Interntionl Journl of Comuter Alictions (975 8887) Volume 5 No.2, August 22 Multi HLR Architecture for Imroving Loction Mngement in PCS Network Rjeev R. Kumr Trithi G. S. Chndel Rvindr Gut SSSIST SSSIST SSSIST Sehore (M.P) Indi Sehore (M.P) Indi Sehore (M.P) Indi ABSTRACT This er rooses multi HLR scheme for loction mngement in PCS network insted of hving single HLR. In conventionl rchitecture we hve single HLR tht cts s centrlized dtse to store the user rofile long with the s loction informtion. This roch of single HLR suffers from two mjor disdvntges: ottleneck nd cll misrouting during the ek lod. In multi HLR rchitecture, we store the users rofile nd s loction informtion zone wise or re wise. By doing so, we minimize the ossiility of oth ottleneck nd cll misrouting. In conventionl rchitecture, we use exlicit de-registrtion scheme for de-registrtion of n from VLR on move. Severl de-registrtion schemes hs roosed reviously in site of exlicit de-registrtion scheme s: imlicit de-registrtion scheme, olling de-registrtion scheme, timeout de-registrtion scheme nd grou deregistrtion scheme. Performnce nlysis of ll these schemes shows tht grou de-registrtion scheme is est mong existing nd roosed schemes. In the roosed multi HLR rchitecture grou de-registrtion scheme is imlemented insted of exlicit de-registrtion scheme. Result shows tht roosed rchitecture with grou de-registrtion scheme is more efficient thn exlicit de-registrtion scheme. Keywords Bottleneck, Cll misrouting, Imlicit de-registrtion scheme, Polling de-registrtion scheme, Timeout de-registrtion scheme, Grou de-registrtion scheme, Exlicit de-registrtion scheme.. INTRODUCTION Current roch to suort user moility in GSM nd IS-4 stndrd, we use two-level dtse rchitecture, one is clled home loction register (HLR) nd nother is clled visitor loction register (VLR). This HLR-VLR rchitecture is centrlized in nture [6]. In GSM nd IS-4 stndrd we hve one HLR nd severl VLRs. In HLR we store the user s informtion on ermnent sis while in VLRs this storge is temorrily. HLR cts s the mster dtse for the entire system. VLRs cche the user s rofile from this mster dtse. As we hve hysiclly nd logiclly only one coy of the HLR hence its mngement is esy ut it suffers from ottleneck [, 2]. As er move, informtion stored in HLR nd VLR chnges. When n leves its RA nd enters into new RA eing served y the sme VLR, this informtion is chnged only in the VLR. But when n leves its RA nd enters into the new RA tht is eing served y the different VLR, the s informtion is udted in HLR nd this informtion is deleted from the old VLR [, 2, 3, 4, nd 5]. Now we cn differentite the user s moility in two wys in conventionl HLR-VLR rchitecture... Intr-VLR Move: This move occurs when s RA chnged ut VLR is sme. In this cse, s current loction informtion is udted in the VLR. This move does not ffect the HLR..2. Inter-VLR Move: This move occurs when chnges its RA nd new RA is eing served y the new VLR. This move is shown in fig.it hs six stes. () (4) VLR (5) (2) (6) (3) MSC Cell Registrtion re oundry Bse sttion Fig : Exlicit de-registrtion HLR Ste : When moile terminl enters into new RA, it sends loction udte to the nery se sttion. Ste 2: The se sttion sends this informtion to the se sttion controller (BSC) through which it is eing governed. The se sttion controller sends this informtion to the MSC/VLR. Ste 3: The new MSC/VLR udtes its VLR which shows tht the moile terminl is residing in its service re nd informs the HLR y sending registrtion messge to the HLR. Ste 4: The HLR sends registrtion cknowledgment messge to the new MSC/VLR together with coy of the suscrier's