Tài liệu Apress - SQL Server 2008 Query Performance Tuning Distilled (2009)02 ppt

You can also manage the configuration values for iejoanraniaiknu and i]toanran iaiknu by using the ol[_kjbecqnaÍÊÞÊÌi“ÍÊ̜Ăi`Í«ĂœVi`ÕĂi°Í/œÍÊiiÍÌ iÍVœ˜vˆ}ÕĂ>̈œ˜ÍÛ>ÕiÊÍvœĂÍ Ì iÊiÍ«>Ă>“i

Dynamic Management Views

/œÍ}iÌÍ>˜ immediate snapshot of a large amount of data that was formerly available only

ˆ˜Í*iĂvœĂ“>˜ViÍœ˜ˆÌœĂ]Í-+Í-iẶiĂ͘œÜ͜vviĂÊÍ̅iÍÊ>“iÍ`>Ì>͈˜ÌiØ>ÞÍÌ…ĂœÕ}…Í>ÍÊiÌ of

`ޘ>“ˆV͓>˜>}i“i˜ÌÍۈiÜÊÍ­6Ê®Í>˜`Í`ޘ>“ˆV͓>˜>}i“i˜ÌÍv՘V̈œ˜ÊÍ­Ê®°Í/…iÊiÍ>ĂiÍ

iÝÌĂi“iÞÍÕÊivՏ͓iV…>˜ˆÊ“ÊÍvœĂÍV>«ÌÕȘ}Í>Íʘ>«Ê…œÌ͜vÍ̅iÍVÕĂĂi˜ÌÍ«iĂvœĂ“>˜Vi͜vÍޜÕĂÍ

ÊÞÊÌi“°Í½Íˆ˜ÌÜ`ÕViÍÊiÛiĂ>ÍœvÍ̅iÊiÍÌ…ĂœÕ}…œÕÌÍ̅iÍLœœŽ]ÍLÕÌͽÍvœVÕÊ͜˜Í>ÍviÜÍ̅>ÌÍ>ĂiÍ̅iÍ

most important for monitoring performance and for establishing a baseline

/…iÍouo*`i[ko[lanbkni]j_a[_kqjpano view displaysÍ̅iÍ-+Í-iẶiĂÍVœÕ˜ÌiĂÊÍ܈̅ˆ˜Í>Í

µÕiĂÞ]Í>œÜˆ˜}ÍޜÕÍ̜Í>««ÞÍ̅iÍvՏÍÊÌĂi˜}̅͜vÍ/‡-+Í̜Í̅iÍ`>Ì>͈““i`ˆ>ÌiÞ°ÍœĂÍiÝ>“«i]Í

this simple query will return the current value for Hkcejo+oa_:

OAHA?P_jpn[r]hqa

BNKIouo*`i[ko[lanbkni]j_a[_kqjpano

SDANAK>FA?P[J=IA9#IOOMH CB.,,46Cajan]hOp]peope_o#

=J@_kqjpan[j]ia9#Hkcejo+oa_#

/…ˆÊÍĂiÌÕØÊÍ̅iÍÛ>Õi͜vÍ£xÍvœĂ͓ÞÍÊiẶiĂ°ÍœĂÍޜÕĂÍÊiẶiĂ]ÍޜսÍ˜ii`Í̜ÍÊÕLÊ̈ÌÕÌiÍ̅iÍ

appropriate server name in the K>FA?P[J=IA comparison

/…iĂiÍ>ĂiÍ>͏>Ă}i͘ՓLiĂ͜vÍ6ÊÍ>˜`ÍÊÍ̅>ÌÍV>˜ÍLiÍÕÊi`Í̜Í}>̅iĂ͈˜vœĂ“>̈œ˜Í

about the server Rather than cover them all, I’ll introduce one more that you will find yourself

accessing on a regular basis, ouo*`i[ko[s]ep[op]po°Í/…ˆÊÍ6ÍʅœÜÊÍ>˜Í>}}Ăi}>Ìi`ÍۈiÜ͜vÍ

̅iÍ̅Ăi>`ÊÍ܈̅ˆ˜Í-+Í-iẶiĂÍ̅>ÌÍ>ĂiÍÜ>ˆÌˆ˜}͜˜ÍÛ>ȜÕÊÍĂiʜÕĂViÊ]ÍVœiVÌi`Íʈ˜ViÍ̅i͏>ÊÌÍ

̈“iÍ-+Í-iẶiĂÍÜ>ÊÍÊÌ>ĂÌi`͜ĂÍ̅iÍVœÕ˜ÌiĂÊÍÜiĂiÍĂiÊiÌ°Í`i˜Ìˆvވ˜}Í̅iÍÌÞ«iÊ͜vÍÜ>ˆÌÊÍ̅>ÌÍ>ĂiÍ

occurring within your system is one of the easiest mechanisms to begin identifying the source

œvÍޜÕĂÍLœÌ̏i˜iVŽÊ°Í9œÕÍV>˜ÍʜĂÌÍ̅iÍ`>Ì>͈˜ÍÛ>ȜÕÊÍÜ>ÞÊ]ÍLÕÌÍvœĂÍ̅iÍvˆĂÊÌÍiÝ>“«i]ͽÍœœŽÍ>ÌÍ

the waits that have the longest current count using this simple query:

OAHA?PPKL$-,%

&

BNKIouo*`i[ko[s]ep[op]po

KN@AN>Us]ep[peia[io@AO?

ˆ}ÕĂiÍӇÓÍ`ˆÊ«>ÞÊÍ̅i͜ÕÌ«ÕÌ°

Figure 2-2 Output from ouo*`i[ko[s]ep[op]po

You can see not only the cumulative time that particular waits have occurred but also

>ÍVœÕ˜Ì͜vͅœÜ͜vÌi˜Í̅iÞͅ>Ûi͜VVÕĂĂi`Í>˜`Í̅i͓>݈“Õ“Í̈“iÍ̅>ÌÍʜ“i̅ˆ˜}ͅ>`Í̜Í

Ü>ˆÌ°ÍܓͅiĂi]ÍޜÕÍV>˜Íˆ`i˜ÌˆvÞÍ̅iÍÜ>ˆÌÍÌÞ«iÍ>˜`ÍLi}ˆ˜ÍÌÜÕLiʅœœÌˆ˜}°Í"˜i͜vÍ̅i͓œÊÌÍ

Vœ““œ˜ÍÌÞ«iÊ͜vÍÜ>ˆÌÊ͈ÊÍĨ"°ÍvÍޜÕÍÊiiÍ=OUJ?D[EK[?KILHAPEKJ, EK[?KILHAPEKJ, HKCICN, SNEPAHKC, or L=CAEKH=P?D in your top tenÍÜ>ˆÌÍÌÞ«iÊ]ÍޜÕ͓>ÞÍLiÍiÝ«iÈi˜Vˆ˜}ÍĨ"ÍVœ˜Ìi˜Ìˆœ˜]Í

>˜`ÍޜÕ͘œÜ͎˜œÜÍ܅iĂiÍ̜ÍÊÌ>ĂÌÍܜΈ˜}°ÍœĂÍ>͓œĂiÍ`iÌ>ˆi`Í>˜>ÞʈÊ͜vÍÜ>ˆÌÍÌÞ«iÊÍ>˜`ͅœÜÍ ÌœÍÕÊiÍ̅i“Í>ÊÍ>͓œ˜ˆÌœĂˆ˜}Í̜œÍ܈̅ˆ˜Í-+Í-iẶiĂ]ÍĂi>`Í̅i͈VĂœÊœvÌÍ܅ˆÌiÍ«>«iĂͺ-+Í -iẶiĂÍÓôôxÍ7>ˆÌÊÍ>˜`Í+ÕiÕiʻͭdppl6++sss*ie_nkokbp*_ki+pa_djap+lnk`pa_djkh+omh+

