Productivity of disk subsystem and SQL Of server 2000

[Silgey Av pedlov]

Gentlemen are specialists!
We put to use Navision Of axapta CIS SP3 before the two-unit realization. Is of the order of 80 users, the base 40 Gb of data, so many indices. Server of the data base on the platform SPSH4 (Shasta), 4 processors Of xeon MP of 25092L3 of cache down 400 MHz to systemic busbar without Hyper Of threading. Memory 12 Gb. Disk system: the two-channel RAID- controller LSI Of acceleRAID 320-2 (on one canal basket with 5 disks Of seagate Of cheetah of 10K.6 (capacity 146.8 Gb, velocity 30000rpm), on other canal basket with 5 disks Of seagate Of cheetah of 15K.4 (capacity 36.7 Gb, velocity 16900rpm); before the first basket of 4 disks in RAID-10 mass (on it basic data), 1 remained under system and SQL Of server with the utility bases; before the second basket of 4 disks in RAID-10 (on it the additional file of data)) and 1 remained under the ravines. Stripe Of size = of 64 K, Cluster of size (in NTFS) = of 64. Net is . One network adapter is begun to operate. WASPS Windows of 2003 Enterprise Of server SP1 of v.1433. Server of the base of data OF MS SQL Of server 2000 SP4. Support AWE in SQL Of server, /PAE, /3GB.
Essence of the problem: the brake SQL Of servera, judging as far as everything because of the low productivity of disk subsystem. => 1) as it is possible to follow the work of RAID- controller and disk masses (are there no apparatus problems); 2) is it possible to entrust to the counters of the standard Of performance Of monitor from Win2K3 from the section Of logical Of disk, such as parameters it is necessary to track what threshold values for these parameters; 3) can help of data of the logical volumes, on the disk RAID- masses?
The prehistory of a question - http://3mety.ro/viewtobec.hdm?d=5134 (the load of processor they managed of the optimization of the code of appendix - now like with this all before the order).

[js]

You will specify the model of controller. Mylex/LSI Of acceleRAID352 or LSI Of megaRAID320-2? Or 320-2?
The layout of disk you have radically inaccurate - it is better down the mass from the rapid disks to place fundamental base. Single disks from the point of view of reliability - generally welds.
You will look the disk counters of reads/sec, writes/sec, queue of lenght separately on all volumes. The first two counters for the evaluation of load, the third - indicator “does manage- not manage disk” (its value before the ideal must be 0, but it is permissible in the limits of units - although this already braking). as a rule will not lie.

[js]

on gives scanty increase.
How much cache on the controller is included on Wright Back?

[js]

Yes, to look not disk, but , after selecting there different volumes.

[Silgey Av pedlov]

Controller OF LSI Of megaRAID320-2 (without the X). Cache 128 Mb, write of back of , cashed IO also. On the slow disks basic given, since place it is not enough on the rapid (now on the mass from the slow disks Of used Of space = 245 Gb, Free Of space = 28,6 Gb).
From the point of view of - given on 10th , and ravines are cut off constantly (layout whimpering). - This layout is optimum with the existing quantity and the parameters of disks before the system.
On disk readings and records: recordings under 100 on the problematic volume, readings - on the order of 800, turn - on the order of 30.
About the logical and physical disks - you will explain, if you please, before than difference?

[js]

Before there are two counters - logical of disk and phisical of disk. Probably it is possible and to look, but we became accustomed . So to us it is simply clearer
Data of do speak, that disk does not manage (I I do understand, that turn 30- this on the average? If peak 30- this is bad, if average - this ). Well and + of under 1000 on four screws - this is limit. I.e. disk works from all forces and does not manage above .
Of recommendations are such:
1. it is budgetary. To pack the complete box of identical 15 of disks, to place the controller of 320-2 with the cache of 512 and everything before 5, after cutting down the moons on the needs. This is maximum, that it is possible to do without getting in the expenditures.
2. it is correct. External disk system of the type HDS AMS200 with the box (but better than two) of screws. As weaker versions - IBM DS4300 or FC-FC Of infortrend. This is more expensive, but it is more right from all points of view. Reliability first of all radically will raise, the secondly will make possible to be scaled before the future (for years forward).

[js]

By the way, it is possible you to ask to remove still data of the counters of read of bytes/sec and write of bytes/sec? Entirely well separately during the carrying out of transactions and reports. This to us before the plan of the collection of statistics for the more distinct recommendations to other users. I will be very grateful.

