Using Network Memory™ to Stop Server Sprawl and Improve Application Delivery

The Evolution of the Intelligent Enterprise Network.

BACKGROUND

Over the past decade, enterprise networks have evolved signifi cantly. Whereas once they simply routed packets based on destination address, today they can also make intelligent decisions based on load, traffi c patterns, applications, and even specifi c information acquired through packet inspection.

There is no doubt that application awareness has helped to increase the “intelligence” of enterprise networks. The tradeoff, however, comes in the form of increased complexity. This has created a point of diminishing returns, whereupon adding more application awareness to a network can actually hinder effi ciency, scale, and performance. (It can be argued that the primary reason that the Internet has been able to scale so successfully is because the network has remained completely independent from the applications that traverse it.)

Fortunately, increasing application awareness is not the only way to add intelligence to an enterprise network. In the case of Wide Area Networks (WANs), signifi cant gains can be made if prior to (or instead of) determining “how” and “where” information should be routed across the network, it is fi rst determined if the information needs to be sent at all. This is the function of Silver Peak’s Network Memory™ technology, a fundamental component of Silver Peak’s groundbreaking solution for Local Instance Networking (LIN).

THE NEW INTELLIGENT PARADIGM – SILVER PEAK NETWORK MEMORY™

All Silver Peak NX Series appliances are equipped with Network Memory™ technology. These devices sit in enterprise locations, such as branch offi ces and corporate headquarters, examining all information that is traveling in and out of a WAN. As packets fl ow through the NX Series appliances, Network Memory technology is used to “fi ngerprint” the information and store a local copy, or instance, of the data.

Network Memory™ uses pattern matching at the byte level. As packets arrive at an NX Series appliance, Network Memory™ computes fi ngerprints of the data contained within the packets, and checks to see whether these fi ngerprints match data that is stored in the local instance of the NX appliance at the destination location. If the remote NX appliance does contain the information, there is no need to resend it over the WAN. Instead, instructions are sent to deliver the data locally.

SOLVING TODAY’S IT DILEMMA - CENTRALIZED OR DISTRIBUTED?

Network Memory™ can revolutionize many aspects of the corporate network infrastructure. However, the biggest, and most immediate impact is to provide a compelling solution to one of IT’s persistent dilemmas: should branch infrastructure be centralized or distributed?

In a distributed implementation, email servers, fi le servers and databases are placed in each branch location. This typically provides the best possible performance to end users in these locations because the applications and their respective data are accessed via a Local Area Network (LAN). Unfortunately, this leads to “server sprawl”, which has many adverse affects. For one, managing this type of distributed infrastructure can be quite costly. That is because there are numerous devices to purchase, install, manage and repair, all of which are distributed across a variety of locations (many of which do not have IT personnel onsite). In addition, more managed devices means more software licenses to purchase and more day-to-day management activities to perform, such as patch updates and backups. Furthermore, storing mission critical data in branch offi ces exposes an enterprise to environmental and security risks that do not exist when servers are located within a secure data center that is equipped with proper power, cooling, and other environmental controls.

The alternative is to consolidate server infrastructure into a select number of data centers, which enables all maintenance, troubleshooting, security policy enforcement, backups and auditing to be performed centrally. This dramatically simplifies server management. In addition, server centralization reduces overall hardware costs and operating expenditures. Unfortunately, as most applications do not run well over a WAN, this deployment scenario often results in unacceptably poor performance.

Network Memory™ was designed to deliver the best of both worlds: the administration, security, and compliance benefi ts of centralized servers and storage, with the enhanced performance that comes with server distribution. This has the following benefits:

APPLICATION TRANSPARENCY

Network Memory™ operates with complete application transparency. It delivers data with 100% coherence and does not alter application semantics. All authentication, fi le and record locking functions are handled by the native application servers, enabling seamless deployment and operations.

This provides big performance gains across a broad range of applications, which include:

Email: When an email is sent to multiple employees in the same offi ce (e.g., using MAPI), the content traverses the WAN only once. Network Memory™ enables subsequent users to retrieve the content locally from the NX series appliance, which occurs at LAN speeds.

Users still authenticate with the central email server, and the master copy of the email still resides there. Network Memory™ simply eliminates the transfer of repetitive information across the WAN, which can provide a 10-100x improvement in email download times (depending on the size of attachments).

File Server: When a user requests a document from a remote fi le server (e.g., using CIFS), there is no need to transfer that information if other users in the same offi ce have already downloaded the fi le. Even when the fi le is not identical to the version that was previously downloaded, signifi cant performance improvements are realized by transporting only the incremental changes across the WAN.

Web: If a web application is generating dynamic pages (e.g., using HTTP), only delta information is transferred. For example, a SharePoint table with many rows could be updated by just transmitting the delta for the row, rather than the whole page.

Video streaming and Video on Demand: If several employees in an offi ce chose to watch the same video (e.g., a distance learning module or a taped CEO address), Network Memory™ eliminates the need to send multiple copies across the wide area network. This has the same advantage if they all are watching the video simultaneously, or at different times. Because Network Memory™ works at the network level, it supports UDP and RTP. This makes it ideal for applications that do not leverage TCP, like video streaming, IP multicast, real-time fi nancial applications, etc.

