Proceedings of 2009 12th International Conference on Computer and Information Technology (ICCIT 2009) 21-23 December, 2009, Dhaka, Bangladesh

A New Hashing and Caching Approach for Reducing Call Delivery Cost and Location Server’s Load in Wireless Mobile Networks Md. Mohsin Ali, Md. Amjad Hossain, Md. Kowsar Hossain, G. M. Mashrur-E-Elahi, Md. Asadul Islam Department of Computer Science and Engineering (CSE) Khulna University of Engineering & Technology (KUET), Khulna – 9203, Bangladesh [email protected], [email protected], [email protected], [email protected], [email protected] Abstract This paper proposes a new approach for reducing average call delivery cost and location server’s load of wireless mobile networks. It uses caches whose up-todate information is responsible for dropping these costs and these caches are updated not only during call arrival moment from the calling Mobile Hosts (MHs) but also during call receiving moment to those MHs. To achieve load balancing among replicated Home Location Registers (HLRs), hashing technique is also used and this load is also affected by up-to-date cache information. The analytical model and experimental results show that our proposed method prepares the cache with up-to-date information more frequently with the increase of average call arrival rate as well as average call receiving rate. This increases probability of finding MH’s location as well as hit ratio of the cache. As a result, both the average call delivery cost and load on a particular HLR are minimized considerably than all other previous approaches.

(distributed databases) download data from HLR concerning current users within the VLR’s specific service areas. Two basic operations in location management are location update and call delivery. Location update is performed whenever an MH crosses the RA boundaries. Call delivery is the process of determining the serving VLR and the cell location of the called MH. It has been shown that average location update cost and call delivery cost can be reduced by caching and hashing [3], [4]. This paper proposes a new approach to improve the overall network performance in terms of average call delivery cost and individual location servers load by considering both the call arrival and receiving rate at the calling MH to update its cache information. The paper is organized as follows: Section II provides an overview of the related recent research work. Our proposed scheme is described in section III. Section IV gives the analytical model and comparison among different methods based on some experimental results. We provide a concluding remark in section V.

Keywords: Call arrival and receiving rate, call delivery cost, hashing and caching, Home Location Register (HLR), load balancing, Visitor Location Resister (VLR).

II. RELATED WORK There has been considerable amount of work done on location management to improve the overall performance of the wireless mobile networks [5], [7], [8]. The main issues of location management are to reduce the average call delivery cost and location update cost. To reduce these costs, the location of MH can be cached at its caller site where the majority of calls originate [5]. Recently, the distributed database architecture based on hierarchical organization is used for locating MH rather than using centralized database architecture [6].

I. INTRODUCTION Location management is concerned with the issues of tracking and finding MH in order to roaming in the network coverage area. To maintain the MHs’ locations, two types of location databases like HLR and Visitor Location Register (VLR) are commonly used in all wireless communications systems like basic IS-41 [1] and GSM [2]. These databases are organized in two levels of data hierarchy. In basic scheme of wireless mobile networks, network coverage is divided into cells. Each cell has a Base Transceiver Station (BTS) to which MH of the cell communicate through a wireless link. Each BTS is connected to a Mobile Switching Center (MSC) through a wired network. To facilitate the tracking of a moving MH, the wireless network is partitioned into many Registration Areas (RAs). Each RA includes tens or hundreds of cells. Each RA has a VLR servicing it and a VLR is designed to monitor only one RA. The HLR is the centralized database which contains the records of all users’ services in addition to location information for an entire network. The VLRs

978-1-4244-6283-4/09/$26.00 ©2009 IEEE

A number of replicated centralized databases, each of which is known as HLR, can be maintained to reduce the load on the database. To achieve the load balancing among the replicated HLRs, a protocol is proposed in [9]. Location updates and queries are multi-casted to subsets of location servers. These subsets vary with time and depends on the location of MH’s querying of MSC and load on the server. At least one common database can exist at each pair of subsets. Hashing technique is used for the construction of the subsets. In [4], a