^aopln]_pe_a+lanbkni]j_a[pqjejc[s]epo[mqaqao*iolt®°Í
Ì…œÕ}…͈ÌÍÜ>ÊÍÜÈÌÌi˜ÍvœĂÍ-+Í -iẶiĂÍÓôôx]͈Ì͈ÊÍiµÕ>ÞÍ>««ˆV>LiÍ̜Í-+Í-iẶiĂÍÓôôn°

Hardware Resource Bottlenecks

/Þ«ˆV>Þ]Í-+Í-iẶiĂÍ`>Ì>L>Êi performance is affected by stress on the following hardware resources:

Í UÍ ˆÊŽÍĨ"

Í UÍ *ÜViÊʜĂ

Í UÍ iÌܜÎ

-ÌĂiÊÊÍLiޜ˜`Í̅iÍV>«>VˆÌÞ͜vÍ>ͅ>Ă`Ü>ĂiÍĂiʜÕĂViÍvœĂ“ÊÍ>ÍLœÌ̏i˜iVŽ°Í/œÍ>``ĂiÊÊÍ̅iÍ

œÛiĂ>Í«iĂvœĂ“>˜Vi͜vÍ>ÍÊÞÊÌi“]ÍޜÕ͘ii`Í̜͈`i˜ÌˆvÞÍ̅iÊiÍLœÌ̏i˜iVŽÊ]ÍLiV>ÕÊiÍ̅iÞÍvœĂ“Í̅iÍ limit on overall system performance

Identifying Bottlenecks

/…iĂi͈ÊÍÕÊÕ>ÞÍ>ÍĂi>̈œ˜Ê…ˆ«ÍLiÌÜii˜ÍĂiʜÕĂViÍLœÌ̏i˜iVŽÊ°ÍœĂÍiÝ>“«i]Í>Í«ĂœViÊʜĂÍ LœÌ̏i˜iVŽÍ“>ÞÍLiÍ>ÍÊޓ«Ìœ“ÍœvÍiÝViÊʈÛiÍ«>}ˆ˜}Í­“i“œĂÞÍLœÌ̏i˜iVŽ®ÍœĂÍ>ÍʏœÜÍ`ˆÊŽÍ­`ˆÊŽÍ LœÌ̏i˜iVŽ®°ÍvÍ>ÍÊÞÊÌi“͈Ê͏œÜ͜˜Í“i“œĂÞ]ÍV>Õʈ˜}ÍiÝViÊʈÛiÍ«>}ˆ˜}]Í>˜`ͅ>ÊÍ>ÍʏœÜÍ`ˆÊŽ]Í̅i˜Í

œ˜i͜vÍ̅iÍi˜`ÍĂiÊՏÌÊÍ܈ÍLiÍ>Í«ĂœViÊʜĂÍ܈̅ͅˆ}…ÍṎˆđ>̈œ˜Íʈ˜ViÍ̅iÍ«ĂœViÊʜĂͅ>ÊÍ̜ÍÊ«i˜`Í

̅iÍĂiÊՏÌ>˜Ìͅˆ}…͘ՓLiĂ͜vÍĨ"ÍĂiµÕiÊÌÊ°Í,i«>Vˆ˜}Í̅iÍ«ĂœViÊʜĂÍ܈̅Í>Ív>ÊÌiĂ͜˜i͓>Þͅi«Í

>͏ˆÌ̏i]ÍLÕÌ͈ÌÍܜՏ`͘œÌÍLiÍ̅iÍLiÊÌ͜ÛiĂ>ÍʜṎœ˜°Í˜Í>ÍV>Êi͏ˆŽiÍ̅ˆÊ]͈˜VĂi>ʈ˜}͓i“œĂÞ͈ÊÍ>Í

“œĂiÍ>««Ăœ«Ăˆ>ÌiÍʜṎœ˜]ÍLiV>ÕÊi͈ÌÍ܈Í`iVĂi>ÊiÍ«ĂiÊÊÕĂi͜˜Í̅iÍ`ˆÊŽÍ>˜`Í«ĂœViÊʜĂÍ>ÊÍÜi°Í

N Note The most common performance problem is usually I/O, either from memory or from the disk

"˜i͜vÍ̅iÍLiÊÌÍÜ>ÞÊ͜v͏œV>̈˜}Í>ÍLœÌ̏i˜iVŽÍˆÊÍ̜͈`i˜ÌˆvÞÍĂiʜÕĂViÊÍ̅>ÌÍ>ĂiÍÜ>ˆÌˆ˜}ÍvœĂÍ some other resource to complete its operation You can use Performance Monitor counters

œĂÍ6ÊÍÊÕV…Í>ÊÍouo*`i[ko[s]ep[op]poÍ̜Í}>̅iĂÍ̅>Ì͈˜vœĂ“>̈œ˜°Í/…iÍĂiÊ«œ˜ÊiÍ̈“i͜vÍ>Í request served by a resource includes the time the request had to wait in the resource queue,

>ÊÍÜiÍ>ÊÍ̅iÍ̈“iÍÌ>Ži˜Í̜ÍiÝiVÕÌiÍ̅iÍĂiµÕiÊÌ]ÍʜÍi˜`ÍÕÊiĂÍĂiÊ«œ˜ÊiÍ̈“i͈ÊÍ`ˆĂiV̏ÞÍ«Ăœ«œĂ-tional to the amount of queuing in a system

`ˆÊŽÍÊÕLÊÞÊÌi“Í>Ăi>`Þͅ>ÊÍ«i˜`ˆ˜}Í`ˆÊŽÍĂiµÕiÊÌÊ͜˜ÍˆÌ]Í>͘iÜÍ`ˆÊŽÍĂiµÕiÊÌͅ>ÊÍ̜ÍÜ>ˆÌÍ՘̈Í̅iÍ

«ĂiۈœÕÊÍ`ˆÊŽÍĂiµÕiÊÌÊÍVœ“«iÌi°ÍvÍ̅iÍ̈“iÍÌ>Ži˜ÍLÞÍ>˜Í>ÛiĂ>}iÍ`ˆÊŽÍÌĂ>˜ÊviĂ͈Ê͜˜iÍÊiVœ˜`]Í

̅i˜Í̅i͘iÜÍ`ˆÊŽÍĂiµÕiÊÌͅ>ÊÍ̜ÍÜ>ˆÌÍvœĂÍ>LœÕÌÍÌi˜ÍÊiVœ˜`ÊÍLivœĂiÍ}iÌ̈˜}Í̅iÍ>ÌÌi˜Ìˆœ˜ÍœvÍ̅iÍ

`ˆÊŽÍÊÕLÊÞÊÌi“°Í/…iĂivœĂi]Í̅iÍ̜Ì>ÍĂiÊ«œ˜ÊiÍ̈“i͜vÍ̅iÍ`ˆÊŽÍĂiµÕiÊÌÍ܈ÍLiÍÌi˜ÍÊiVœ˜`ÊÍÜ>ˆÌÍ