user rofile. This informtion is used to vlidte the suscrier. Ste 5: The HLR sends registrtion cncelltion messge to the old MSC/VLR Ste 6: The old MSC/VLR removes the record of the moile terminl t its ssocited VLR nd sends cncelltion cknowledgment messge to the HLR. Vrious HLR imlementtions re ville. SONOFON GSM reort tells tht if HLR suorts 3,customers nd user erform 2 HLR oertions er dy including loction udte, routing, uthentiction nd network ttchments, we hve totl 6, oertions er dy. During ek hour s 2-3% oertions for loction udte, network ttchment nd 42
Interntionl Journl of Comuter Alictions (975 8887) Volume 5 No.2, August 22 uthentiction come nd totl lod ecomes 8, queries er hour (222 trnsctions er second). This ek lod even my e 5% higher. Such hevy lod cnnot e suorted y stndlone HLR. As consequence HLR suffers from ottleneck nd cnnot gurntee the qulity of service [6]. In conventionl HLR-VLR scheme, de-registrtion of n from VLR is lwys exlicit. Exlicit in the sense tht stle entries of VLRs re removed with the hel of HLR. Actully HLR sends de-registrtion messge to the VLR to remove the stle entries when n chnges its VLR. This exlicit deregistrtion increses the totl cost y incresing the trffic lod. To reduce the trffic lod following de-registrtion strtegies were roosed. residing in registrtion re moves to nother registrtion re. & re eing served y the sme VLR, VLR. The VLR is eing served y the resident-hlr. Due to movement of, this loction udte is chnged t VLR not t resident-hlr. Resident HLR (A) Distnce Bsed De-registrtion Scheme. (B) Time-Bsed De-registrtion Scheme. (C) Polling-Bsed De-registrtion Scheme. (D) Grou De-registrtion Scheme. (E) Movement-Bsed De-registrtion Scheme. Performnce nlysis of de-registrtion strtegies in Personl Communiction Network shows tht the grou de-registrtion scheme is est scheme mong time nd olling sed deregistrtion schemes. This er long with the limittions of the conventionl HLR-VLR roch motivtes to imlement the grou de-registrtion scheme in the multi HLR rchitecture []. 2. REPLICATION OF HLR Chen nd Lee roosed n ide of relicting the HLR to void the ottleneck. In this roch we mintin severl relics of the HLR. In cse of high lod, relics re used insted of the mster-hlr [6]. If roility of filure in n HLR is then vilility of n HLR under high lod condition will e (- n ), if we hve n HLRs (including mster-hlr nd (n-) relics of HLR)[7]. However to mintin consistency of informtion in the existing relics will introduce the extr costs. Agin it is to e noted tht this rchitecture is even centrlized not distriuted hving only one dvntge of removl of ottleneck ut in cse of high lod when relics re not correctly udted, misrouting of clls will tke lce. 3. PROPOSED MULTI HLR ARCHITECTURE In the roosed rchitecture, we hve severl HLRs zone wise or circle wise insted of single HLR. It reduces the storge overhed of the HLR. Ech HLR cn serve more thn one VLR nd ech VLR cn serve more thn one RAs. Simly we cn sy tht this rchitecture contins severl conventionl HLR-VLR rchitectures. For ech we define two tyes of HLRs: resident-hlr nd serving-hlr. Resident-HLR is the HLR where often resides. While on move, it cn enter into the RA eing served y nother HLR (serving-hlr). When will e served y the HLR other thn resident-hlr, we will refer it s roming. In the roosed rchitecture we define following tyes of move s: 3.. Intr-VLR-Resident-HLR Move: In this tye of move, the chnges its RA nd the new RA is still eing served y the sme VLR. The serving VLR is eing served y the resident-hlr. Now it is ovious tht the loction udte is tking lce only t VLR not t resident-hlr. Intr-VLR-Resident-HLR