[Silgey Av pedlov]

The many thanks at the point of the operational response and the commentaries!
About “it is budgetary” and “it is correct” - we “will think”.
About the counters - operational reports are created in essence during the workday. Transactions also not only at night (in the daytime in essence). I do think, you will satisfy data, which me will succeed in gathering above workday?

[js]

Certainly. We try to gather statistics based on the work of different appendices with the disks - so that it is more distinctly the recommendation to give. But that before this matter blunder frequently strongly dearly is envelopped - as in you for example

[Silgey Av pedlov]

Thus, for the information - could not you , threshold values for read \ of sec and write \ of sec before our layout?

[js]

Well so you see exactly limiting values. The presence of turn means that the disk achieved the maximum of its possibilities, one cannot more rapidly spit no longer and commands are built before the turn to the disk. But processors smoke bamboo (by the way beside manager in this case completely it can be 100% load - simply they thresh time of waiting IO).

[Fadih]

Sergey Of petrov wrote (a):
On disk readings and records: recordings under 100 on the problematic volume, readings - on the order of 800, turn - order 30
" spread” load based on it to the remaining masses (unclearly, true, which on them now is created). If inputoutput is unevenly distributed (common reason - “they did not guess”, dividing off on the file groups, with 10 disks, on top of that different-sized, imho it was not generally worthwhile this to venture), to breathe will begin more easily, but nevertheless not the panacea
Record
- load (reads, writes, avg. queue of length) on all masses
- vacant place on the masses
- as is now beaten tempdb
- you do use the file groups
- the results of dbcc of sqlperf (waitstats)
possibly, anything yes will be devised
P.S I did correctly comprehend that the size is narrower than 80 gigas-byte?

[Silgey Av pedlov]

