A very important method of increasing data access efficiency is to cache frequently accessed data. like country list or top articles.. in situations where the data changes less frequently, such as once a day or every few hours, it is inefficient to hit the database for every page request.
As many would suggest to cache the HTML Output of a portion of a page(Fragment Caching) as a country select box or caching the entire page(Output Caching), there are cases where Data Caching will be more powerful. for example you may need to display that list of countries in more than one page/Portion with different look and/or use that list programmatically.
In a previous post [Data Access Layer for Classic ASP]: I proposed DAL Classes to encapsulate Data Access code for a certain Table/Object, and now I'll extend the DAL to handle Data caching too, please refer to my previous post before continuing..
I've modified the "Users" class to handle saving and reading from cache, which is saved to ADO XML format or Advanced Data Tablegram (ADTG) Binary format using "Save" method of ADO Recordset, Which has many advantages :
- You don't need to change much of your Data Access code to read from cache
- Opening the recordset from XML will act the same as Opening Database when you use Paging
- You can use The "Filter" method of ADO Recordset to filter records as you would do in a SQL Query
Here is the New ClassClass Users
'private Arr to hold users Objects
Private arr
'0 : Binary ADTG , 1: XML
Public CacheType 'as Byte
' Cache File Path
Public CachePath 'as String
''''' Paging Variables
Private mPageCount 'as Long
Public PageIndex 'as Long
Public PageSize 'as Long
'for getting top records if larget than 0
Public TopRecords 'as Long
'destroy arr
Private Sub Class_Terminate()
Erase arr
end Sub
Private Sub Class_Initialize()
Initialize
TopRecords = 0
PageIndex = 0
PageSize = 0
CacheType = 0
CachePath = ""
end Sub
'Initialize collection
Public Sub Initialize()
redim Arr(-1)
mPageCount = 0
End Sub
'return Upper Bound of objects array
Public Property get UpperBound()
UpperBound = UBound(arr)
End Property
'return object at index i
Public Property Get ElementAt(i)
set ElementAt = arr(i)
End Property
'return Upper Bound of objects array
Public Property Get PageCount()
PageCount = mPageCount
End Property
'Select users by Status ( u can add more search parameters or create another selectUsers method)
Public Sub SelectUsers(Status)
Initialize
Dim rs : Set rs = server.CreateObject("ADODB.Recordset")
'Do paging
If PageSize>0 And PageIndex>0 Then
RS.CursorLocation = adUseClient
RS.CacheSize = PageSize
RS.PageSize = PageSize
End If
'if open from XML
If CachePath>"" Then
If Status>0 Then rs.Filter = "Status="& Status
Rs.Open CachePath,,,,adCmdFile
'if Open from DB
Else
If TopRecords>0 Then
rs.MaxRecords = TopRecords
Top = " top "& TopRecords &" "
End If
Dim SQL : SQL= "SELECT "& Top &" * From users"
If Status>0 Then SQL = SQL & " where Status="& Status
rs.Open SQL , ConnStr, adOpenForwardOnly,adLockReadOnly,adCmdText
End If
' if paging : get page count
If Not rs.EOF And PageSize>0 And PageIndex>0 Then
RS.AbsolutePage = PageIndex
mPageCount = RS.PageCount
End If
Dim i : i=0
'(TopRecords=0 Or i<TopRecords) condition is needed to get correct Top N records when opening from Cache
' ,MaxRecords doesn't seem to work in that case
While Not rs.EOF And (TopRecords=0 Or i<TopRecords) And (PageSize=0 Or i<PageSize)
'loop until EOF or Paging Size reached
'create Object and set Values then add to collection
Dim u : Set u = New User
u.ID = CLng(Rs("ID"))
u.Name = Rs("Name")
u.Email = Rs("Email")
u.Password = Rs("Password")
u.LastLogin = cdate(rs("LastLogin"))
u.Status = cbyte(Rs("Status"))
ReDim Preserve arr(i)
Set arr(i) = u
set u = Nothing
rs.MoveNext
i=i+1
Wend
rs.Close
Set rs = Nothing
End Sub
' Open Recordset and Save it
Public Sub CacheUsers(Status)
If CachePath="" Then Exit Sub