̈“i]Í«ÕÊ͜˜iÍÊiVœ˜`Í`ˆÊŽÍÌĂ>˜ÊviĂÍ̈“i°

iÍ>Ü>ĂiÍ̅>ÌÍ̅iÍ>LÊi˜Vi͜vÍ>͵ÕiÕiÍ`œiÊ͘œÌ͓i>˜Í̅>ÌÍ̅iĂi͈Ê͘œÍLœÌ̏i˜iVŽ°Í7…i˜Í

µÕiÕi͏i˜}̅ÊÍÊÌ>ĂÌÍ}Ü܈˜}]ͅœÜiÛiĂ]͈Ì͈ÊÍ>ÍÊÕĂiÍʈ}˜Í̅>ÌÍ̅iÍÊÞÊÌi“͈Ê͘œÌÍ>LiÍ̜͎ii«ÍÕ«Í

with the demand

œÌÍ>ÍĂiʜÕĂViÊͅ>ÛiÍÊ«iVˆvˆVÍVœÕ˜ÌiĂÊÍ̅>ÌÍʅœÜ͵ÕiՈ˜}͏iÛiÊ]ÍLÕÌ͓œÊÌÍĂiʜÕĂViÊÍ

…>ÛiÍʜ“iÍVœÕ˜ÌiĂÊÍ̅>ÌÍĂi«ĂiÊi˜ÌÍ>˜ÍœÛiĂVœ““ˆÌÌ>ÍœvÍ̅>ÌÍĂiʜÕĂVi°ÍœĂÍiÝ>“«i]͓i“œĂÞÍ

has no such counter, but a large number of hard page faults represents the overcommittal of

«…ÞʈV>Í“i“œĂÞÍ­…>Ă`Í«>}iÍv>ՏÌÊÍ>ĂiÍiÝ«>ˆ˜i`͏>ÌiĂ͈˜Í̅iÍV…>«ÌiĂ͈˜Í̅iÍÊiV̈œ˜Íº*>}iÊĨÊiVÍ VœÕ˜ÌiĂÊÍ̜͈˜`ˆV>ÌiÍ̅i͏iÛiÍœv͵ÕiՈ˜}°ÍœĂÍiÝ>“«i]Í̅i counter L]caHebaAtla_p]j_u

ˆ˜`ˆV>ÌiÊͅœÜ͏œ˜}Í>Í«>}iÍ܈ÍÊÌ>Þ͈˜Í̅iÍLÕvviĂÍ«œœÍ܈̅œÕÌÍLiˆ˜}ÍĂiviĂi˜Vi`°Í/…ˆÊ͈ÊÍ>˜Í

ˆ˜`ˆV>̜Ă͜vͅœÜÍÜiÍ-+Í-iẶiĂ͈ÊÍ>LiÍ̜͓>˜>}i͈ÌÊ͓i“œĂÞ]Íʈ˜ViÍ>͏œ˜}iĂ͏ˆvi͓i>˜ÊÍ̅>ÌÍ

>Í«ˆiVi͜vÍ`>Ì>͈˜Í̅iÍLÕvviĂÍ܈ÍLiÍ̅iĂi]Í>Û>ˆ>Li]ÍÜ>ˆÌˆ˜}ÍvœĂÍ̅i͘iÝÌÍĂiviĂi˜Vi°ÍœÜiÛiĂ]Í

>ÍʅœĂÌiĂ͏ˆvi͓i>˜ÊÍ̅>ÌÍ-+Í-iẶiĂ͈Ê͓œÛˆ˜}Í«>}iÊ͈˜Í>˜`͜ÕÌ͜vÍ̅iÍLÕvviĂ͵ՈVŽÞ]Í«œÊʈLÞÍ

ÊÕ}}iÊ̈˜}Í>͓i“œĂÞÍLœÌ̏i˜iVŽ°

You will see which counters to use inÍ>˜>Þđˆ˜}Íi>V…ÍÌÞ«i͜vÍLœÌ̏i˜iVŽÍʅœĂ̏ް

Bottleneck Resolution

"˜ViÍޜÕͅ>Ûi͈`i˜Ìˆvˆi`ÍLœÌ̏i˜iVŽÊ]ÍޜÕÍV>˜ÍĂiʜÛiÍ̅i“͈˜ÍÌܜÍÜ>ÞÊ\

Í UÍ 9œÕÍV>˜Íˆ˜VĂi>ÊiÍĂiʜÕĂViÍÌ…ĂœÕ}…«ÕÌ°

Í UÍ 9œÕÍV>˜Í`iVĂi>ÊiÍ̅iÍ>ĂÈÛ>ÍĂ>Ìi͜vÍĂiµÕiÊÌÊÍ̜Í̅iÍĂiʜÕĂVi°

˜VĂi>ʈ˜}Í̅iÍÌ…ĂœÕ}…«ÕÌÍÕÊÕ>ÞÍĂiµÕˆĂiÊÍiÝÌĂ>ÍĂiʜÕĂViÊÍÊÕV…Í>Ê͓i“œĂÞ]Í`ˆÊŽÊ]Í«Ăœ-ViÊʜĂÊ]͜Ă͘iÌܜÎÍ>`>«ÌiĂÊ°Í9œÕÍV>˜Í`iVĂi>ÊiÍ̅iÍ>ĂÈÛ>ÍĂ>ÌiÍLÞÍLiˆ˜}͓œĂiÍÊiiV̈ÛiÍ>LœÕÌÍ

̅iÍĂiµÕiÊÌÊÍ̜Í>ÍĂiʜÕĂVi°ÍœĂÍiÝ>“«i]Í܅i˜ÍޜÕͅ>ÛiÍ>Í`ˆÊŽÍÊÕLÊÞÊÌi“ÍLœÌ̏i˜iVŽ]ÍޜÕÍV>˜Í

iˆÌ…iĂ͈˜VĂi>ÊiÍ̅iÍÌ…ĂœÕ}…«ÕÌ͜vÍ̅iÍ`ˆÊŽÍÊÕLÊÞÊÌi“ÍœĂÍ`iVĂi>ÊiÍ̅iÍ>“œÕ˜Ì͜vÍĨ"ÍĂiµÕiÊÌÊ°

˜VĂi>ʈ˜}Í̅iÍÌ…ĂœÕ}…«ÕÌ͓i>˜ÊÍ>``ˆ˜}͓œĂiÍ`ˆÊŽÊ͜ĂÍÕ«}Ă>`ˆ˜}Í̜Ív>ÊÌiĂÍ`ˆÊŽÊ°ÍiVĂi>Ê-ˆ˜}Í̅iÍ>ĂÈÛ>ÍĂ>Ìi͓i>˜Ê͈`i˜Ìˆvވ˜}Í̅iÍV>ÕÊi͜vͅˆ}…ÍĨ"ÍĂiµÕiÊÌÊÍ̜Í̅iÍ`ˆÊŽÍÊÕLÊÞÊÌi“Í>˜`Í