Load " smeared” already long ago, two greatest and loaded tables in us before another file group on other disk mass (which is less, but more rapid). By itself, the disk mass, beyond which lies a basic quantity of tables (which before appendix 906), tests large loads in comparison with the second.
On the great, slow and loaded mass it is free only more than 10% disk space. On smaller - it is more than 2/3.
Data on the counters I will lay out today nearer down to the evening (for the completeness of picture).
Still one question - although before our system begins strongly to predominate reading above the recording, there is one table, with which constantly are manufactured the operations of insertion-editing- removal. How on it is better to dispose the parameter of fill of factor for the indices? Zero now stand. The average value of Page split/sec for the server of approximately 2. does cost to generally ?
On waitstats:
Wait Type Requests Wait Time Signal wait time
MISCELLANEOUS 34.0 0.0 0.0
LCK_M_SCH_S 106.0 7.6105128E+740144.0
LCK_M_SCH_M 13.0 2.419348E+7 16.0
LCK_M_S 1224.0 2.8600186E+82782.0
LCK_M_U 14990.0 4.7855043E+810255.0
LCK_M_X 583.01743721.0 582.0
LCK_M_IS 194.0 1.9078072E+74709.0
LCK_M_IU 2777.0 7.3213870E+73580.0
LCK_M_IX 478.0 4.078748E+71167.0
LCK_M_SIU 0.0 0.0 0.0
LCK_M_SIX 0.0 0.0 0.0
LCK_M_UIX 0.0 0.0 0.0
LCK_M_BU 0.0 0.0 0.0
LCK_M_RS_S 0.0 0.0 0.0
LCK_M_RS_U 0.0 0.0 0.0
LCK_M_RIn_NL 0.0 0.0 0.0
LCK_M_RIn_S 0.0 0.0 0.0
LCK_M_RIn_U 0.0 0.0 0.0
LCK_M_RIn_X 0.0 0.0 0.0
LCK_M_RX_S 0.0 0.0 0.0
LCK_M_RX_U 0.0 0.0 0.0
LCK_M_RX_X 0.0 0.0 0.0
SLEEP 1739576.0 7.3647059E+8 7.0037563E+8
IO_COMPLETION 2822379.0 1.472788E+786546.0
ASYNC_IO_COMPLETION 81.02384.0 0.0
RESOURCE_SEMAPHORE 224.01809229.0 0.0
DTC 0.0 0.0 0.0
OLEDB 5364298.0 2.5484860E+8 4.0179436E+9
FAILPOINT 0.0 0.0 0.0
RESOURCE_QUEUE 9923891.0 2.094291E+9 7.071886E+8
ASYNC_DISKPOOL_LOCK 546.0 0.0 0.0
UMS_THREAD 0.0 0.0 0.0
PIPELINE_INDEX_STAT 42.050127.02156.0
PIPELINE_LOG 0.0 0.0 0.0
PIPELINE_VLM 0.0 0.0 0.0
WRITELOG 4365916.0 1.9332E+71815581.0
LOGBUFFER 1004.0120131.0 63.0
PSS_CHILD 0.0 0.0 0.0
EXCHANGE 0.0 0.0 0.0
XCB 0.0 0.0 0.0
DBTABLE 0.0 0.0 0.0
EC 1.0 16.0 16.0
TEMPOBJ 0.0 0.0 0.0
XACTLOCKINFO 0.0 0.0 0.0
LOGMGR 0.0 0.0 0.0
CMEMTHREAD 3107.0 424.0 424.0
CXPACKET 0.0 0.0 0.0
PAGESUPP 0.0 0.0 0.0
SHUTDOWN 0.0 0.0 0.0
WAITFOR 0.0 0.0 0.0
CURSOR 0.0 0.0 0.0
EXECSYNC 0.0 0.0 0.0
LATCH_NL 0.0 0.0 0.0
LATCH_KP 0.0 0.0 0.0
LATCH_SH 143.0 31.0 16.0
LATCH_UP 1588.0513988.0 390.0
LATCH_EX 3440.05800.0 93.0
LATCH_DT 0.0 0.0 0.0
PAGELATCH_NL 0.0 0.0 0.0
PAGELATCH_KP 6.0 0.0 0.0
PAGELATCH_SH 6509328.0336101.0224339.0
PAGELATCH_UP 798382.0196261.0136963.0
PAGELATCH_EX 8083369.0447986.0744415.0
PAGELATCH_DT 0.0 0.0 0.0
PAGEIOLATCH_NL 0.0 0.0 0.0
PAGEIOLATCH_KP 0.0 0.0 0.0
PAGEIOLATCH_SH 1.0625978E+8 3.2912651E+9 1.8549222E+7
PAGEIOLATCH_UP 23422.0048419.02173.0
PAGEIOLATCH_EX 5733438.0 1.4723337E+8358188.0
PAGEIOLATCH_DT 0.0 0.0 0.0
TRAN_MARK_NL 0.0 0.0 0.0
TRAN_MARK_KP 0.0 0.0 0.0
TRAN_MARK_SH 0.0 0.0 0.0
TRAN_MARK_UP 0.0 0.0 0.0
TRAN_MARK_EX 0.0 0.0 0.0
TRAN_MARK_DT 0.0 0.0 0.0
NETWORKIO 700286.0 5.9791853E+7 0.0
Total 2.4162867E+8 7.5059084E+9 5.4582007E+9
PAGEIOLATCH_SH - leader!

[Silgey Av pedlov]

tempdb lies beyond the disk mass for the sake of the fundamental base together, it is decomposed beside 4 files of data on 2 Gb each (about the number of processors).

[Silgey Av pedlov]

The file groups of 5 pieces, moreover that lies beyond the different masses, relates down the different file groups.





Down page 1,2, 3, 4, 5, 6 the trail

[Silgey Av pedlov]

This of turn down “slow” and “great” disk
D - Avg. Disk Of queue
Description:
Filesize: 186.65 KB
Viewed: 393 Time (s)
D - Avg. Disk Of queue

[Silgey Av pedlov]

These are disk turns on “rapid” mass with 2 tables
E - Avg. Disk Of queue
Description:
Filesize: 163.46 KB
Viewed: 283 Time (s)
E - Avg. Disk Of queue

[Silgey Av pedlov]

D - Disk
Description:
Filesize: 201.98 KB
Viewed: 273 Time (s)
D - Disk
D - Disk
Description:
Filesize: 196.84 KB
Viewed: 280 Time (s)
D - Disk

[Silgey Av pedlov]

E - Disk
Description:
Filesize: 175.83 KB
Viewed: 275 Time (s)
E - Disk
E - Disk
Description:
Filesize: 207.51 KB
Viewed: 273 Time (s)
E - Disk

[Silgey Av pedlov]

One admonition - today at night finished calculation statistician according to tables (95%); therefore large load was at night. Usually, if nothing it is launched beyond the night, then it is considerably less.
In other respects - common day load.

[js]

Well that to you to say. If we count these graphs at the point of the typical load, then you burn on money. Disk stands by crab. I already indicated the versions of decision.