move is shown in fig (2). An Fig 2: Intr-VLR-Resident-HLR move 3.2. Intr-VLR-Serving-HLR Move: In this tye of move, the chnges its RA nd the new RA is still eing served y the sme VLR. The serving VLR is eing served y the serving- HLR. Agin this informtion is only udted t VLR not t serving-hlr nd resident-hlr. Serving HLR Fig 3: Intr-VLR-Serving-HLR move Intr-VLR-Serving-HLR move is shown in fig (3). An residing in registrtion re moves to nother registrtion re. & re eing served y the sme VLR, VLR. The VLR is eing served y the serving-hlr. Due to movement of, this loction udte is chnged t VLR not t serving-hlr nd resident-hlr. 3.3. Inter-VLR-Resident-HLR Move: In this tye of move, the chnges its RA nd the new RA is eing served y the new VLR. The serving VLR is eing served y the resident- HLR. Now in this cse registrtion of will tke lce t new VLR, de-registrtion of will tke lce t old VLR nd finlly resident-hlr will udte this informtion in its dtse. Inter-VLR-Resident-HLR move is shown in fig (4). An is residing in registrtion re which is eing served y the 43
Interntionl Journl of Comuter Alictions (975 8887) Volume 5 No.2, August 22 VLR, VLR. This VLR is eing served y the resident-hlr. On move chnges its registrtion re nd comes in. This is eing served y the VLR 2.Now the registrtion of will tke lce t VLR 2 nd de-registrtion will tke lce t VLR. This chnge in loction udte will tke lce on resident-hlr. Resident HLR 2) On recetion of this messge serving-hlr de-registers this from its dtse nd informs to its VLR (where ws registered reviously) for de-registrtion. This cse is shown in fig (6). is residing in the registrtion re of VLR 2 of serving-hlr. On move comes under VLR 3 of resident-hlr. In this cse oth VLRs nd HLRs chnge. The is now in its resident- HLR. Registrtion of will tke lce t VLR 3 nd resident-hlr while it s de-registrtion will tke lce t serving-hlr nd its corresonding VLR where ws reviously resided. Serving HLR Resident HLR VLR 3 VLR 4 Fig 4: Inter-VLR-Resident-HLR move 3.4. Inter-VLR-Serving-HLR Move: In this tye of move, the chnges its RA nd the new RA is eing served y the new VLR. The serving VLR is eing served y the serving-hlr. Now in this cse registrtion of will tke lce t new VLR, de-registrtion of will tke lce t old VLR nd finlly serving-hlr will udte this informtion in its dtse. However resident-hlr will not e udted. Serving HLR Fig 5: Inter-VLR-Serving-HLR move Inter-VLR-Serving-HLR move is shown in fig (5). An is residing in registrtion re which is eing served y the VLR, VLR. This VLR is eing served y the serving-hlr. On move chnges its registrtion re nd comes in. This is eing served y the VLR 2.Now the registrtion of will tke lce t VLR 2 nd de-registrtion will tke lce t VLR. This chnge in loction udte will tke lce on serving-hlr. Resident-HLR will remin unffected. 3.5. Inter-VLR-Inter-HLR Move: This tye of move occurs into two cses. ) This tye of move occurs when n comes ck to its resident-hlr from serving-hlr. In this move s registrtion occurs t serving VLR of resident-hlr nd this informtion is sent ck to the serving-hlr. RA 3 Fig 6: Inter-VLR-Inter-HLR move 3) When n moves to nother serving-hlr. In this cse registrtion of tkes lce t new serving-hlr nd its VLR under which comes, de-registrtion occurs t the revious serving- HLR nd its ssocited VLR from where is coming nd resident-hlr is udted. 