>««Þˆ˜}ÍĂiʜṎœ˜ÊÍ̜Í`iVĂi>ÊiÍ̅iˆĂ͘ՓLiĂ°Í9œÕ͓>ÞÍLiÍ>LiÍ̜Í`iVĂi>ÊiÍ̅iÍĨ"ÍĂiµÕiÊÌÊ]Í

vœĂÍiÝ>“«i]ÍLÞÍ>``ˆ˜}Í>««Ăœ«Ăˆ>Ìi͈˜`iÝiÊ͜˜Í>ÍÌ>LiÍ̜͏ˆ“ˆÌÍ̅iÍ>“œÕ˜Ì͜vÍ`>Ì>Í>VViÊÊi`͜ĂÍ

by partitioning a tableÍLiÌÜii˜Í“Տ̈«iÍ`ˆÊŽÊ°

Memory Bottleneck Analysis

Memory can be aÍ«ĂœLi“>̈VÍLœÌ̏i˜iVŽÍLiV>ÕÊiÍ>ÍLœÌ̏i˜iVŽÍˆ˜Í“i“œĂÞÍ܈Í“>˜ˆviÊÌ͜˜Í

œÌ…iĂÍĂiʜÕĂViÊ]Í̜œ°Í/…ˆÊ͈ÊÍ«>Ă̈VՏ>ÏÞÍÌẮiÍvœĂÍ>ÍÊÞÊÌi“ÍẮ˜˜ˆ˜}Í-+Í-iẶiĂ°Í7…i˜Í-+Í

-iẶiĂÍẮ˜Ê͜ÕÌ͜vÍV>V…iÍ­œĂ͓i“œĂÞ®]Í>Í«ĂœViÊÊÍ܈̅ˆ˜Í-+Í-iẶiĂÍ­V>i`Ílazy writer®Í…>Ê

̜ÍܜÎÍiÝÌi˜ÊˆÛiÞÍ̜͓>ˆ˜Ì>ˆ˜Íi˜œÕ}…ÍvĂii͈˜ÌiØ>Í“i“œĂÞÍ«>}iÊÍ܈̅ˆ˜Í-+Í-iẶiĂ°Í/…ˆÊÍ

L>VŽÍ̜Í`ˆÊŽ°

SQL Server Memory Management

-+Í-iẶiĂ͓>˜>}iÊ memory for databases, including memory requirements for data and

µÕiĂÞÍiÝiVṎœ˜Í«>˜Ê]͈˜Í>͏>Ă}iÍ«œœÍœv͓i“œĂÞÍV>i`Í̅iÍmemory pool°Í/…i͓i“œĂÞÍ«œœÍ

consists of a collection of 8KB buffers to manage data pages and plan cache pages, free pages,

>˜`ÍʜÍvœĂ̅°Í/…i͓i“œĂÞÍ«œœÍˆÊÍÕÊÕ>ÞÍ̅i͏>Ă}iÊÌÍ«œĂ̈œ˜ÍœvÍ-+Í-iẶiĂ͓i“œĂÞ°Í-+Í -iẶiĂ͓>˜>}iÊ͓i“œĂÞÍLÞÍ}Ü܈˜}͜ĂÍÊ…Ăˆ˜Žˆ˜}͈ÌÊ͓i“œĂÞÍ«œœÍʈđiÍ`ޘ>“ˆV>Þ°

You

canÍVœ˜vˆ}ÕĂiÍ-+Í-iẶiĂÍvœĂÍ`ޘ>“ˆV͓i“œĂÞ͓>˜>}i“i˜Ì͈˜Í-+Í-iẶiĂÍ>˜>}i-“i˜ÌÍ-ÌÕ`ˆœÍ­ -®°ÍœÍ̜Í̅iÍi“œĂÞÍvœ`iĂ͜vÍ̅iÍ-iẶiĂÍ*Ü«iĂ̈iÊÍ`ˆ>œ}ÍLœÝ]Í>ÊÍʅœÜ˜Íˆ˜Í

ˆ}ÕĂiÍӇΰ

Figure 2-3 SQL Server memory configuration

/…iÍ`ޘ>“ˆV͓i“œĂÞÍĂ>˜}i͈ÊÍVœ˜Ì܏i`ÍÌ…ĂœÕ}…ÍÌܜÍVœ˜vˆ}ÕĂ>̈œ˜Í«Ăœ«iĂ̈iÊ\Í

Iejeiqi$I>% and I]teiqi$I>%

Í UÍ Iejeiqi$I>%]Í>ÊœÍŽ˜œÜ˜Í>ÊÍiejoanraniaiknu]ÍܜÎÊÍ>ÊÍ>ÍvœœĂÍÛ>ÕiÍvœĂÍ̅i͓i“œĂÞÍ

«œœ°Í"˜ViÍ̅i͓i“œĂÞÍ«œœÍĂi>V…iÊÍ̅iÍÊ>“iÍʈđiÍ>ÊÍ̅iÍvœœĂÍÛ>Õi]Í-+Í-iẶiĂÍV>˜Í

Vœ˜Ìˆ˜ÕiÍVœ““ˆÌ̈˜}Í«>}iÊ͈˜Í̅i͓i“œĂÞÍ«œœ]ÍLÕÌ͈ÌÍV>˜˜œÌÍLiÍʅẮ˜ŽÍ̜͏iÊÊÍ̅>˜Í

̅iÍvœœĂÍÛ>Õi°Í œÌiÍ̅>ÌÍ-+Í-iẶiĂÍ`œiÊ͘œÌÍÊÌ>ĂÌÍ܈̅Í̅iÍiejoanraniaiknu

con-figuration value but commits memory dynamically, as needed

Í UÍ I]teiqi$I>%]Í>ÊœÍŽ˜œÜ˜Í>Ê i]toanraniaiknu, serves as a ceiling value to limit the

“>݈“Õ“Í}ÜÜ̅͜vÍ̅i͓i“œĂÞÍ«œœ°Í/…iÊiÍVœ˜vˆ}ÕĂ>̈œ˜ÍÊiÌ̈˜}ÊÍÌ>ŽiÍivviVÌ͈““i-diately and do not require a restart

ˆVĂœÊœvÌÍĂiVœ““i˜`ÊÍ̅>ÌÍޜÕÍÕÊiÍ`ޘ>“ˆV͓i“œĂÞÍVœ˜vˆ}ÕĂ>̈œ˜ÍvœĂÍ-+Í-iẶiĂ]Í

where iejoanraniaiknu will be 0 and i]toanraniaiknuÍ܈ÍLiÍ̅i͓>݈“Õ“Í«…ÞʈV>Í

memory of the system, assuming a single instance on the machine You should not run other

“i“œĂއˆ˜Ìi˜ÊˆÛiÍ>««ˆV>̈œ˜Ê͜˜Í̅iÍÊ>“iÍÊiẶiĂÍ>ÊÍ-+Í-iẶiĂ]ÍLÕÌ͈vÍޜÕ͓ÕÊÌ]ÍÍĂiVœ“-mend you first get estimates on how much memory is needed by other applications and then