61

HLR 0

HLR 1

VLR

HLR m-1

Network

PSTN

Cache

…

HLR 2

PSTN

MSC

MSC BS

LR

Cache MH

Cell

Fig. 1. Overall architecture of the wireless mobile networks with hashing and caching.

saving the information in the cache. During call arrival time, if that location information is found in the cache, then it is also checked to see whether it is up-to-date or not. The call is established in case of cache’s up-to-date information. If location information is obsolete or absent in the cache, then it is collected from the location server selected by means of hashing function and updated or saved it to the cache. In case of handoff, MH’s new location information is updated in one of the new location servers among m by applying hashing function. Then, this information is multi-casted to remaining m-1 location servers by the selected server. The flow chart in Fig. 2 describes our proposed approach.

hashing and caching scheme on location management is proposed to reduce average call delivery cost where only call arrival rate is used to cache the MHs location. Only one location server is chosen by using the hashing technique which simplifies the load balancing protocol and reduces the database operation. Fig. 1 shows the architecture of the wireless mobile networks which uses caches to store the location of MHs and hashing function for load balancing among replicated HLRs. III. PROPOSED APPROACH In our proposed approach, a set of location servers are used which are identical in nature. Hashing function is used to select one of these servers dynamically. Every MH, location server, and VLR has specific Mobile Host Identifier (MHID), Server Identifier (SERVERID), and VLR Identifier (VLRID), respectively. Assume, there are m location servers (0 to m-1) in the system. We use the function SERVERID = f (MHID, VLRID, m) = (MHID + VLRID) mod m to find the specific location server among m servers which contains the MH’s location identified as MHID.

IV. ANALYTICAL MODEL AND EXPERIMENTAL RESULTS We know that the database access cost and signaling cost mainly depend on the database query delay and signal transmission delay. We mainly focused on the call delivery cost because the location update cost remains almost the same as basic IS-41 scheme [1] and as hashing and caching scheme [4] if we use lazy caching. In IS-41 scheme, total three queries are required to make a call. The first query is at calling VLR (Dv), the second at HLR (Dh), and the third at called VLR (Dv). So, the total database access cost is 2Dv + Dh [4]. Fig. 3 shows that total signaling cost depends on the following things

The location information is also cached at VLR to reduce average call delivery cost and server’s load. The cached location information of an MH is updated not only during call arrival moment from the calling MH but also during call receiving moment to that calling MH. At first, every call at VLR is checked to identify whether it is arrival call, receiving call, or handoff. Then, the cache is checked for MHs’ location information during call arrival time as well as during call receiving time. If that information is available in the cache during call receiving time and up-to-date, then the cache remains unchanged. Otherwise, the location information in the cache is obsolete or unavailable. In this case, cache is updated by extracting calling MH’s location information from signaling message or by

1. 2. 3.

62

Two messages exchange between calling VLR and HLR (2Ch), Two messages exchange between HLR and called VLR (2Ch), and One Message exchanges between calling VLR and called VLR (Cv).

Start

Call arrival, or receiving or handoff?

Call receiving

No

MHs new location information is updated in that new server MH’s location information in cache?

Yes

No

Select new location server for MHs new location using hashing function

Call arrival

MH’s location information in cache?

Cache hit?

Handoff

No This server multicasts this information to remaining m-1 location servers

Yes

Yes

Call established

Yes

Cache hit? No

Update or save MH’s location information in cache

Come back to that location server

Find MHs location from server selected by hashing function

Go to VLR containing the address of called MH

Fig. 2. Flow chart of the proposed method.

So, the total signaling cost is 4Ch + Cv. As a result, the average call delivery cost of basic IS-41 scheme is Cb = 4Ch+ Cv+2Dv + Dh [4].

location information may not exists in the cache and then the total cost depends on the cost of one query access in the cache, the total cost of basic IS-41 scheme (Cb), and an update cost in the cache. Query access and update cost of the cache is considered negligible. Consider Cc as average call delivery cost of hashing and caching scheme, Cch as average call delivery cost if the location information is in the cache, p as cache hit ratio, and q as probability that the location information is in the cache. So, Cc is defined as follows [4].