4. GROUP DE-REGISTRATION SCHEME IN PCS NETWORK In grou de-registrtion scheme, HLR mintins n OML (old moile list) for ech VLR. When n chnges its VLR on move, HLR kees this into the OML of reviously serving VLR. When loction registrtion request comes for n, HLR sends this OML to the VLR s rt of registrtion cknowledgement. VLR uses this OML to remove the entries of s who hve lredy left it. This scheme cn e descried into four stes: ) When n chnges its VLR, the new serving VLR sends registrtion request to the HLR. 2) On recetion of this request, HLR udtes the loction of nd uts this s identifiction into the OML of revious VLR from where is coming. 3) HLR sends registrtion cknowledgement to the VLR long with its OML. Sent OML is lwys ket emty. 4) On recetion of this cknowledgement, VLR finds the informtion of those who hve left it nd remove their informtion for its dtse. Thus the invlid s identifictions re removed from the VLR every time new enters the RA [, 3]. 44
Interntionl Journl of Comuter Alictions (975 8887) Volume 5 No.2, August 22 5. PERFORMANCE ANLAYSIS OF VARIOUS DEREGISTRATION STRATETGIES IN PCS NETWORK Performnce nlysis of vrious de-registrtion strtegies like exlicit de-registrtion, imlicit de-registrtion, timeout deregistrtion nd grou de-registrtion strtegies re evluted []. Result shows tht cost incurred in the grou de-registrtion strtegy is greter thn timeout nd olling de-registrtion scheme ut less thn exlicit de-registrtion strtegy. However in cse of grou de-registrtion strtegy no stle entry regrding the moile terminl resides in the VLRs. In timeout deregistrtion strtegy, removl of stle entries of s re deendent on the time. In this scheme, s eriodiclly register themselves to VLRs so tht it cn ensure its resence into the VLRs. This scheme is refinement to the olling de-registrtion scheme in which VLR rodcsts olling messges to collect the cknowledgement of the s residing in the RAs ssocited with the VLR. The olling scheme is lso deendent on the time. Process of de-registrtion in exlicit de-registrtion scheme is sed on moility of the terminls. Grou deregistrtion strtegy rovides us to de-register the moile terminls on moility sis hving less cost thn the exlicit one. In this roosed rchitecture, we re imlementing the grou de-registrtion strtegy in multi HLR rchitecture nd evluting its erformnce ginst exlicit de-registrtion scheme. 6. PERFORMANCE ANALYSIS An nlyticl model to evlute the erformnce of multi HLR rchitecture hs een resented here long with the grou deregistrtion strtegy imlemented in sme rchitecture. Here ll HLRs re locted t the sme lyer nd they re communicting ech other in oint-to-oint sis. In this nlysis, hierrchl trees of R lyers re eing used. The lyer R contins the roots of ll trees nd leves of ll trees re t the level. It mens oth roots nd leves reside on the sme lyer. Following terms re eing used in the erformnce nlysis:- m x,y Lyer of the closest common node to RA x nd RA y. Proility tht the move is intr-vlr. n New RA of the. Old RA of the. α Proility of n to reside in its resident-hlr. Proility of n to reside in first serving-hlr. β Proility of n to reside in second serving-hlr. P (m x,y =i) is defined s the roility tht the closest common node to LA x (RA x) nd LA y(ra y) is in lyer i. This roility cn e given y the following eqution. P(m,n =i) = (-) i- for i =,2..R- (-) i- for i = R (2) T (i, j): Cost of trnsmitting messge over link etween two djcent lyers i nd j. C m (i): Cost of ccessing or udting dtse in lyer i. M multihlr-vlr (exlicit) : Estimted cost of loction udte in the exlicit multi HLR-VLR scheme. M mutihlr-vlr (grou) : Estimted cost of loction udte using grou de-registrtion scheme in multi HLR-VLR rchitecture. Estimted cost of loction udte in exlicit stnd lone HLR- VLR scheme is given s: [ ] { } The first rt of Eq. (3) is the cost of loction udte in intr- VLR move. The second rt illustrtes the scenrio fter n inter-vlr move. T (, L) =T (, 2) +T (2, 3) +..