Software patch distribution and upgrades. Rather than downloading a separate software patch for each remote system that requires an upgrade, Network Memory™ enables this information to be stored once and delivered locally upon future requests. Because Network Memory™ is application independent, this works with whatever centralized patch management system IT chooses to employ.

Remote backups are a big consumer of bandwidth. Even for a full backup, Network Memory™ can drastically reduce the amount of time and bandwidth required by identifying content that is repeated in multiple fi les, across multiple directories, and in different user accounts. Once the fi rst backup is completed, future “full” backups are effectively reduced to “incremental backups” as far as WAN traffi c is concerned.

In addition to the above, Network Memory™ benefi ts almost all enterprise client/server implementations, such as ERP, CRM, and other software applications, including SAP, Siebel, Oracle, and Citrix. Network Memory™ can also be used to improve remote software development, where signifi cant traffi c can be generated checking-in and checking-out the same or modifi ed software code, and it can improve the performance of SQL or other database access, where successive aggregate queries request overlapping parts of the database

Byte level visibility enables Network Memory™ to work across multiple enterprise applications simultaneously. For instance, if a PDF fi le is copied to a fi le server using the CIFS protocol, downloaded from an Intranet using HTTP, and then emailed using SMTP, Network Memory™ treats the content the same, regardless of the application used. This provides a signifi cant benefi t over alternative solutions that perform application-specifi c enhancements.

ENTERPRISE SCALABILITY

Silver Peak’s Network Memory™ technology is designed to scale to hundreds of megabits per second (Mbps) of WAN bandwidth. This is possible because Network Memory™ operates at the network or IP level instead of the TCP fl ow level. In addition, Network Memory™ implements extremely fast pattern matching algorithms (fi ngerprinting decisions are made in microseconds), enabling it to process more information in a shorter period of time.

The Silver Peak solution scales to 155 Mbps in a single NX Series appliance, supporting many hundred of thousands of fl ows. In contrast, competing application acceleration solutions scale only to several thousand fl ows, limiting their effectiveness in a large distributed enterprise.

A NETWORK WIDE SOLUTION

Each Silver Peak NX Series appliance is constantly exchanging information about what’s contained in its local Network Memory with its peers, creating a network-wide picture of what information is stored where. This is similar to the way that routers use OSPF to exchange reachability information. In doing this, Network Memory™ works at the network rather than link level. In other words, its data reference is not limited to a particular WAN link – a local instance can be updated from any source. This becomes especially powerful in WANs that are confi gured as a mesh network, where branch offi ces communicate directly.
 
The same discovery algorithm makes Network Memory™ systems extremely effi cient to cluster. Each system in a cluster may maintain its own disk subsystem, but can quickly learn what information is contained in other member’s local instances.

SEAMLESS INTEGRATION AND OPERATIONS

Network Memory™ operates at the network layer of the ISO model, making it very easy to deploy and support on an ongoing basis. Installation is simple. There is no need to reconfi gure clients or servers (e.g., re-mapping share drives), enabling benefi ts to be realized without changing the way people work. Because Network Memory™ does not rely on TCP stream reconstruction to operate, NX Series appliances can be safely deployed in environments that employ asymmetric routing. (Other alternative approaches are often stymied by asymmetric routing, because they need to reconstruct TCP or even application level fl ows in order to operate). Furthermore, the Silver Peak solution operates in a completely complementary fashion to application front end equipment (also known as application delivery controllers). The Silver Peak solution also facilitates ongoing operations. The NX series appliances can be easily disabled for fault isolation and troubleshooting. In this case, the network operates exactly the same as it did prior to install. (Other, more invasive approaches such fi le proxy based WAFS devices can be very diffi cult to extract or disable, because they require ancillary changes to desktops and servers, which are diffi cult to reverse in the event of problems.)

THE WHOLE PICTURE – BUILDING A LOCAL INSTANCE NETWORK

Network memory™ is the cornerstone of the Silver Peak solution. When combined with advanced Quality of Service capabilities, adaptive tools for latency and loss mitigation, and innovative compression techniques, Silver Peak is delivering the industry’s premier platform for Local Instance Networking. The Silver Peak solution is the fi rst enterprisegrade platform for the re-centralization of branch offi ce server and storage infrastructure, enhancing enterprise-wide data security and compliance, while providing signifi cant performance improvements over existing application acceleration approaches. By improving the way that networks intelligently handle information, Silver Peak’s innovative Network Memory™ technology is changing the face of enterprise application delivery.

Bizi Arayın; Uzman kadromuz güvenlik taraması yapsın. Güvenlik konusunda risklerinizi belirlesin. Daha iyi bir güvenlik politikası için önerilerde bulunsun. Bize; +90 (216) 550 80 20 - 21 nolu telefondan arayarak yada, guvenlik@vedius.com adresine mail atarak ulaşabilirsiniz.

yada; Formu doldurun biz sizi arayalım Ad Soyad :   Firma/Görev :   Telefon :   E-posta :