[Jpret]

Good day! We based on Sergey Of petrov before one boat, so to say…
What pricks up ears - today objectively feels itself better. In particular, blockings, “long” queries and the rest it is by an order less than at the previous days. I would say that concerning these indicators we in effect are located before our common regime. It is subjective - even users practically do not squeak. In connection with this the question: and will not lie counters? With such turns, , we together with the base had to today entire day very quietly smoke before ......

[js]

Do can they simply they did become accustomed?
For the beginning try my first version - this on the money is not strongly dear, but at least must improve.
But that take at the point of the test . But then return backwards. You will see that it will be begun

[Fadih]

Sergey Of petrov wrote (a):
Load " smeared” already long ago, two greatest and loaded tables in us before another file group on other disk mass (which is less, but it is more rapid)
I had in mind reverse - thus far it leaves so that one mass in you it is bent, the second against this background almost stays (all, it goes without saying, relatively). I would examine version with one file group - and “hot” will be serviced by the large number of disks, and in “warm” it is more than chances to be placed on the bright mass
Quotation:
On the great, slow and loaded mass it is free only more than 10% disk space
It is unhealthy. In you there or for the development? You take away . The first - impact under the in the case of the failure of mass, the second regularly " draws off” the cache
Quotation:
The average value of Page split/sec for the server of approximately 2. does cost to generally ?
now - it does not stand
Quotation:
PAGEIOLATCH_SH - leader!
the fact
Quotation:
tempdb lies beyond the disk mass for the sake of the fundamental base together, it is decomposed beside 4 files of data on 2 Gb each (about the number of processors)
That it is decomposed - it is good. But that which lies beyond the mass is not. I had to leave the script, which ranks files on the activity of input-output. Tempdb, if me it does not change the memory, was loaded very even. On the whole, you resound - down the rapid mass, down the disk to the system (whole shaft, by the way, it stays)
Thus, it is time already to fast. But that light blinks - and regards. But then I will continue

[Fadih]

Sergey Of petrov wrote (a):
This of turn down “slow” and “great” disk
well that here you will say - it suggests. it is more accurate, it suppresses
Quotation:
These are disk turns on “rapid” mass with 2 tables
already more easily - reserve exists
Quotation:

oops, beside bytes/sec to be oriented more complexly than beside transfer/sec
nevertheless, load beyond the recording decent - very it seems that fouls tempdb, 80 to it is heavy to ensure the flow of introduced data under 2 megas-byte per second. You resound its
gs wrote (a):
Disk stands by the crab
stands, but not entire, but partially
we will try to play on this
gs wrote (a):
But that take at the point of the test . But then return backwards. You will see that it will be begun
- piece is good

[Silgey Av pedlov]

Apropos and the like - are better to be healthy and rich, than poor and patient.
On the occasions of the transfers: I apply pictures:
E - Avg. Disk
Description:
Filesize: 184.35 KB
Viewed: 280 Time (s)
E - Avg. Disk
D - Avg. Disk
Description:
Filesize: 172.13 KB
Viewed: 266 Time (s)
D - Avg. Disk

[Fadih]

Sergey Of petrov wrote (a):
Apropos and the like - are better to be healthy and rich, than poor and patient.
On the occasions of the transfers: I apply pictures:
do not count at the point of the impudence - it is possible instead of avg. disk of sec/transfer to give disk of transfers/sec?

[Silgey Av pedlov]

Now I will launch on transfer/sec, or tomorrow results I can publish down to the evening.

[Silgey Av pedlov]

By the way, let us in the evening throw all files of tempdb down “rapid mass”. Does can ? Although a little…

[Silgey Av pedlov]

Yesterday's something for the transfers of tempdb and reloading SQL of server. But thus far - that which before the frequency of transactions in yesterday.
E - Avg. Disk of transfers_
Description:
Filesize: 195.11 KB
Viewed: 264 Time (s)
E - Avg. Disk of transfers_
D - Avg. Disk of transfers_
Description:
Filesize: 168.67 KB
Viewed: 268 Time (s)
D - Avg. Disk of transfers_

[Silgey Av pedlov]

Thus, week we lived based on the new disk stable (IBM DS4300, 14 disks on 68 Gb each, 60000Rpm, RAID 10 down all 14 disks, it was decomposed beside 2 logical ; of 2 controllers it is begun to operate only 1 (as they said, down the period of testing - the second is under working conditions necessary for purposes of redundancy)). But so that we do have against the given moment? I will compare with the data, which I published before our times without the counter (on one of the logical disks, which is most of all loaded).
Disk turns on the counter:
G - Avg Of disk Of queue
Description:
Filesize: 218.91 KB
Viewed: 224 Time (s)
G - Avg Of disk Of queue

[Silgey Av pedlov]

Average value - 5. it was - 17.