Vœ˜vˆ}ÕĂiÍ-+Í-iẶiĂÍ܈̅Í>Íi]toanraniaiknu value set to prevent the other applications

vܓÍÊÌ>Ặˆ˜}Í-+Í-iẶiĂ͜v͓i“œĂÞ°Í"˜Í>ÍÊÞÊÌi“Í܅iĂiÍ-+Í-iẶiĂ͈ÊÍẮ˜˜ˆ˜}͜˜ÍˆÌÊ͜ܘ]ÍÍ

«ĂiviĂÍ̜ÍÊiÌÍ̅i͓ˆ˜ˆ“Õ“ÍÊiẶiĂ͓i“œĂÞÍiµÕ>Í̜Í̅i͓>ÝÍÛ>ÕiÍ>˜`Íʈ“«ÞÍ`ˆÊ«>ÌV…Í܈̅Í

`ޘ>“ˆV͓>˜>}i“i˜Ì°Í"˜Í>ÍÊiẶiĂÍ܈͓̅Տ̈«iÍ-+Í-iẶiĂ͈˜ÊÌ>˜ViÊ]ÍޜսÍ˜ii`Í̜Í>`ÕÊÌÍ

̅iÊi͓i“œĂÞÍÊiÌ̈˜}ÊÍ̜Íi˜ÊÕĂiÍi>V…͈˜ÊÌ>˜Viͅ>ÊÍ>˜Í>`iµÕ>ÌiÍÛ>Õi°ÍÕÊÌ͓>ŽiÍÊÕĂiÍޜսÛiÍ

ivÌÍi˜œÕ}…Í“i“œĂÞÍvœĂÍ̅i͜«iĂ>̈˜}ÍÊÞÊÌi“Í>˜`ÍiÝÌiØ>Í«ĂœViÊÊiÊ]Í>ÊÍÜiÍ>Ê͘œ˜‡LÕvviÇ

pool memory (which used to be called IaiPkHa]ra®°

i“œĂÞÍ܈̅ˆ˜Í-+Í-iẶiĂÍV>˜ÍLiÍÜÕ}…ÞÍ`ˆÛˆ`i`͈˜ÌœÍLÕvviĂÍ«œœÍ“i“œĂÞ]Í܅ˆV…ÍĂi«-ĂiÊi˜ÌÊÍ`>Ì>Í«>}iÊÍ>˜`ÍvĂiiÍ«>}iÊ]Í>˜`͘œ˜LÕvviĂ͓i“œĂÞ]Í܅ˆV…ÍVœ˜ÊˆÊÌÊ͜vÍ̅Ăi>`Ê]ÍÊ]Í

ˆ˜Ži`ÍÊiẶiĂÊ]Í>˜`̅͜iĂÊ°ÍœÊÌ͜vÍ̅i͓i“œĂÞÍÕÊi`ÍLÞÍ-+Í-iẶiĂÍ}œiÊ͈˜ÌœÍ̅iÍLÕvviĂÍ«œœ°

N Note SQL Server does consume more memory than simply that specified by the i]t[oanran[iaiknu

setting

You can also manage the configuration values for iejoanraniaiknu and i]toanran

iaiknu by using the ol[_kjbecqnaÍÊÞÊÌi“ÍÊ̜Ăi`Í«ĂœVi`ÕĂi°Í/œÍÊiiÍ̅iÍVœ˜vˆ}ÕĂ>̈œ˜ÍÛ>ÕiÊÍvœĂÍ

̅iÊiÍ«>Ă>“iÌiĂÊ]ÍiÝiVÕÌiÍ̅iÍol[_kjbecqna stored procedure as follows:

ata_ol[_kjbecqna#iejoanraniaiknu$I>%#

ata_ol[_kjbecqna#i]toanraniaiknu$I>%#

ˆ}ÕĂiÍӇ{ÍʅœÜÊÍ̅iÍĂiÊՏÌ͜vÍẮ˜˜ˆ˜}Í̅iÊiÍVœ““>˜`Ê°

Figure 2-4 SQL Server memory configuration properties

œÌiÍ̅>ÌÍ̅iÍ`iv>ՏÌÍÛ>ÕiÍvœĂÍ̅iÍiejoanraniaiknu setting is 0MB, and for the i]t oanraniaiknuÍÊiÌ̈˜}͈Ì͈ÊÍÓ£{Ì{nÎỈ{Ì °Í
Êœ]Íi]toanraniaiknu cannot be set to less

̅>˜Í{ °

You can also modify these configuration values using the ol[_kjbecqna stored procedure

œĂÍiÝ>“«i]Í̜ÍÊiÌÍi]toanraniaiknu to 200MB and iejoanraniaiknuÍ̜ͣôô ]ÍiÝiVÕÌiÍ the following set of statements (oap[iaiknu*omh͈˜Í̅iÍ`œÜ˜œ>`®\

QOAi]opan

ATA?ol[_kjbecqna#odks]`r]j_a`klpekj#(#-#

NA?KJBECQNA

ata_ol[_kjbecqna#iejoanraniaiknu$I>%#(-,,

ata_ol[_kjbecqna#i]toanraniaiknu$I>%#(.,,

NA?KJBECQNASEPDKRANNE@A

/…iÍiejoanraniaiknu and i]toanraniaiknu configurations are classified as advanced options By default, the ol[_kjbecqna stored procedure does not affect/display the advanced options Setting odks]`r]j_a`klpekj to - as shown previously enables the ol[_kjbecqna stored procedure to affect/display the advanced options

/…iÍNA?KJBECQNA statement updates the memory configuration values set by ol[_kjbecqna Since ad hoc updates to the system catalog containing the memory configuration values are not recommended, the KRANNE@A flag is used with the NA?KJBECQNA statement to force the memory configuration If you do the memory configuration through Management Studio,

>˜>}i“i˜ÌÍ-ÌÕ`ˆœÍ>Õ̜“>̈V>ÞÍiÝiVÕÌiÊÍ̅iÍNA?KJBECQNASEPDKRANNE@A statement after the configuration setting

˜Íʜ“iÍĂ>ĂiÍVˆĂVՓÊÌ>˜ViÊ]ÍޜÕ͓>Þ͘ii`Í̜Í>œÜÍvœĂÍ-+Í-iẶiĂÍʅ>Ș}Í>ÍÊÞÊÌi“½ÊÍ

œ˜ÍˆÌ°Í œÌ…ÍÊiẶiĂÊÍ>Ăiͅi>ÛÞÍÕÊiĂÊ͜v͓i“œĂÞÍ>˜`Í̅ÕÊ͎ii«Í«Õʅˆ˜}Íi>V…̅͜iĂÍvœĂ͓i“œĂÞ°Í

“i“œĂÞ͓>˜>}i“i˜Ì͜ÛiÅi>`ÍLÞÍVœ˜vˆ}ÕȘ}Í-+Í-iẶiĂÍvœĂÍ>ÍvˆÝi`͓i“œĂÞÍʈđi°ÍœÜiÛiĂ]Í

«i>Êi͎ii«Íˆ˜Í“ˆ˜`Í̅>ÌÍʈ˜ViÍ-+Í-iẶiĂ͈ÊÍ>˜ÍiÝÌĂi“iÞÍĂiʜÕĂVi‡ˆ˜Ìi˜ÊˆÛiÍ«ĂœViÊÊ]͈Ì͈ÊÍ …ˆ}…ÞÍĂiVœ““i˜`i`Í̅>ÌÍޜÕͅ>ÛiÍ>Í`i`ˆV>Ìi`Í-+Í-iẶiĂÍ«Ăœ`ÕV̈œ˜Í“>V…ˆ˜i°

œÜÍ̅>ÌÍޜÕÍ՘`iĂÊÌ>˜`Í-+Í-iẶiĂ͓i“œĂÞ͓>˜>}i“i˜Ì]͏i̽ÊÍVœ˜Êˆ`iĂÍ̅iÍperfor-“>˜ViÍVœÕ˜ÌiĂÊÍޜÕÍV>˜ÍÕÊiÍ̜Í>˜>ÞđiÍÊÌĂiÊÊ͜˜Í“i“œĂÞ]Í>ÊÍʅœÜ˜Íˆ˜Í/>LiÍӇ£°

