Contents   Index

Tests

The test environment for use of TACO has already been described, see Section 6.1.1. Weakly connected mode were tested using a 14400B modem. Another environment used in order to evaluate $\mathcal{P}$$e$$\mathcal{S}$$t$$\mathcal{O}$ without use of TACO has been with two hp9000s700 machines in DIKU's net; embla.diku.dk as server and skade.diku.dk as client. In these environments I have performed the following tests (on the files listed in Table 7.1):

1. Session: Reading - fully connected.
   1. Start with an empty cache.
   2. Communications status is fully connected.
   3. Open all the files for reading strict and close them.
      The files are not cached. Files are fetched from the server. This measures the overhead compared with TACO for fetching when fully connected, and compared to remote NFS access.
   4. Open all the files for reading strict and close them.
      The files are cached and consistent. Files are acknowledged from the server. This measures the gain of caching compared with TACO for fetching when fully connected.
   5. Communication status is disconnected.
   6. Open all the files for reading optimistic and close them.
      The files are cached and within the consistency time bound. Files are opened locally, and there is no server contact. This measures the overhead compared with local UNIX for reading.
   7. Communication status is fully connected.
   8. Open all the files for reading pessimistic and close them.
      The files are cached and consistent. Files are acknowledged from and locked on the server. This measure the overhead of locking compared with not locking (in one of the above tests) when fully connected.
2. Session: Reading - weakly connected.
   1. Start with an empty cache.
   2. Communications status is weakly connected.
   3. Open all the files for reading strict and close them.
      The files are not cached. Files are fetched from the server. This measures the overhead compared with TACO for fetching when weakly connected.
   4. Open all the files for reading strict and close them.
      The files are cached and consistent. Files are acknowledged from the server. This measure the gain of caching compared with TACO for fetching when weakly connected.
   5. Open all the files for reading optimistic and close them.
      The files are cached and within the consistency time bound. File are opened locally, ant here is no server contact. This measures the overhead compared with local UNIX for reading.
   6. Open all the files for reading pessimistic and close them.
      The files are cached and consistent. Files are acknowledged from and locked on the server. This measures the overhead of locking compared with no locking (in one of the above tests) when weakly connected.
3. Session: Writing - fully connected.
   1. Start with an empty cache.
   2. Communication status is fully connected.
   3. Open all the files for writing strict, write to them, and close them.
      The files are not cached, and they are updated. Files are fetched from and send to the server. This measures the overhead compared with TACO for fetching and sending when fully connected, and compared to remote NFS access.
   4. Communication status is disconnected.
   5. Open all the files for writing optimistic, write to them, and close them.
      The files are cached and within the modification time bound, and they are updated. Files are opened locally, and send to the server, later. This measures the overhead compared with ANSI-C on top of UNIX for writing.
   6. Communication status is fully connected.
   7. Open all the files for writing optimistic, write to them, and close them.
      The files are cached and within the modification time bound, and they are updated. Files are opened locally, and send to the server. This measures the overhead compared with TACO for sending when fully connected.
4. Session: Writing - weakly connected.
   1. Start with an empty cache.
   2. Communication status is weakly connected.
   3. Open all the files for writing strict, write to them, and close them.
      The files are not cached, and they are updated. Files are fetched from and send to the server. This measures the overhead compared with TACO for fetching and sending when weakly connected.
   4. Open all the files for writing optimistic, write to them, and close them.
      The files are cached and within the modification time bound, and they are updated. Files are opened locally, and send to the server. This measures the overhead compared with TACO for sending when weakly connected.

Table 7.1: Files used for testing

No.	File	Size	Type
1	chb.ps	63499	Postscript
2	ctb2.ps	24258	Postscript
3	dataflow.ps	83551	Postscript
4	etb.ps	14812	Postscript
5	files2.ps	38401	Postscript
6	included.ps	49241	Postscript
7	mtb2.ps	24258	Postscript
8	rwc-isde.ps	91006	Postscript
9	rwc-none.ps	83218	Postscript
10	rwc-unde.ps	83218	Postscript
11	so2pesto.ps	83881	Postscript
12	states2.ps	57876	Postscript
13	speciale.log	9890	text
14	speciale.idx	10597	text
15	speciale.aux	20467	text
16	speciale.dvi	278036	DVI
17	speciale.toc	7454	text
18	speciale.lof	1625	text
19	speciale.lot	1675	text
20	speciale.ind	6403	text
21	speciale.ilg	323	text
22	speciale.tex	193688	LATEX 2$\epsilon$
23	pclient.h	41968	C
24	pesto.h	9019	C
25	pinfo.h	3031	C
26	plock.h	4745	C
27	pname.h	1525	C
28	ppclient.h	3873	C
29	ppserver.h	1113	C
30	pserver.c	21376	C
31	ptaco.h	977	C
32	ptacod.c	2730	C
		1317734

  Contents   Index