Dim rs : Set rs = server.CreateObject("ADODB.Recordset")
Dim Top
If TopRecords>0 Then
rs.MaxRecords = TopRecords
Top = " top "& TopRecords &" "
End if
Dim SQL : SQL= "SELECT "& Top &" * From users"
If Status>0 Then SQL = SQL & " where Status="& Status
rs.Open SQL , ConnStr, adOpenForwardOnly,adLockReadOnly,adCmdText
Call SaveRS(rs)
rs.Close
Set rs = Nothing
End Sub
'Handle saving Recordset to Stream
Private Sub SaveRS(rs)
Const adTypeText = 2
Const adTypeBinary = 1
Const adSaveCreateOverWrite = 2
Const adPersistXML = 1
Const adPersistADTG = 0
Dim Stream : Set Stream = Server.CreateObject("ADODB.Stream")
If CacheType=1 Or CacheType=2 Then
Stream.Type = adTypeText
Stream.Open
rs.Save Stream, adPersistXML
Else
Stream.Type = adTypeBinary
Stream.Open
rs.Save Stream, adPersistADTG
end If
Application.Lock
Stream.SaveToFile CachePath,adSaveCreateOverWrite
Stream.Close
Application.UnLock
Set Stream = Nothing
End Sub
End Class
Sample Usage
1- Save Cache when Target Data is updatedDim aUsers : Set aUsers = New Users
aUsers.CacheType = 1 'XML cache
aUsers.CachePath = Server.MapPath("users.xml")
aUsers.TopRecords = Top
Call aUsers.CacheUsers(0)
Set aUsers = Nothing
2- Read CacheDim aUsers : Set aUsers = New Users
aUsers.CacheType = 1 'XML cache
aUsers.CachePath = Server.MapPath("users.xml")
aUsers.TopRecords = Top
Call aUsers.SelectUsers(0)
Dim i,aUser
For i=0 To aUsers.UpperBound
Set aUser = aUsers.ElementAt(i)
'do something
Set aUser = Nothing
Next
Set aUsers = Nothing
Time Comparison
I did a timing test to compare between reading from a SQL Server and from XML/Binary file, but Reading from a local SQL Server (on the same machine as IIS) will be always faster, so to get real world results i simulated a LAN/WAN Connection on SQL Server [How to? check: Simulating WAN Connection for SQL Server Performance Testing].
And to simulate concurrent connections on IIS and calculate timing i used the very neat [Microsoft Web Stress Tool] on a PC of P3.2GHZ, 1GB Ram, Windows XP, SQL Server 2005 Express.
method / Records | 100 | 500 | 1000 | 3000 |
---|---|---|---|---|
Open WAN SQL Server | 559.25 | 614.80 | 980.85 | 2489.85 |
Open LAN SQL Server | 276.70 | 436.00 | 777.60 | 2368.90 |
Open XML Cache | 72.25 | 369.95 | 821.10 | 2469.31 |
Open Binary Cache | 60.45 | 311.25 | 666.45 | 2062.20 |
As you can see reading less than 1000 records from XML Cache is slightly faster than reading from LAN SQL Server and time saving gets bigger when compared to WAN SQL server(reading >1000 records gets slower in XML due to loading&parsing overhead).
You can also notice that reading from Binary(ADTG) is faster than reading from XML and LAN/WAN SQL Server in all cases, and files created in binary are Approx. 50% smaller than the equivalent XML files.
I'm curious, why are you saving via a stream instead of rs.Save(path, ADTG)? Is there an advantage? Something wrong with using Open/Save from the recordset instead?
Also, why the Application.Lock/Unlock? Does that prevent simultaneous Save calls from colliding/file "deadlocking"?
I'm working on something for my company's website, we have some legacy sections of the site that were terribly written and when the crawlers hit them, it really starts to drag. So I'm looking into this method, by storing the saved recordsets data on the web server and having a worker process which handles the deleting of cache (which the systems will automatically recreate if the cache file doesn't exist) either periodically or via a separate table populated by triggers to signal a cache refresh is needed.
Thanks and great work!
@Rob,
thanks rob for visiting my blog..
I guess I used stream to overwrite existing files..
Yes, application lock to avoid deadlocks..
You can delete cache every time the data is modified, and create it when cache file is not found. In the case that data get modified few times and accessed many times..