Table 2-1 Performance Monitor Counters to Analyze Memory Pressure

Object(Instance[,InstanceN]) Counter Description Values

dependent L]cao+oa_ Rate of hard page faults Average

Û>Õi͐Íxô L]caB]qhpo+oa_ Rate of total page faults

its baseline value for trend analysis L]caoEjlqp+oa_ Rate of input page faults

L]caoKqplqp+oa_ Rate of output page faults

Object(Instance[,InstanceN]) Counter Description Values

OMHOanran6>qbbanI]j]can >qbban_]_da

depn]pek

Percentage of requests served out of buffer cache

Average value q 90%

L]caHeba

Atla_p]j_u

/ˆ“i page spends in buffer

Average Û>ÕìÍÎôô

?da_glkejp

L]cao+oa_

PagesÍÜÈÌÌi˜Í̜Í`ˆÊŽÍLÞÍ

V…iVŽ«œˆ˜Ì

Average Û>Õi͐ÍÎô H]vusnepao+oa_ ˆĂÌÞ aged pages flushed

from buffer

Average value < 20 OMHOanran6IaiknuI]j]can IaiknuCn]jpo

Laj`ejc

ՓLiĂ of processes waiting for memory grant

Average value = 0 P]ncapOanran

Iaiknu$G>%

>݈“Õ“ physical

“i“œĂÞÍ-+Í-iẶiĂÍV>˜Í Vœ˜ÊՓi͜˜Í̅iÍLœÝ

of physical memory Pkp]hOanran

Iaiknu$G>%

Physical memory cur-Ăi˜ÌÞÍ>Êʈ}˜i`Í̜Í-+Í

Iaiknu$G>%

that this process has allocated that cannot

be shared with other processes

½Í˜œÜÍÜ>ŽÍޜÕÍÌ…ĂœÕ}…Í̅iÊiÍVœÕ˜ÌiĂÊÍ̜Í}iÌÍ>ÍLiÌÌiĂ͈`i> of what you can use them for

Available Bytes

/…iÍ=r]eh]^ha>upaoÍVœÕ˜ÌiĂÍĂi«ĂiÊi˜ÌÊÍvĂiiÍ«…ÞʈV>Í“i“œĂÞ͈˜Í̅iÍÊÞÊÌi“°ÍœĂÍ}œœ`Í«iĂ-vœĂ“>˜Vi]Í̅ˆÊÍVœÕ˜ÌiĂÍÛ>ÕiÍʅœÕ`͘œÌÍLiÍ̜œÍœÜ°ÍvÍ-+Í-iẶiĂ͈ÊÍVœ˜vˆ}ÕĂi`ÍvœĂÍ`ޘ>“ˆVÍ

memory usage, then this value will be controlled by calls to a Windows API that determines

-+Í-iẶiĂ͓i“œĂÞ͘œÌÍV…>˜}ˆ˜}͈˜`ˆV>ÌiÊÍ̅>ÌÍ̅iÍÊiẶiĂ͈ÊÍ՘`iĂÍÊiÛiĂi͓i“œĂÞÍÊÌĂiÊÊ°

Pages/sec and Page Faults/sec Counters

/œÍ՘`iĂÊÌ>˜`Í̅i͈“«œĂÌ>˜Vi͜vÍ̅iÍL]cao+oa_ and L]caB]qhpo+oa_ counters, you first need

to learn about page faults A page fault occurs when a process requires code or data that is not

in its working setÍ­ˆÌÊÍÊ«>Vi͈˜Í«…ÞʈV>Í“i“œĂÞ®°ÍÌ͓>Þ͏i>`Í̜Í>ÍʜvÌÍ«>}iÍv>ՏÌ͜ĂÍ>ͅ>Ă`Í«>}iÍ

fault If the faulted page is found elsewhere in physical memory, then it is called a soft page

fault A hard page fault͜VVÕĂÊÍ܅i˜Í>Í«ĂœViÊÊÍĂiµÕˆĂiÊÍVœ`i͜ĂÍ`>Ì>Í̅>Ì͈Ê͘œÌ͈˜ÍˆÌÊÍܜΈ˜}Í

ÊiÌ͜ĂÍiÊi܅iĂi͈˜Í«…ÞʈV>Í“i“œĂÞÍ>˜`͓ÕÊÌÍLiÍĂiÌÈiÛi`ÍvܓÍ`ˆÊŽ°

/…iÍÊ«ii`͜vÍ>Í`ˆÊŽÍ>VViÊÊ͈Ê͈˜Í̅i͜Ă`iĂ͜v͓ˆˆÊiVœ˜`Ê]Í܅iĂi>ÊÍ>͓i“œĂÞÍ>VViÊÊ͈Ê͈˜Í

̅i͜Ă`iĂ͜v͘>˜œÊiVœ˜`Ê°Í/…ˆÊͅÕ}iÍ`ˆvviĂi˜Vi͈˜Í̅iÍÊ«ii`ÍLiÌÜii˜Í>Í`ˆÊŽÍ>VViÊÊÍ>˜`Í>͓i“-œĂÞÍ>VViÊÊ͓>ŽiÊÍ̅iÍivviVÌ͜vͅ>Ă`Í«>}iÍv>ՏÌÊÍʈ}˜ˆvˆV>˜ÌÍVœ“«>Ăi`Í̜Í̅>Ì͜vÍʜvÌÍ«>}iÍv>ՏÌÊ°

/…iÍL]cao+oa_ÍVœÕ˜ÌiĂÍĂi«ĂiÊi˜ÌÊÍ̅i͘ՓLiĂ͜vÍ«>}iÊÍĂi>`Ívܓ͜ĂÍÜÈÌÌi˜Í̜Í`ˆÊŽÍ«iĂÍ

ÊiVœ˜`Í̜ÍĂiʜÛiͅ>Ă`Í«>}iÍv>ՏÌÊ°Í/…iÍL]caB]qhpo+oa_ performance counter indicates the

total page faults per second—soft page faults plus hard page faults—handled by the system

>Ă`Í«>}iÍv>ՏÌÊ]͈˜`ˆV>Ìi`ÍLÞÍL]cao+oa_, should not be consistently high If this counter

ˆÊÍVœ˜ÊˆÊÌi˜ÌÞÍÛiĂÞͅˆ}…]Í̅i˜Í-+Í-iẶiĂ͈ÊÍ«ĂœL>LÞÍÊÌ>Ặˆ˜}̅͜iĂÍ>««ˆV>̈œ˜Ê°Í/…iĂiÍ>Ăi͘œÍ hard and fast numbers for what indicates a problem, because these numbers will vary widely LiÌÜii˜ÍÊÞÊÌi“ÊÍL>Êi`͜˜Í̅iÍ>“œÕ˜ÌÍ>˜`ÍÌÞ«i͜v͓i“œĂÞÍ>ÊÍÜiÍ>ÊÍ̅iÍÊ«ii`͜vÍ`ˆÊŽÍ>VViÊÊÍ

on the system

If the L]cao+oa_ counter isÍÛiĂÞͅˆ}…]Í̅i˜ÍޜÕÍV>˜ÍLĂi>ŽÍˆÌÍի͈˜ÌœÍL]caoEjlqp+oa_ and L]caoKqplqp+oa_:

Í UÍ L]caoEjlqp+oa_: An application will wait only on an input page, not on an output page

Í UÍ L]caoKqplqp+oa_: Page output will stress the system, but an application usually does not see this stress Pages output are usually represented by the application’s dirty

