7. About the Third Cache-Off

7.1 What is the timeline for the Cache-Off?

Please see our deadlines page.

7.2 Where is the Third Cache-Off being held?

The facilities have been provided by the Compaq Corporation. Here is the address:

Attn: TMF Cache-off
27816 Parkway Plaza (SH249)
Tomball, TX 77375
USA

See the logistics page for additional information, including nearby hotels.

7.3 For soft entry into the third cache-off event, you state that you require an auto-generated Polygraph report. What does that mean?

It does not refer to the binary log files that Polygraph generates while it is running. What we need is the report generated by the make_report command. See the ReportGen documentation for more information.

The report should be generated based on the client side log(s). Please keep all logs around if possible. We might ask you for the logs later if needed.

7.4 Why do you require the auto-generated report?

For two reasons:

1. We want to make sure that you've run Polygraph before coming to the Cache-Off.
2. We want to make sure that your expectations are not too high.

In the past, some participants thought they could achieve higher throughput than was really possible. They spent too much time trying for the high throughput and did not have any successful runs.

In the past, we have seen many problems when a vendor would prepare for the test using *what they thought* an equivalent but "faster" workload. If such a vendor fails a full PolyMix-3 test, it would complain to us that "it worked perfectly in our lab at home!" There was no way we could verify those claims. In most cases where verification was possible, we would find significant changes to the workload or testing environment.

7.5 Cache Filling

With PolyMix-3, are we still filling the cache with 2 x physical disk size?

Due to many objections, we are using twice the configured cache size now. The configured cache size includes "configured disk cache capacity" plus total RAM size.

7.6 Participation Fees

The Third Cache-Off is being organized by The Measurement Factory.

Please make checks payable to:

The Measurement Factory, Inc.
PO Box 3320
Boulder CO 80307

Participation fees are calculated as follows:

The cost to participate is proportional to the number of Polygraph client/server pairs that an entry requires -- $2,000 per pair.

The minimum amount is $2,000, for one pair.

Each entry must pay a non-refundable deposit of $2,000 by July 10, 2000. The deposit is a credit to your final payment amount.

The total participation fee depends on product peak request rate and on the time of payment as detailed by the rules below:

1. Maximum per-client load level is set at 400 req/sec
2. Each participant must declare maximum request rate when registering (R)
3. At the cache-off, a participant is guaranteed to have enough client-server pairs to generate R
4. Extra pairs might be available, but are not guaranteed
5. If full registration fee is received by Aug 18th, the fee is $400 * [ R / 100 ] with a $2,000 minimum.
6. If full registration fee is received after Aug 18th, the fee is $500 * [ R / 100 ] with a $2,000 minimum.
7. Entries that are not paid in full by Sep 18th may not be tested
8. The value of R is not reported; only the actual achieved rate is reported

A table is provided below for your convenience (but you may want to double check our calculations):

   Registered         Total payment
   request                 ($) 
   rate (R)    by 8/18    by 9/18   cacheoff#2  
   ----------  -------   --------     -----
         1       2000       2000       2000
       100       2000       2000       2000
       400       2000       2000       2000
       500       2000       2500       4000
       600       2400       3000       4000
       700       2800       3500       4000
       800       3200       4000       4000
       900       3600       4500       6000
      1000       4000       5000       6000
      2000       8000      10000      10000
      3000      12000      15000      16000
      4000      16000      20000      20000
      5000      20000      25000      26000
      6000      24000      30000      30000
      7000      28000      35000      36000
      8000      32000      40000      40000
      9000      36000      45000      46000
     10000      40000      50000      50000
     12800      51200      64000      64000

The second cache-off figures (last column) are provided for comparison only. We wanted to illustrate that we are trying to address concerns of vendors that want to bring large entries as well as small ones.

7.7 When are the results going to be made public?

Please refer to the Cache-off schedule.