.

In hashing and caching scheme [4], location information is stored in the cache at the moment of call arrival. If it is in the cache, then cache hit may occur and at that time total cost depends on the cost of one query access in the cache and one message exchanges between calling VLR and called VLR (Cv). Otherwise, cache miss occurs means the information is obsolete. At that time, the total cost depends on the cost of one query access in the cache, one message exchanges between calling VLR and called VLR (Cv), the total cost of basic IS-41 scheme (Cb) and an update cost in the cache. The

1

(1)

1

(2)

Where,

63

If

HLR

Ch

Ch

Ch

Cv

Calling VLR

, then (8) becomes (9)

Ch

Fig. 4 shows different average call delivery cost, obtained by varying average call arrival or receiving rate for three different values of

Called VLR

. It depicts that with

increasing average call arrival or receiving rate, average call delivery cost of basic IS-41 scheme remains constant, because it does not depend on average call arrival or receiving rate relating to cache. But, this cost for both hashing and caching and proposed scheme is

Fig. 3. The basic signaling process of the wireless mobile networks.

Assume, the call arrivals to MHs and MHs’ mobility follow the Poisson distribution. The mean call arrival rate is λ and the mean mobility rate is µm. LCMR is the Local Call to Mobility Ratio. According to [4], p is defined as follows

smaller for each of the three values of

since caches

are used to find out MH’s location to deliver call instead of finding out it in the location server with more penalties. However, for proposed method, this cost decreases significantly than that of the remaining methods with respect to . Because, we use not only call arrival rate but also call receiving rate for cache update considering both of them as equal. So, the cache is updated more frequently than hashing and caching scheme. As a result, cache hit ratio increases and the average call delivery cost decreases.

(3) According to [8],

and

. So, (3) is written as

follows (4)

The load performance of location server is measured by throughput of location servers. For basic IS-41 scheme, the HLR is the only location server. For measuring the load performance of this scheme, we define X as throughput and µ as HLR query rate (service rate). This scheme is modeled as an M/M/1 queuing system. The call arrival times of this system are independent and exponentially distributed with parameter . Service times are also exponentially distributed with parameter µ. A single server is used in this system. So, the system throughput is equal to .

From (3), it is shown that in hashing and caching scheme, p depends only on λ and µm, but in our proposed scheme, p depends not only on λ and µm, but also on the mean call receiving rate, . So, it can be defined as follows (5) If

, then (5) becomes

For hashing and caching scheme, we assume that there are m location servers. MSC sends requests to serveri, as the throughput of where i = 1,2,···, m. We define one location server, as call arrival rate at serveri, as as one of the location server’s service rate, and the probability of querying to serveri. We assume that . the query service rate of one location server is The server is selected uniformly using basic hashing

(6) as maximum call arrival rate, then If we consider q in hashing and caching scheme depends only on and . So, it can be defined as follows (7)

method. So,

is defined as

follows [4]

Consider, as maximum call receiving rate. In our proposed scheme, q depends on λ, , , and . So, for this scheme, it can be defined as follows

1

is defined as . Thus,

1 Now, (10) becomes

(8)

64

(10)

Cb Cc Ccr

Dh/Cv = 19 Dh/Cv = 10 Dh/Cv = 1

56 52 48

Average Call Delivery Cost

44 40 36 32 28 24 20 16 12 8 4 0 0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

Lamda (λ) Fig. 4. Performance comparison of basic IS-41, hashing and caching and proposed schemes based on average call delivery cost.

proposed scheme initially increases until a particular value of and then decreases with the increase of call arrival as well as call receiving rate for any value of m. Because, there are m location servers to serve these two schemes and there are caches to find out MH’s location instead of directly searching in the location servers. Moreover, in proposed scheme, server load decreases more than the remaining schemes because it updates the cache more frequently during both call arrival and call receiving moment at calling MH instead of only during call arrival moment at that MH like hashing and caching scheme. But, larger values of m similarly make the server load of these two schemes smaller.

