The real issue is IP everywhere. Why shouldn’t storage be IP also?
Because it is not the right solution for every application and for every criticality (life-critical, mission critical, business critical, task critical, boss-critical). This is yet another one-size-fits-all false war driven by vendor hype similar to SAN vs NAS where I swear there are people on the verge on arguing that asynchronous mirroring is better than synchronous mirroring regardless of price-performance considerations.<g>
EMC continues its blue sky research on IP because it believe that it is the network transport layer of the future. It is not ready NOW, hoever, except for certain narrow applications that primarily trade off performance for costs. EMC has been working with CMNT on SRDF (synchronous mirroring software used for disaster recovery, data migration, etc) for several years now. They were actually selling SRDF over IP for about a year before they announced its general availability earlier this year. That announcement contained references to continuing work being done at MCI/WorldCom (FC core at the data centers/all-IP backbone network) and Excite (asynchronous mirroring and point in time copies).
........WorldCom Inc. is a believer. MCI, before merging with WorldCom, had worked over the past few years to rid itself of the multitude of dedicated links between data centers. Keeping Fibre Channel connections inside the data center, MCI consolidated the external links onto a single IP network—a move that cut support costs for per-megabyte transfers tenfold.
"Today, with the high-speed fiber optic links all over the country, and our entire backbone an IP network, it [made sense to] combine everything over the IP network," said Bob Oliver, the com pany's chief architect and strategist, in Colorado Springs, Colo. MCI is using EMC's SRDF (Symmetrix Remote Data Facility) over IP.
"[SRDF] was the first step. That allowed us to get rid of the proprietary, special networks," Oliver said. "At one point, we had literally 100 different networks. We had gateways and protocol translators. It was a mess.
zdnet.com
According to CMNT, some Fortune 500 companies are alrealdy using SRDF over IP alongside SRDF over dedicated lines to average down their telecommunication costs while the dotcoms and xSPs with all-IP networks are more willing to settle for asynchronous mirroring because it matches their cash burn stage and the criticality of the data they are accumulating. During the last quarter, CMNT booked $1.9M in revenues from EMC for SRDF over IP and $3+M in revenues from Compaq for SAN over WAN. CMNT has also submitted its layer 4 enhancements to the standards process.
Here's another article on this matter.
Storage over IP: More than superSAN
Newest idea in convergence has potential, but it faces some large hurdles
By Henry Baltazar, eWEEK
June 5, 2000 12:00 AM ET
As if we needed yet another group lobbying for our convergence upgrade dollars, a new cadre of vendors has popped up, aggressively preaching the benefits of pushing storage traffic through the IP network.
Like other fill-in-the-blank-over-IP vendors (video, voice and so forth), storage-over-IP companies are trying to capitalize on the powerful WAN IP backbones that are being set up within corporations and the ubiquitous Internet to act as carriers of storage traffic.
eWeek Labs has found that once you get past the hype, storage over IP does have the potential to extend the reach of SANs (storage area networks) further than previously imagined — but only if IT managers are selective in the type of applications they run over the network.
Storage-over-IP vendors, including Nishan Systems, of San Jose, Calif., are banking on these networks to provide clients with the ability to access data from all over the world, allowing SANs to grow far beyond the current 10-kilometer capabilities of Fibre Channel architectures.
Storage-over-IP convergence would also provide another major benefit: consolidation of management. IT managers would have to train staffers to use only one protocol—IP—and Fibre Channel would be limited to small local networks. Essentially, IT managers could devote their time and energy to maintaining one network—another very attractive selling point.
Furthermore, storage over IP would enable IT managers to take advantage of the aggressive performance growth of IP networking, specifically the 10G-bit Ethernet specification, which is expected to be available in about two years and will be several times faster than Fibre Channel.
[my comment: Which Fibre Channel? The 2G-bit version now making its way to the marketplace or the 4G-bit version expected to follow fairly quickly? And when 10G-bit Ethernet does increase available bandwidth on the LAN, what does one think is going to happen to applications and usage patterns that will prevent new types of congestion from happening again especially as the blend of alpha-numeric and rich media data becomes richer and more unpredictable? ]
Initial storage-over-IP products from vendors such as Nishan, Gadzoox Networks Inc. and CNT Corp. will be Fibre Channel-to-IP routers that will be used to connect Fibre Channel SANs. The first applications to run on these networks will primarily be remote backup and restore applications. In the event of a disaster, offline tapes at the mirror site will allow IT managers to quickly bring replacement sites online.
Challenges ahead
Latency is the biggest obstacle that storage-over-IP technology faces. Many applications that will utilize SANs assume the data they are trying to access is local, but Fibre Channel can place storage several kilometers away from servers.
It may be somewhat hard to believe, but the bottleneck that will hamper storage over IP in point-to-point connections (between company sites) will be the speed of light. Although packets traveling on fiber networks travel at the speed of light, a fair amount of latency should be expected if these packets have to travel great distances.
Distributed processing applications for high-transaction environments will not be a good starting place for IT managers looking to converge storage traffic on IP backbones because latency will prevent geographically dispersed storage subsystems connected by IP from maintaining cache coherence. In the event of a disaster or system failure, the threat of lost transactions is very real.
IT managers must also consider the impact that storage over IP will have on their bandwidth usage. Quality of service is a necessity for storage over IP, and it will be important for IT managers to ensure that their servers are not starved of precious storage resources.
An interesting point of confusion that is starting to develop deals with the nomenclature for the new storage-over-IP technology.
SoIP, which would be the logical abbreviation for storage over IP, is copyrighted by Nishan Systems. In contrast, CNT and its partners are leaning toward SAN over IP as their marketing vehicle. As the technologies of Nishan Systems and CNT develop, the two names may come to stand for vastly different technologies
zdnet.com |
