, which is an extension to AFS. Send mail
- to info@citi.umich.edu. I've looked at eNFS briefly, but haven't the
- guts to try out the demo yet.
-
- --Dave
- --
- System Administrator, Population Research Institute barr@pop.psu.edu
- loose: v. to set free, or adj. not securely fastened.
- lose: v. to miss from one's possession.
-
- [[see [2] below]]
-
- ---------------------------------------
- From: sdi@sun-valley.Stanford.EDU (Steven D. Ims)
-
-
- Paul,
-
- We also evaluated the eNFS/Cache product. Our network is about 16
- machines (SS1, IPC, and SS2) with a Sun4/370 server (SCSI disks) and
- clients. The typical load for our server during the day is about 2.5
- to 3.
-
- We didn't see much noticeable improvement in performance due to the
- eNFS/Cache, but we attribute that to the following: (1) relatively low
- load on the server and (2) access time on the client disks and the
- server disks are comparable.
-
- However, I would expect that if the server load were higher or the
- clients' disks faster, then eNFS could make a noticeable impact.
-
- I'd be interested to read a summary of the responses that you get
- about this product. Thanks.
-
- -- Steve
- sdi@sun-valley.stanford.edu
-
- [[see [3] below]]
- ----------------------------------------
- From: Barry Margolin <barmar@Think.COM>
-
- We had some problems that may have been a result of eNFS/Cache. Some of
- the clients using it were getting corrupted data.
- --
- Barry Margolin
- System Manager, Thinking Machines Corp.
-
- barmar@think.com {uunet,harvard}!think!barmar
-
- [[I didn't see anthing of this nature,I was using version 1.0.1, I used it
- to compile X11R5 a couple of times and had no problems with that.]]
- ----------------------------
- From: David_Boyd@sparky.IMD.Sterling.COM (David Boyd)
-
- In article <6515@m1.cs.man.ac.uk> you write:
- >Is there anyone who is using this product in a large network who can comment
- >on its performance/reliability etc. I will summarise to these newsgroups if
- >you email me. I understand that Sun may have a similar product in a release of
- >Solaris2.1 for later next year,one of the factors affecting whether we purchase
- >will be the date at which we upgrade these machines (in teaching laboratories)
- >from SunOS4.1.X to Solaris2.X,we have a limited budget!
- >
- Although we are not using it on a large network we have done some
- evaluations of its effect on network load. The first thing we noticed is that
- if the response time from the server is fast you acually suffer a small
- performance penalty. When reading lots and lots of image data over the
- network I found it could reduce actual transfers by quite a signigicant
- amount.
-
- Now, a question. Where did you here that Sun may be introducing
- a similar product? This is the first we have heard about it. What is the
- projected availability and cost if you know. Thanks.
- --
- David W. Boyd UUCP: uunet!sparky!dwb
- Sterling Software IMD INTERNET: dwb@IMD.Sterling.COM
- 1404 Ft. Crook Rd. South Phone: (402) 291-8300
- Bellevue, NE. 68005-2969 FAX: (402) 291-4362
-
- --------------------------------------------------------
- From: Mike_Wertzberger@sparky.IMD.Sterling.COM (Mike Wertzberger)
-
- I have eNFS/Cache. I found it to reduce my network load in a local area
- network by 40%.. My current configuration is a SUN630MPx2 server with SUN
- SPARC 2 clients. I have configured my clients with a 100 MB cache directory.
- My intent is to migrate this all to a very wide area and see what happens.
-
- I can only say good things about Interstream, I found them to be very
- responsive to questions, and open to ideas for future technology.
-
- The current products from sun I dont think do the same type of work, they do a
- NFS write accelleration, and that is not a lot of help for my situation.
- --
- Mike Wertzberger UUCP: uunet!sparky!wertz
- Sterling Software IMD INTERNET: wertz@IMD.Sterling.COM
- 1404 Ft. Crook Rd. South Phone: (402) 291-8300
- Bellevue, NE. 68005-2969 FAX: (402) 291-4362
-
- [[ I feel that there are several products addressing different aspects of the
- "problem" eNFS & Prestoserve for NFS writes, Interphase boards for network
- throughput, eNFS/Cache for client side read performance.]]
- -------------------------------------------------------
- --------------------------------------------------------
- Notes
- -----
- [1] Sun caching filesystems.
- I can only pass on what I was told as unattributable/unsubstantiated rumour:
- Sun may provide a caching filesystem in Solaris2.2 or 2.3. - I heard that
- it might be in the 4/93 release.(Nothing was mentioned about 4.1.X systems).
- Perhaps Sun will disclose something amongst their Nov 10 revelations.
-
- [2] IFS/AFS
- Re IFS:
- "Many of our technical reports are available online via AFS or anonymous
- ftp to citi.umich.edu in /afs/umich.edu/group/itd/citi/public/techreports.
- We also have an IFS Overview document available in postscript via the
- same access methods but in /afs/umich.edu/group/itd/citi/public/ifs"
- Re AFS:
- This is marketed by Transarc:
- afs-sales@transarc.com
- Tel: 412-338-4432
-
- [3] Performance
- As I only installed it on one client I found it hard to extrapolate what
- the overall performance on our network would be. It is reasonable to assume
- that a fast NFS server would respond quicker than a client's local disc for
- one client but what would N clients produce in throughput?? Generally I observed
- things going slower as the cache was loaded and then varying speed changes but
- accompanied by a fall in network packets.It was quite easy to construct
- tests like doing a sum on the files within an NFS mounted filesystem that
- showed things in a favourable light!
-
- ------------------------
-
- Interstream contacts
- Interstream Inc, 1501 Reedsdale Street, Pittsburgh, PA 15223-2329
- Tel: 412-323-8000
- Fax: 412-323-1930
- slf%interstream.com@netserver.andrew.cmu.edu (Stacy Fuechslin International
- Account Representative)
- bjd%interstream.com@netserver.andrew.cmu.edu (Bruce DaCosta President/CEO)
-