+T (L-, L) is equl to the cost of trversing links etween node of lyer (i.e., VLR) nd the node of lyer R (i.e., where n HLR is locted). This cost is multilied y 4 ecuse new VLR sends registrtion request to the HLR, the ltter sends cncelltion request to the old VLR, old VLR sends n cknowledgement in resonse to the HLR nd finlly HLR confirms the registrtion of new t the new VLR. Trnsmission cost of the messge is descried s follows: T(,L)=T(,2)+T(2,3)+..+T(L-,L) T (, 2) will give the result 2; T (2, 3) will give the result 3 nd so on. Estimted cost of loction udte with grou de-registrtion scheme is given s follows: [ ] { } The first rt of this Eq. No.(4) is the cost of loction udte in intr-vlr move. The second rt illustrtes the scenrio fter n intr-vlr move. When n leves its RA nd enters into new RA the new VLR sends registrtion request to the HLR. HLR kees the identifiction of the into the OML of the old VLR. After erforming the s rofile udte y ccessing its dtse HLR sends the cknowledgement messge long with the OML of new VLR. We see tht HLR dtse is eing consulted three times. The first ccess is done for utting the s identifiction into the old VLR s OML, second time for udting the s current loction informtion nd third time for emtying the OML of new VLR, further the entries of this OML is sent ck with the cknowledgement. At the VLR side dtse is eing consulted twice, first for the registrtion of new nd second for de-registrtion of the entries sent y the HLR. Using equ. No. (3) nd (4) we cn derive the exressions for loction mngement schemes for multi HLR-VLR rchitectures s: We furthermore denote the costs of vrious oertions used in this nlysis s follows: 45
.2.4.6.8 Totl Cost Reltive Cost in Percentge..2.3.4.5.6.7.8.9.2.4.6.8 Interntionl Journl of Comuter Alictions (975 8887) Volume 5 No.2, August 22 First rt of equ. (5) is the cost when is not leving its resident-hlr. In this cse moility of is locl to the HLR; it my chnge its serving VLR nd cn enter into new VLR eing served y resident-hlr. The second rt of eqn (5) is the cost when n leves resident-hlr nd enters into new RA eing served y nother HLR clled serving-hlr. A cost of 2 is eing dded only to inform the resident-hlr tht its hs chnged its loction nd resident-hlr cknowledges the current serving-hlr. In rt 3 of this eqution, we re generlizing the movement of n when leves its resident-hlr nd enters into new serving-hlr nd then fter it gin chnges its serving-hlr to nother serving-hlr. A cost of 4 is eing dded to it ecuse of the following resons: Let n leves its resident-hlr nd enters into serving-hlr sy serving-hlr. Agin the sme leves this serving- HLR nd enters into nother serving-hlr sy serving-hlr2. In registrtion rocess of in serving-hlr2, it sends loction udte to resident-hlr (cost incurred is ), on recetion of this messge, resident-hlr udtes the loction informtion of nd sends loction cncelltion messge to serving-hlr (cost incurred is ). On recetion of this messge, the serving-hlr deletes the loction informtion of this. The serving-hlr cknowledges the resident-hlr out the loction cncelltion messge (cost incurred is ). Finlly the resident-hlr cknowledges the serving-hlr2 (cost incurred is ) nd loction registrtion tkes lce t the serving-hlr2. As ll HLRs re the sme level hence messge exchnge cost etween the two HLRs is T (, ) =. In eqn(6), ech rt cn e exlined in the sme mnner s in eqn(5). Only difference is tht in eqn (5) exlicit de-registrtion scheme is eing used while in eqn (6) grou de-registrtion scheme is eing used. 7. RESULTS In this section the numericl vlues of exlicit de-registrtion scheme nd grou de-registrtion scheme imlemented in the multi HLR-VLR rchitecture re evluted nd comred. Fig (6) nd (7) show the erformnce of loction udte schemes with R=5 nd R=3 resectively. 