«>}iÊÍ̅>Ì͘ii`Í̜ÍLiÍL>VŽi`͜ÕÌÍ̜Í̅iÍ`ˆÊŽ°ÍL]caoKqplqp+oa_ is an issue only when

`ˆÊŽÍœ>`ÍLiVœ“iÍ>˜ÍˆÊÊÕi°

Êœ]ÍV…iVŽÍLnk_aoo6L]caB]qhpo+oa_Í̜Ívˆ˜`͜ÕÌÍ܅ˆV…Í«ĂœViÊÊ͈ÊÍV>Õʈ˜}ÍiÝViÊʈÛiÍ paging in case of high L]cao+oa_°Í/…iÍLnk_aoo object is the system component that provides performance data for the processes running on the system, which are individually represented

by their corresponding instance name

œĂÍiÝ>“«i]Í̅iÍ-+Í-iẶiĂÍ«ĂœViÊÊ͈ÊÍĂi«ĂiÊi˜Ìi`ÍLÞÍ̅iÍomhoanrn instance of the Lnk_aoo͜LiV̰͈}…͘ՓLiĂÊÍvœĂÍ̅ˆÊÍVœÕ˜ÌiĂÍÕÊÕ>ÞÍ`œÍ˜œÌ͓i>˜Í“ÕV…Í՘iÊÊÍL]cao+oa_ is high L]caB]qhpo+oa_ can range all over the spectrum with normal application behavior, with Û>ÕiÊÍvܓÍôÍ̜ͣ]ôôôÍ«iĂÍÊiVœ˜`ÍLiˆ˜}Í>VVi«Ì>Li°Í/…ˆÊÍi˜ÌˆĂiÍ`>Ì>ÍÊiÌ͓i>˜ÊÍ>ÍL>Êiˆ˜i͈ÊÍ iÊÊi˜Ìˆ>Í̜Í`iÌiӈ˜iÍ̅iÍiÝ«iVÌi`͘œĂ“>ÍLi…>ۈœĂ°

Buffer Cache Hit Ratio

/…iÍbuffer cache is the pool of buffer pages into which data pages are read, and it is often the

Lˆ}}iÊÌÍ«>ĂÌ͜vÍ̅iÍ-+Í-iẶiĂ͓i“œĂÞÍ«œœ°Í/…ˆÊÍVœÕ˜ÌiĂÍÛ>ÕiÍʅœÕ`ÍLiÍ>Êͅˆ}…Í>ÊÍ«œÊʈLi]Í iÊ«iVˆ>ÞÍvœĂÍ"/*ÍÊÞÊÌi“ÊÍ̅>ÌÍʅœÕ`ͅ>ÛiÍv>ˆĂÞÍĂi}ˆ“i˜Ìi`Í`>Ì>Í>VViÊÊ]Í՘ˆŽiÍ>ÍÜ>Ăi-…œÕÊi͜ĂÍĂi«œĂ̈˜}ÍÊÞÊÌi“°ÍÌ͈ÊÍiÝÌĂi“iÞÍVœ““œ˜Í̜Ívˆ˜`Í̅ˆÊÍVœÕ˜ÌiĂÍÛ>ÕiÍ>Ê͙™Í«iĂVi˜ÌÍ

or more for most production servers A low >qbban_]_dadepn]pek value indicates that few requests could be served out of the buffer cache, with the rest of the requests being served vܓÍ`ˆÊŽ°

7…i˜Í̅ˆÊͅ>««i˜Ê]ÍiˆÌ…iĂÍ-+Í-iẶiĂ͈ÊÍÊ̈ÍÜ>ӈ˜}Íի͜ĂÍ̅i͓i“œĂÞÍĂiµÕˆĂi“i˜Ì͜vÍ

̅iÍLÕvviĂÍV>V…i͈Ê͓œĂiÍ̅>˜Í̅i͓>݈“Õ“Í“i“œĂÞÍ>Û>ˆ>LiÍvœĂ͈ÌÊÍ}ÜÜ̅°ÍvÍ̅ˆÊ͈ÊÍVœ˜ÊˆÊ-tently low, you should consider getting more memory for the system

Page Life Expectancy

L]caHebaAtla_p]j_u indicates how long a page will stay in the buffer pool without being ref-iĂi˜Vi`°Íi˜iĂ>Þ]Í>͏œÜ͘ՓLiĂÍvœĂÍ̅ˆÊÍVœÕ˜ÌiĂ͓i>˜ÊÍ̅>ÌÍ«>}iÊÍ>ĂiÍLiˆ˜}ÍĂi“œÛi`ÍvÜ“Í the buffer, lowering the efficiency of the cache and indicating the possibility of memory pres-ÊÕĂi°Í"˜ÍĂi«œĂ̈˜}ÍÊÞÊÌi“Ê]Í>Ê͜««œÊi`Í̜Í"/*ÍÊÞÊÌi“Ê]Í̅ˆÊ͘ՓLiĂ͓>ÞÍĂi“>ˆ˜Í>ÌÍ>͏œÜiĂÍ Û>ÕiÍʈ˜Vi͓œĂiÍ`>Ì>͈ÊÍ>VViÊÊi`ÍvܓÍĂi«œĂ̈˜}ÍÊÞÊÌi“Ê°Í
ÍĂi>ʜ˜>LiÍÛ>ÕiÍ̜ÍiÝ«iVÌÍ̜ÍÊiiÍ …iĂi͈ÊÍÎôôÍÊiVœ˜`Ê͜Ă͓œĂi°

Checkpoint Pages/sec

/…iÍ?da_glkejpL]cao+oa_ÍVœÕ˜ÌiĂÍĂi«ĂiÊi˜ÌÊÍ̅i͘ՓLiĂ͜vÍ«>}iÊÍ̅>ÌÍ>Ăi͓œÛi`Í̜Í`ˆÊŽÍLÞÍ

>ÍV…iVŽ«œˆ˜Ì͜«iĂ>̈œ˜°Í/…iÊi͘ՓLiĂÊÍʅœÕ`ÍLiÍĂi>̈ÛiÞ͏œÜ]ÍvœĂÍiÝ>“«i]͏iÊÊÍ̅>˜ÍÎôÍ«iĂÍ

ÊiVœ˜`ÍvœĂ͓œÊÌÍÊÞÊÌi“Ê°Í
ͅˆ}…iĂ͘ՓLiĂ͓i>˜Ê͓œĂiÍ«>}iÊÍ>ĂiÍLiˆ˜}͓>Îi`Í>ÊÍ`ˆĂÌÞ͈˜Í̅iÍ

V>V…i°Í
Í`ˆĂÌÞÍ«>}i͈Ê͜˜iÍ̅>Ì͈Ê͓œ`ˆvˆi`Í܅ˆi͈˜Í̅iÍLÕvviĂ°Í7…i˜ÍˆÌ½Ê͓œ`ˆvˆi`]͈̽Ê͓>Îi`Í

>ÊÍ`ˆĂÌÞÍ>˜`Í܈Í}iÌÍÜÈÌÌi˜ÍL>VŽÍ̜Í̅iÍ`ˆÊŽÍ`ÕȘ}Í̅i͘iÝÌÍV…iVŽ«œˆ˜Ì°Íˆ}…iĂÍÛ>ÕiÊ͜˜Í̅ˆÊÍ

counter indicate a larger number of writes occurring within the system, possibly indicative of

Ĩ"Í«ĂœLi“Ê°

Lazy writes/sec

/…iÍH]vusnepao+oa_ counter records the number of buffers written each second by the buffer

“>˜>}iĂ½Ê͏>đÞÍÜÈÌiÍ«ĂœViÊÊ°Í/…ˆÊÍ«ĂœViÊÊ͈ÊÍ܅iĂiÍ̅iÍ`ˆĂÌÞ]Í>}i`ÍLÕvviĂÊÍ>ĂiÍĂi“œÛi`ÍvܓÍ

the buffer by a system process that frees the memory up for other uses A dirty, aged buffer