[Silgey Av pedlov]

Disk Of transfers/sec on the counter:
G - Disk Of transfers_
Description:
Filesize: 193.27 KB
Viewed: 188 Time (s)
G - Disk Of transfers_

[Silgey Av pedlov]

The average value of now 340, was 602.

[Fadih]

with the calculation of average values the meaning, probably had not to consider periods of low load (night)
about the doubly low of disk of transfers - thus far without the commentaries
number by numbers, and as sensation- that after upgrade?

[Silgey Av pedlov]

We adhere as always. By the way, yesterday's graphs are not possibly so significant, because the close of period yesterday occurred. The number of blockings of the user processes was increased. On the whole, not panacea. The first sensations are such.

[Silgey Av pedlov]

Before the addition - data on Disk of read of bytes/sec:
G - Disk Of read Of bytes_
Description:
Filesize: 223.81 KB
Viewed: 178 Time (s)
G - Disk Of read Of bytes_

[Silgey Av pedlov]

Sorry, mistake moved out…

[Silgey Av pedlov]

Not, all like before the order.
Thus, average with the stable: 5055578, it is earlier than 7637289

[Silgey Av pedlov]

Data on Disk of write of bytes/sec now:
G - Disk Of write Of bytes_
Description:
Filesize: 220.85 KB
Viewed: 186 Time (s)
G - Disk Of write Of bytes_

[Silgey Av pedlov]

Average now with the stable: 164387, were earlier than 1616873.
On the whole, so… But day yesterday surely actually was uncharacteristic.
Now I will try to lay out data in the more characteristic day - medium.

[Silgey Av pedlov]

Given above on Wednesday March 15:
G - Avg Of disk Of queue
Description:
Filesize: 218.02 KB
Viewed: 189 Time (s)
G - Avg Of disk Of queue

[Silgey Av pedlov]

Transmissions of the data
G - Disk Of transfers_
Description:
Filesize: 186.45 KB
Viewed: 184 Time (s)
G - Disk Of transfers_

[Silgey Av pedlov]

Readings
G - Disk Of read Of bytes_
Description:
Filesize: 219.86 KB
Viewed: 175 Time (s)
G - Disk Of read Of bytes_

[Silgey Av pedlov]

Recordings:
G - Disk Of write Of bytes_
Description:
Filesize: 203.57 KB
Viewed: 177 Time (s)
G - Disk Of write Of bytes_

[js]

You see what plug. In you the data were on two volumes, on which the turn was respectively 17 and 4. now one volume with turn 4, screws it became temporarily more. Strictly already it is substantially better. But it is small, tons to turn in effect constant. It is apparently must still more than the screws
By the way, why not to do two moon and not to give out them through two controllers? The total capacity of cache will become more.
But as on the sensations?

[Silgey Av pedlov]

Turns: there were 17, it became 4 on the average; the transmission of the data: there were 602, it became 298; reading: were 7637349, became 6096508; the recording: were 1376408, became 1719910. Given everything I lead “on the average”, because the type of load is approximately identical (night period without the particular temporary changes, day also). Output: disk turns were reduced with the retention of throughput capacity. How to explain the decrease of disk transfers/sec - I do not know.

[js]

But not of by hour?

[Silgey Av pedlov]

I do not know about - it is necessary to request in them. But it was changed like radically nothing. About LUN - on the advice of coworkers Trinity was created one great mass RAID10, on which were cut 2 logical (by primary for them it stands the existing controller, Write of caching of with of mirroring of ). The second controller to us simply they did not give with the assembly they were stable. I rang on this score - Gennadiy said that he, before principle, was not necessary. I.e., and without it will work. Is good b, of course, of the second, but also so it is possible. For the testing. Sergey Seleznev promised to appear on this week, but so he did not appear. Obviously, I must itself was about itself mention. I hope that on the following week, which will mandatorily be better than flowing (), we will be met.