6 5 4 3 2 Loction Udte Cost VLR(grou) VLR(exlicit) Figures (8) & (9) show the reltive cost in ercentge (((M multihlr-vlr (grou) / (M multihlr-vlr (exlicit) )*) when R=5 nd 3. Proility shows tht the will reside in its resident-hlr. Proility nd indicte tht will e in serving-hlrs. When hs vlue then is not chnging its resident-hlr. Proility defines the s intr VLR move. If equls to then is not chnging its loction. When tends to, it mens s move is not locl. When nd re equl to, is in its resident-hlr nd not chnging its loction nd hence loction udte cost is. When nd equls to, it mens is not in its resident-hlr nd its move is not locl with resect to serving-hlr. When nd equl to, it shows the mximum degree of movement, in this cse grou de-registrtion scheme is 72.73% efficient thn the conventionl exlicit de-registrtion scheme when R=5 nd 85.7% in cse when R=3. 4 2 8 6 4 2 Proility of 's Movement 2 Loction Udte Cost Fig 7: Loction udte cost when R=3. 8 6 4 2 Fig 8: Reltive loction udte cost when R=5. VLR(grou) VLR(exlicit) Reltive Cost in Percentge Proility of 's Movement Reltive Cost Proility of 's Movement Fig 6: Loction udte cost when R=5. 46
Reltive Cost in Percentge..2.3.4.5.6.7.8.9 Interntionl Journl of Comuter Alictions (975 8887) Volume 5 No.2, August 22 2 8 6 4 2 Reltive Cost in Percentge Proility of 's Movement Fig9: Reltive loction udte cost when R=3. 8. CONCLUSION Conventionl rchitecture hs single HLR nd tht s why it suffers from cll misrouting nd ottleneck during ek lod. To remove this, severl conventionl rchitectures re grou together to form multi HLR rchitecture. In this scheme we store the user rofile in HLRs zone wise. This roch reduces the storge lod on HLR nd hence minimizes the ossiility of ottleneck. Now in this rchitecture even in high lod rorite informtion is fetched from HLRs nd we significntly minimize the ossiility of cll misrouting. Anlysis done in the lst section shows tht totl cost incurred into the loction mngement in the roosed multi HLR-VLR rchitecture using grou de-registrtion scheme is efficient thn the exlicit de-registrtion scheme. The roosed rchitecture is free from the rolem of ottleneck s we re not entirely [8] Reltive Cost relying on one HLR. We hve not ny stle entry of in ny VLR s we hve ssocited the de-registrtion rocess of the with its movement nd sved sufficient cost y imlementing grou de-registrtion scheme insted of conventionl exlicit de-registrtion. 9. REFERENCES [] Rjeev R. Kumr Trithi, Sudhir Agrwl nd Swti Tiwri, Performnce Anlysis of De-registrtion Strtegy in Personl Communiction Network, IJCA, vol.24-no., June 2. [2] Rjeev R. Kumr Trithi, Sudhir Agrwl nd Swti Tiwri, Modified HLR-VLR Loction Mngement Scheme in PCS Network, IJCA, vol.6-no.5, Set 2. [3] Z. Mo, C. Douligeris, A novel deregistrtion strtegy for moile network, IEEE VTS 275-2754, vol.6, Aug 22. [4] Hider Sf nd Smuel Pierre, A New Architecture for Imroving Loction mngement in PCS Network, Journl of Comuter Science (2):249-258, 25. [5] Srvl H. Singh, Sudhir Agrwl nd Vijy Kumr Performnce Evlution of Loction Mngement Techniques in PCS Networks, IJCA, vol.5, no-8, Fe- 2. [6] Gun-Chi Chen nd Suh-Yin Lee, Evlution of Distriuted nd Relicted HLR for Loction Mngement in PCS Network," Journl Of Informtion Science And Engineering 9, 85- (23). [7] George Couloris, Jen Dollimore nd Tim Kinderg, Distriuted Systems Concets nd Design, Person Eduction, 4e. 47