ˆÊ͜˜iÍ̅>Ìͅ>ÊÍV…>˜}iÊÍ>˜`͘ii`ÊÍ̜ÍLiÍÜÈÌÌi˜Í̜Í̅iÍ`ˆÊŽ°Íˆ}…iĂÍÛ>ÕiÊ͜˜Í̅ˆÊÍVœÕ˜ÌiĂÍ

«œÊʈLÞ͈˜`ˆV>ÌiÍĨ"͈ÊÊÕiÊ͜ĂÍiÛi˜Í“i“œĂÞÍ«ĂœLi“Ê°Í/…iÍH]vusnepao+oa_ values should

consistently be less than 20 for the average system

Memory Grants Pending

/…iÍIaiknuCn]jpoLaj`ejc counter represents the number of processes pending for a

mem-œĂÞÍ}Ă>˜ÌÍ܈̅ˆ˜Í-+Í-iẶiĂ͓i“œĂÞ°ÍvÍ̅ˆÊÍVœÕ˜ÌiĂÍÛ>Õi͈Êͅˆ}…]Í̅i˜Í-+Í-iẶiĂ͈ÊÍʅœĂÌ͜vÍ

“i“œĂÞ°Í1˜`iĂ͘œĂ“>ÍVœ˜`ˆÌˆœ˜Ê]Í̅ˆÊÍVœÕ˜ÌiĂÍÛ>ÕiÍʅœÕ`ÍVœ˜ÊˆÊÌi˜ÌÞÍLiÍôÍvœĂ͓œÊÌÍ«Ăœ-duction servers

Another way to retrieve this value, on the fly, is to run queries against theÍ6Íouo*`i[

ata_[mqanu[iaiknu[cn]jpo A jqhh value in the column cn]jp[peia indicates that the process

ˆÊÍÊ̈ÍÜ>ˆÌˆ˜}ÍvœĂÍ>͓i“œĂÞÍ}Ă>˜Ì°Í/…ˆÊ͈Ê͜˜i͓i̅œ`ÍޜÕÍV>˜ÍÕÊiÍ̜ÍÌÜÕLiʅœœÌ͵ÕiĂÞÍ

̈“iœÕÌÊÍLÞ͈`i˜Ìˆvވ˜}Í̅>ÌÍ>͵ÕiĂÞÍ­œĂ͵ÕiÈiʮ͈ÊÍÜ>ˆÌˆ˜}͜˜Í“i“œĂÞ͈˜ÍœĂ`iĂÍ̜ÍiÝiVÕÌi°

Target Server Memory (KB) and Total Server Memory (KB)

P]ncapOanranIaiknu$G>% indicatesÍ̅iÍ̜Ì>Í>“œÕ˜Ì͜vÍ`ޘ>“ˆV͓i“œĂÞÍ-+Í-iẶiĂ͈ÊÍ

willing to consume Pkp]hOanranIaiknu$G>% indicates the amount of memory currently

>Êʈ}˜i`Í̜Í-+Í-iẶiĂ°Í/…iÍPkp]hOanranIaiknu$G>% counter value can be very high if the

ÊÞÊÌi“͈ÊÍ`i`ˆV>Ìi`Í̜Í-+Í-iẶiĂ°ÍvÍPkp]hOanranIaiknu$G>% is much less than P]ncap

OanranIaiknu$G>%]Í̅i˜ÍiˆÌ…iĂÍ̅iÍ-+Í-iẶiĂ͓i“œĂÞÍĂiµÕˆĂi“i˜Ì͈Ê͏œÜ]Í̅iÍi]toanran

iaiknuÍVœ˜vˆ}ÕĂ>̈œ˜Í«>Ă>“iÌiĂ͜vÍ-+Í-iẶiĂ͈ÊÍÊiÌÍ>ÌÍ̜œÍœÜÍ>ÍÛ>Õi]͜ĂÍ̅iÍÊÞÊÌi“͈Ê͈˜Í

warm-up

phase°Í/…iÍÜ>Ӈիͫ…>Êi͈ÊÍ̅iÍ«iȜ`Í>vÌiĂÍ-+Í-iẶiĂ͈ÊÍÊÌ>ĂÌi`Í܅i˜Í̅iÍ`>Ì>-L>ÊiÍÊiẶiĂ͈Ê͈˜Í̅iÍ«ĂœViÊÊ͜vÍiÝ«>˜`ˆ˜}͈ÌÊ͓i“œĂÞÍ>œV>̈œ˜Í`ޘ>“ˆV>ÞÍ>Ê͓œĂiÍ`>Ì>Í

sets are accessed, bringing more data pages into memory

9œÕÍV>˜ÍVœ˜vˆĂ“Í>͏œÜ͓i“œĂÞÍĂiµÕˆĂi“i˜ÌÍvܓÍ-+Í-iẶiĂÍLÞÍ̅iÍ«ĂiÊi˜Vi͜vÍ>͏>Ă}iÍ

˜Õ“LiĂ͜vÍvĂiiÍ«>}iÊ]ÍÕÊÕ>ÞÍx]ôôô͜Ă͓œĂi°

Memory Bottleneck Resolutions

When there is high stress on memory, indicated by a large number of hard page faults, you can ĂiʜÛi͓i“œĂÞÍLœÌ̏i˜iVŽÍÕʈ˜}Í̅iÍvœÜV…>ĂÌÍʅœÜ˜Íˆ˜Íˆ}ÕĂiÍӇx°

Figure 2-5 Memory bottleneck resolution chart

ÍviÜ͜vÍ̅iÍVœ““œ˜ÍĂiʜṎœ˜ÊÍvœĂ͓i“œĂÞÍLœÌ̏i˜iVŽÊÍ>ĂiÍ>ÊÍvœœÜÊ\

Í UÍ "«Ìˆ“ˆđˆ˜}Í>««ˆV>̈œ˜ÍܜΏœ>`

Í UÍ
œV>̈˜}͓œĂi͓i“œĂÞÍ̜Í-+Í-iẶiĂ

Í UÍ ˜VĂi>ʈ˜}ÍÊÞÊÌi“Í“i“œĂÞ

Í

Í

Í UÍ 1ʈ˜}͓i“œĂÞÍLiޜ˜`Í{ Í܈̅ˆ˜ÍÎӇLˆÌÍ-+Í-iẶiĂ

i̽ÊÍÌ>ŽiÍ>͏œœŽÍ>ÌÍi>V…ÍœvÍ̅iÊi͈˜ÍÌÕØ°

>`ấvè/iiịôấếấế>ịèi>}iấèôèv-+-iii? ?-+  -ii>>}iấiịLị}ĩ}ấ}èấiịôấi`ị>V>ị

You

canVv}ếi-+-iiv`ị>Viị>>}ii? ?-+ -ii>>}i-i? ?-? ?ế`ư -? ?èèiiịv`ivèi-ii*ôièiấ`>}Lí]>ấấĩ...

èiiấếViấ]è/ấấô>èVế>ịèiv>ấịấèi }-+ -ii7i-+

-iiấếèvV>Viưiịđ]>ôViấấĩ? ?-+ -iiưV>i`lazy writerđ>ấ

èĩiíèiấiịè>è>iế}viièi>iịô>}iấĩ? ?-+ -ii/ấ

L>Vè`ấ