[Stlanger03]

Sergey Of petrov wrote (a):
I do not know about - it is necessary to request in them. But it was changed like radically nothing
, before this layout the second controller to you is not actually necessary. The second controller to you will be brought together with the server, .

[Stlanger03]

Indeed we with you have conversation about the fact that the existing disk system survives not better time. Here therefore to you was the second time given disk counter earlier. Originally we planned to let counter for the sake of the server together.
To arrive before you is on Monday must was Peter, but his journey was cancelled at the point of some reasons (I it was at this time in Moscow). On Dmitriy 's letter our aid to you before this question is not necessary, now I do see that it is nevertheless necessary?

[Silgey Av pedlov]

You do see, Gennadiy, the aid of Peter would be at the point of us necessary, if I independently could not assemble counter, and do a system of worker. Question before doing of a system still of of more productive. It was last Monday at the point of me possible to be connected for the sake of your by coworker Sergey , who recently returned himself from the journey to London (it seems). I attempted to pose to it my questions - only for the confirmation of that fact that I did all correctly and my activities will not be covered for the system by irreparable consequences. It answered me, that all like before the order, but better it itself will arrive before us with the second controller, will establish it, will dispose system down the work with two controllers and on this all. Unfortunately, I myself was connected for the sake of Sergey on Tuesday (on on the telephone, not on ICQ), and he also did not make contact. But now still and server before us will arrive on the following week. It is interesting, will save us further systemic ? :. thus far it is not confident.

[CyberDloke]

I please forgiveness at the point of my nonarrival, organizational tie plate occurred, apparently we with Gennadiy comprehended each other.
If only one logical disk is planned, then the addition of the second controller will raise fault tolerance, but it will not give an increase in the productivity, since queries to the disk of by one controller. In this case the controller- owner of logical disk can be replaced by hand. By the way it is possible to try to do two logical disks on the mass, which should be joined beside stripe due to means OS, then with the work are begun to operate 2 controllers, which in a number of cases gives a considerable increase in the productivity, but, in addition this must be tested on the specific objectives.
The last time it edited: CyberDrake (Mar 20, 2006 the 12:28 Of ), in all edited 1 time

[Stlanger03]

Sergey Of petrov wrote (a):
You do see, Gennadiy, the aid of Peter would be at the point of us necessary, if I independently could not assemble counter, and do a system of worker. Question before doing of a system still of of more productive.
Perhaps not this we with you did try to do, until now? Can you me ring or beside for resolution of our questions? I think to us with you sufficient to .

[Stlanger03]

And already Sergey, with Sergey was a real example on task one beside one similar with your task. Why I to you recommended the doing of one logical disk, it is exact because based on our example the laying-out of mass down three groups under the data, ravines and indices led down a decrease in the productivity 3 times. In this case although the turn to the disks and greater, is nevertheless less than in the case 320- GO of controller. Perhaps no?

[Silgey Av pedlov]

Before principle, sole, that he wished to explain - the correctness of the selection of options during the installation and the initialization they were stable.
Namely, the option “Of enable of write of caching of without of batteries” I disconnected, since the pile exists, the option” Of enable of write of caching of with of mirroring " was switch oned.
Further, during the initialization of logical before I indicated the type of “dynamic” (it like it makes it possible to dynamically increase the size of volume) and before the properties of volume it established “Policies \ Of enable of advanced of performance”, which permits at the point of system to hope at the point of and to write data down the disk somewhat more rapidly. Although, before principle, surely this option has beyond us little effect, since prevails reading, but not recording.
Now all before principle does work, but it would be desirable to know, can I something not so or not to the end he did do? I am turned down your experience.
Still one important moment - at the point of us is necessary recurrent testing. This time with the server from IBM. If disk mass with already available on it data it is simple to translate on new server - possible whether to do this then so as to anew not all given and as this it will be purely technically realized? How long ? We simply do catch volume to and base?
Now about one logical - before principle, all possibly and we can try before the new layout with the new server not to divide off the existing space beside several logical , but to create one great and on it to already accommodate all data. What in that case to make with the file groups in SQL Of server? To translate all files of data beside one or to leave as are (now our basic given they are decomposed beside 2 file groups for the diversity on the different logical volumes)? Theoretically difference for the base none, and here practically…
I think, a latter question must be considered, also, with our programmer.
About reduction in the disk turns - result available. To what extent is acceptable that level, which is present now? What threshold value for this parameter with our stable? On the general productivity of appendix - little which was changed (however strangely). Can, still before by something? How it is possible to reveal?