Re: Virtual I/O Caching Alpha/Oracle
Date: 1996/01/29
Message-ID: <4eioeq$tog_at_shore4.intercom.net>#1/1
In article <slrn4ghkes.jc039g.AI26_at_a400.sternwarte.uni-erlangen.de>, AI26_at_a400.sternwarte.uni-erlangen.de (Lemke) says:
>
>On 26 Jan 96 09:55:55 GMT, Alan Greig <agreig_at_river.tay.ac.uk> wrote:
>>In article <NEWTNews.822594935.7964.Bob_at_bobmc.gtionline.com>, Bob McGuire <Bob_at_Alpha> writes:
>>> Greetings netters.
>>>
>>> Has anyone out in Netland attempted to change the
>>> VCC_MAXSIZE parameters on an Alpha from the default of
>>> 3200 to something greater? I'm trying to get the Alpha
>>> running as best as possible with Oracle.
>>
>>On a general multi-user system we see a read hit rate of around
>>90% with a 50MB cache:
>>
>> System Memory Resources on 26-JAN-1996 09:50:41.02
>>
>>Physical Memory Usage (pages): Total Free In Use Modified
>> Main Memory (512.00Mb) 65536 23467 37827 4242
>>
>>Virtual I/O Cache
>> Total Size (Kbytes) 50000 Read IO Count 5953058
>> Free Kbytes 12648 Read Hit Count 5355658
>> Kbytes in Use 37352 Read Hit Rate 89%
>> Write IO Bypassing Cache 7268 Write IO Count 457392
>> Files Retained 91 Read IO Bypassing Cache 252048
>>
>>>
>>> If so, what have you changed the value to, and if there
>>> were changes, how were they, good/bad/no change or simply
>>> not worth the effort.
>>
>>On an system with 3.2 MB (default) I typically see a hit rate
>>of under 50% if there's a lot of activity. So it is worth experimenting
>>at least.
>
>
> Sorry about this followup but we are rather ignorant here (our
> doc wall is grey). If I do a $ sh mem/cach/fu I get:
>
>A400> sh mem/ca/fu
> System Memory Resources on 26-JAN-1996 12:48:58.13
>
>Virtual I/O Cache
> Total Size (Kbytes) 3200 Read IO Count 0
> Free Kbytes 3200 Read Hit Count 0
> Kbytes in Use 0 Read Hit Rate 0%
> Write IO Bypassing Cache 0 Write IO Count 0
> Files Retained 0 Read IO Bypassing Cache 0
>
>A member of this cluster does not have Virtual I/O Cache capability.
>A400>
>
> To me this seems to mean our VIOC is not working at all. Is this
> because we failed to enable if (if so, how do we enable it?) or
> because we are running a mixed cluster (the 5.5 systems are
> VAXen, the rest Alphas):
>
>A400> sh clus
>View of Cluster from system ID 1424 node: A400 26-JAN-1996 13:00:18
>-------------------------------
>| SYSTEMS | MEMBERS |
>| NODE | SOFTWARE | STATUS |
>| A400 | VMS V6.2 | MEMBER |
>| SAT2 | VMS V5.5 | MEMBER |
>| SAT1 | VMS V5.5 | MEMBER |
>| ALF001 | VMS V6.2 | MEMBER |
>| M3300 | VMS V5.5 | MEMBER |
>| ALF002 | VMS V6.2 | MEMBER |
>-------------------------------
>
>
> Thanks for any hint,
> Michael
>
The VAXes will have to be running 6.1 or better to enable VIOC. Received on Mon Jan 29 1996 - 00:00:00 CET