(11) Where, (1-q) is the probability that the location information is not in the cache. As there are m servers, we can model the system as m number of M/M/1 queues. Now, the throughput of location serveri is as follows [4] (12) In our proposed scheme, we have considered both call arrival and receiving rate for cache update. So, will be the same as (12); but, q of this equation will be defined by (8) instead of (7).

V. CONCLUSION In this paper, we propose a new approach to improve the overall performance of the wireless mobile networks in terms of average call delivery cost and location server’s load. We show that hit ratio of the cache used for storing MHs location information increases as we consider both call arrival rate and call receiving rate for updating cache information. A hashing function is used for load balancing among location servers and this load depends on the up-to-date cache information. The analytical model and experimental results show that with the improvement of hit ratio, the probability of

Fig. 5 shows different server loads obtained with respect to average call arrival or receiving rate for three different values of m. It shows that with increasing average call arrival or receiving rate, the server load increases linearly in basic IS-41 scheme. This is because, there is only one location server to serve this scheme, no caching is used and all the calls are delivered by searching in the single server. On the other hand, the server load of hashing and caching and

65

Cb Cc Ccr

m= 1 m= 3 m = 10

6 5.5 5

Server Load

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

Lamda (λ) Fig. 5. Performance comparison of basic IS-41, hashing and caching and proposed schemes based on location servers load.

finding MHs location in cache also increases and thus average call delivery cost and location servers load are reduced significantly compared to all other previous methods. We are currently working with more appropriate hashing function for selecting one of the location servers with low congestion and minimized load.

[5]

[6]

REFERENCES [1] EIA/TIA. “Cellular radio-telecommunications intersystem operations,” Tech. Rep. IS-41 Revision B, EIA/TIA, December 1991. [2] M. Mouly and M.B. Pautet, “The GSM system for mobile communications,” 49 rue Louise Bruneau, Palaiseau, France, Telecom Publishing, January 1992. [3] Chang Woo Pyo, Jie Li, Hisao Kameda, and Xiaohua Jia, “Dynamic Location Management with Caching in Hierarchical Databases for Mobile Networks,” DNIS 2002, LNCS 2544, SpringerVerlag, vol. 2544, pp. 253–267, Berlin/Heidelberg, December 2002. [4] Weiping He and Athman Bouguettaya, “Using Hashing and Caching for Location Management in Wireless Mobile Systems,” MDM 2003, LNCS

[7]

[8] [9]

66

2574, Springer-Verlag, vol. 2574, pp. 335–339, Berlin/Heidelberg, January 2003. I. Akyildiz, J. McNair, J. Ho, H. Uzunalioglu, and W. Wang, “Mobility management in nextgeneration wireless systems,” In Proceedings of the IEEE, vol. 87, no. 8, pp. 1347–1384, August 1999. Y. Bejerano and I. Cidon, “An efficient mobility management strategy for personal communication systems,” In The Fourth Annual ACM/IEEE International Conference on Mobile Computing and Networking, ACM Publishers, pp. 215-222, Dallas, Texas, United States, October 1998. A. Bouguettaya, “On the construction of mobile database management systems,” In The Australian Workshop on Mobile Computing & Databases & Applications, Melbourne, Australia, February 1996. E. Pitoura and G. Samaras, “Data Management for Mobile Computing,” vol. 10, Kluwer Academic Publishers, 1998. R. Prakash and M. Singhal, “Dynamic hashing + quorum = efficient location management for mobile computing systems,” In Proceedings of the Sixteenth Annual ACM Symposium on Principles of Distributed Computing, ACM Press Publisher, pp. 291, August 1997.